1 2018-05-23 Richard Biener <rguenther@suse.de>
3 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4 of fixed offset from memset VN.
6 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
8 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
10 (alloc_cand_and_find_basis): Initialize first_interp field.
11 (slsr_process_mul): Modify first_interp field.
12 (slsr_process_add): Likewise.
13 (slsr_process_cast): Modify first_interp field for each new
15 (slsr_process_copy): Likewise.
16 (dump_candidate): Dump first_interp field.
17 (replace_mult_candidate): Process all interpretations, not just
19 (replace_rhs_if_not_dup): Likewise.
20 (replace_one_candidate): Likewise.
22 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
24 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
26 (aarch64_needs_frame_chain): New function.
27 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
29 2018-05-23 Sudakshina Das <sudi.das@arm.com>
32 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
33 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
34 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
35 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
36 as true for strict-align.
37 (aarch64_can_inline_p): Perform checks even when callee has no
38 attributes to check for strict alignment.
39 * doc/extend.texi (AArch64 Function Attributes): Document
41 * doc/invoke.texi: (AArch64 Options): Likewise.
43 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
45 PR tree-optimization/85853
46 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
47 the handling of the root of the node to...
48 (vect_slp_analyze_node_operations_1): ...this new function,
49 and run the whole thing with the child nodes' def types
50 set according to their SLP node's def type.
52 2018-05-23 Richard Biener <rguenther@suse.de>
55 * tree-data-ref.c (create_runtime_alias_checks): Defer
56 and ignore overflow warnings.
58 2018-05-23 Yury Gribov <tetra2005@gmail.com>
60 PR tree-optimization/85822
61 * tree-vrp.c (is_masked_range_test): Fix handling of negative
64 2018-05-23 Richard Biener <rguenther@suse.de>
66 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
67 memset constants via native_interpret_expr.
69 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
72 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
74 (cgraph_node::create_alias): Likewise.
75 (cgraph_node::get_availability): Check ifunc_resolver instead
76 of looking up ifunc attribute.
77 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
78 * varasm.c (do_assemble_alias): Likewise.
79 (assemble_alias): Likewise.
80 (default_binds_local_p_3): Likewise.
81 * cgraph.h (cgraph_node): Add ifunc_resolver.
82 (cgraph_node::only_called_directly_or_aliased_p): Return false
84 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
86 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
87 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
88 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
89 instead of looking up ifunc attribute.
91 2018-05-22 Luis Machado <luis.machado@linaro.org>
93 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
95 2018-05-22 Martin Sebor <msebor@redhat.com>
98 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
99 only when expasion succeeds.
100 (expand_builtin_strcmp): Same.
101 (expand_builtin_strncmp): Same.
103 2018-05-22 Martin Sebor <msebor@redhat.com>
105 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
107 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
108 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
110 * config/aarch64/aarch64-ldpstp.md: Replace uses of
111 aarch64_mem_pair_operand with memory_operand and delete operand swapping
113 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
114 Add check for legitimate_address.
115 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
116 (aarch64_swap_ldrstr_operands): New.
117 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
120 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
121 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
123 * config/aarch64/aarch64.md: New patterns to generate stp
125 (store_pair_sw, store_pair_dw): New patterns to generate stp for
126 single words and double words.
127 (load_pair_sw, load_pair_dw): Likewise.
128 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
130 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
132 * config/aarch64/aarch64-ldpstp.md: Modify peephole
133 for different mode ldpstp and add peephole for merged zero stores.
135 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
137 (aarch64_gen_store_pair): Rename calls to match new patterns.
138 (aarch64_gen_load_pair): Rename calls to match new patterns.
139 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
140 (load_pair<DREG:mode><DREG2:mode>): ... This.
141 (store_pair<mode>): Rename to...
142 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
143 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
145 (V_INT_EQUIV): Handle SImode.
146 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
149 2018-05-22 Martin Sebor <msebor@redhat.com>
152 * calls.c (maybe_warn_nonstring_arg): Use string length to set
153 or ajust the presumed bound on an operation to avoid unnecessary
156 2018-05-22 Martin Sebor <msebor@redhat.com>
158 PR tree-optimization/85826
159 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
160 assuming that a DECL necesarily has a constant size.
162 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
165 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
167 2018-05-22 Richard Biener <rguenther@suse.de>
169 PR tree-optimization/85834
170 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
171 non-constant and non-zero memset arguments.
173 2018-05-22 Martin Liska <mliska@suse.cz>
176 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
178 2018-05-22 Richard Biener <rguenther@suse.de>
180 PR tree-optimization/85863
181 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
182 comparisons when vectype is specified.
183 (vectorizable_condition): Do not specify vectype for
184 vect_is_simple_cond when SLP vectorizing.
186 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
189 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
190 define __ibm128 as long double.
191 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
192 as a distinct type when IEEE 128-bit support is enabled.
193 (init_float128_ieee): Fix up conversions between IFmode and IEEE
194 128-bit types to use the correct functions.
195 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
196 convert between 128-bit floating point types that have different
197 modes but the same representation, instead of using gen_lowpart to
199 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
201 (IFKF_reg): New attributes to give the register constraints for
203 (extend<mode>tf2_internal): New insns to mark an explicit
204 conversion between 128-bit floating point types that have a
205 different mode but share the same representation.
207 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
209 PR tree-optimization/85814
210 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
211 a null return from get_strinfo when unsharing the next
212 strinfo in the chain.
214 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
217 * varasm.c (weak_finish): Clean up weak_decls.
219 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
221 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
222 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
223 UNSPEC_UADALP values.
224 * config/aarch64/iterators.md (ABAL): New int iterator.
227 (sur): Add mappings for the above.
228 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
230 (aarch64_<sur>abal<mode>_4): Likewise.
231 (aarch64_<sur>adalp<mode>_3): Likewise.
232 (<sur>sadv16qi): New define_expand.
234 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
236 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
237 (*movdf_internal): Ditto.
238 (*rcpsf2_sse): Ditto.
239 (*rsqrtsf2_sse): Ditto.
240 (*sqrt<mode>2_sse): Ditto.
242 2018-05-21 Tamar Christina <tamar.christina@arm.com>
244 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
246 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
247 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
248 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
249 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
251 * config/aarch64/arm_neon.h: Likewise.
252 * config/aarch64/iterators.md (VQ_I): New.
254 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
256 * config.gcc: Add arc/t-multilib-linux to tmake_file for
258 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
261 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
263 * config/nds32/constraints.md (S): New constraint.
264 * config/nds32/nds32.md (call_internal): Use constraint S.
265 (call_value_internal): Likewise.
266 (sibcall_internal): Likewise.
267 (sibcall_value_internal): Likewise.
269 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
270 Chung-Ju Wu <jasonwucj@gmail.com>
272 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
275 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
276 Chung-Ju Wu <jasonwucj@gmail.com>
278 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
279 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
280 (nds32_rtx_costs_impl): Simplify.
281 (nds32_address_cost_impl): Simplify.
282 (nds32_init_rtx_costs): New function.
283 (nds32_rtx_costs_speed_prefer): Likewise.
284 (nds32_rtx_costs_size_prefer): Likewise.
285 (nds32_address_cost_speed_prefer): Likewise.
286 (nds32_address_cost_speed_fwprop): Likewise.
287 (nds32_address_cost_size_prefer): Likewise.
288 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
289 * config/nds32/nds32.c (nds32_option_override): Use
290 nds32_init_rtx_costs function.
292 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
294 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
295 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
296 (TARGET_PIPELINE_N8): Likewise.
297 (TARGET_PIPELINE_N10): Likewise.
298 (TARGET_PIPELINE_N13): Likewise.
299 (TARGET_PIPELINE_GRAYWOLF): Likewise.
301 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
303 * config/nds32/nds32-fpu.md: Update copyright year.
305 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
307 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
309 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
311 * config/nds32/nds32.c
312 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
313 * config/nds32/nds32.opt (minline-asm-r15): New option.
315 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
317 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
319 * config/nds32/nds32.md (abssi2): New pattern.
321 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
323 * config/i386/i386.md (rex64namesuffix): New mode attribute.
324 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
325 Merge insn pattern from sse_cvtsi2ss<round_name> and
326 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
327 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
328 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
329 using SWI48 mode iterator.
330 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
331 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
332 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
333 pattern from sse_cvttss2si<round_saeonly_name>
334 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
335 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
336 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
337 using SWI48 mode iterator.
338 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
339 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
340 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
341 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
342 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
343 using SWI48 mode iterator.
344 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
345 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
346 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
347 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
348 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
350 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
351 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
352 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
353 pattern from sse_cvttsd2si<round_saeonly_name>
354 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
356 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
358 * config/nds32/nds32-md-auxiliary.c
359 (nds32_valid_smw_lwm_base_p): Refine.
360 (nds32_output_smw_single_word): Refine.
361 (nds32_output_smw_double_word): New.
362 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
364 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
366 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
367 (nds32_output_stack_pop): Refine.
368 (nds32_expand_unaligned_load): Refine.
369 (nds32_expand_unaligned_store): Refine.
371 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
372 Chung-Ju Wu <jasonwucj@gmail.com>
374 * config/nds32/constants.md: Add TP_REGNUM constant.
375 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
376 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
378 * config/nds32/nds32-doubleword.md: Consider flag_pic.
379 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
380 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
381 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
382 and PIC code generation.
383 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
385 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
387 * config/nds32/nds32.md: Support TLS and PIC.
388 * config/nds32/nds32.c: Support TLS and PIC.
389 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
390 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
393 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
395 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
398 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
399 Chung-Ju Wu <jasonwucj@gmail.com>
401 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
402 * config/nds32/nds32-md-auxiliary.c
403 (symbolic_reference_mentioned_p): New.
404 (nds32_legitimize_ict_address): New.
405 (nds32_expand_ict_move): New.
406 (nds32_indirect_call_referenced_p): New.
407 (nds32_symbol_binds_local_p): Delete.
408 (nds32_long_call_p): Modify.
409 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
410 * config/nds32/nds32-protos.h
411 (symbolic_reference_mentioned_p): Declare.
412 (nds32_legitimize_ict_address): Declare.
413 (nds32_expand_ict_move): Declare.
414 (nds32_indirect_call_referenced_p): Declare.
415 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
416 (nds32_relax_group): Use nds32_ict_const_p as condition.
417 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
418 (nds32_asm_file_start): Output ict_model directive in asm code.
419 (nds32_legitimate_address_p): Consider indirect call.
420 (nds32_print_operand): Consider indirect call.
421 (nds32_print_operand_address): Consider indirect call.
422 (nds32_insert_attributes): Handle "indirect_call" attribute.
423 (TARGET_LEGITIMATE_ADDRESS_P): Define.
424 (TARGET_LEGITIMATE_CONSTANT_P): Define.
425 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
426 (TARGET_DELEGITIMIZE_ADDRESS): Define.
427 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
428 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
429 (TARGET_ICT_MODEL_SMALL): Define.
430 (TARGET_ICT_MODEL_LARGE): Define.
431 * config/nds32/nds32.md (movsi): Consider ict model.
432 (call, call_value): Consider ict model.
433 (sibcall, sibcall_value): Consider ict model.
434 * config/nds32/nds32.opt (mict-model): New option.
435 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
438 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
439 Monk Chiang <sh.chiang04@gmail.com>
440 Jim Wilson <jimw@sifive.com>
442 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
443 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
444 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
445 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
446 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
447 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
448 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
449 compute save_libcall_adjustment properly.
450 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
451 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
452 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
453 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
454 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
455 (ABI_SPEC): Handle mabi=ilp32e.
456 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
458 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
459 <-march>: Add rv32e as an example.
461 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
464 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
465 (intra_create_variable_infos): Handle C++ constructors.
467 2018-05-18 Martin Liska <mliska@suse.cz>
469 * passes.def: Remove a redundant pass.
471 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
474 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
476 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
478 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
480 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
482 (armv2, armv2a, armv3, armv3m): Delete architectures.
483 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
484 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
485 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
487 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
488 (*mulsidi3adddi): Likewise.
489 (mulsidi3): Likewise.
490 (*mulsidi3_nov6): Likewise.
491 (umulsidi3): Likewise.
492 (umulsidi3_nov6): Likewise.
493 (umaddsidi4): Likewise.
494 (*umulsidi3adddi): Likewise.
495 (smulsi3_highpart): Likewise.
496 (*smulsi3_highpart_nov6): Likewise.
497 (umulsi3_highpart): Likewise.
498 (*umulsi3_highpart_nov6): Likewise.
499 * config/arm/arm.h (arm_arch3m): Delete.
500 * config/arm/arm.c (arm_arch3m): Delete.
501 (arm_option_override_internal): Update armv3-related comment.
502 (arm_configure_build_target): Delete use of isa_bit_mode32.
503 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
504 (arm_rtx_costs_internal): Delete check of arm_arch3m.
505 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
508 * config/arm/arm-protos.h (arm_arch3m): Delete.
509 * config/arm/arm-tables.opt: Regenerate.
510 * config/arm/arm-tune.md: Likewise.
511 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
512 deleted architectures.
514 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
516 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
517 (armv5t, armv5te): New features.
518 (ARMv5, ARMv5e): Delete fgroups.
519 (ARMv5t, ARMv5te): Adjust for above changes.
521 (armv5, armv5e): Delete arches.
522 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
524 (*call_reg_arm): Likewise.
525 (*call_value_reg_armv5): Likewise.
526 (*call_value_reg_arm): Likewise.
527 (*call_symbol): Likewise.
528 (*call_value_symbol): Likewise.
529 (*sibcall_insn): Likewise.
530 (*sibcall_value_insn): Likewise.
532 (prefetch): Likewise.
533 (define_split and define_peephole2 dependent on arm_arch5):
535 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
537 (TARGET_ARM_QBIT): Likewise.
538 (TARGET_DSP_MULTIPLY): Likewise.
539 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
540 (arm_arch5, arm_arch5e): Delete.
541 (arm_arch5t, arm_arch5te): Declare.
542 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
543 (arm_arch5t): Declare.
544 (arm_option_reconfigure_globals): Update for the above.
545 (arm_options_perform_arch_sanity_checks): Update comment, replace
546 use of arm_arch5 with arm_arch5t.
547 (use_return_insn): Likewise.
548 (arm_emit_call_insn): Likewise.
549 (output_return_instruction): Likewise.
550 (arm_final_prescan_insn): Likewise.
551 (arm_coproc_builtin_available): Likewise.
552 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
553 arm_arch5e with arm_arch5t and arm_arch5te.
554 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
555 (arm_arch5t, arm_arch5te): Declare.
556 * config/arm/arm-tables.opt: Regenerate.
557 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
558 * config/arm/t-multilib: Likewise.
559 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
560 instead of arm_arch5.
561 (*call_reg_thumb1): Likewise.
562 (*call_value_reg_thumb1_v5): Likewise.
563 (*call_value_reg_thumb1): Likewise.
564 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
566 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
568 2018-05-18 Martin Liska <mliska@suse.cz>
570 PR gcov-profile/84846
571 * doc/gcov.texi: Document -t option of gcov tool.
573 2018-05-18 Martin Liska <mliska@suse.cz>
575 PR gcov-profile/84846
576 * gcov.c (print_usage): Add new -t option.
577 (process_args): Handle the option.
578 (generate_results): Use stdout as output when requested by
581 2018-05-18 Martin Liska <mliska@suse.cz>
583 PR gcov-profile/84846
584 * coverage.c (coverage_init): Write PWD to .gcno file.
585 * doc/gcov.texi: Document how working directory is printed.
586 * gcov-dump.c (dump_gcov_file): Print PWD.
587 * gcov.c (output_intermediate_file): Likewise.
588 (read_graph_file): Read PWD string.
589 (output_lines): Print PWD.
591 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
594 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
595 for retval and return false if all args to phi are zero.
597 2018-05-18 Richard Biener <rguenther@suse.de>
599 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
601 (evrp_dom_walker::before_dom_children): Call it.
603 2018-05-18 Richard Biener <rguenther@suse.de>
605 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
606 results when processing array refs with variable index.
608 2018-05-18 Toon Moene <toon@moene.org>
610 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
611 directly after that of -floop-interchange. Indicate that both
612 options are enabled by default when specifying -O3.
614 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
616 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
617 iterator. Delete separate integer-mode vec_set<mode> expander.
618 (aarch64_simd_vec_setv2di): Delete.
619 (vec_setv2di): Delete.
620 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
621 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
622 the "w, r" alternative.
624 2018-05-18 Martin Liska <mliska@suse.cz>
626 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
627 * tree-pass.h (make_pass_lower_switch_O0): New function.
628 * tree-switch-conversion.c (node_has_low_bound): Remove.
629 (node_has_high_bound): Likewise.
630 (node_is_bounded): Likewise.
631 (class pass_lower_switch): Make it a template type and create
633 (pass_lower_switch::execute): Add template argument.
634 (make_pass_lower_switch): New function.
635 (make_pass_lower_switch_O0): New function.
636 (do_jump_if_equal): Remove.
637 (emit_case_nodes): Simplify to just handle all 3 cases and leave
638 all the hard work to tree optimization passes.
640 2018-05-18 Martin Liska <mliska@suse.cz>
642 * dbgcnt.c (limit_low): Renamed from limit.
643 (limit_high): New variable.
644 (dbg_cnt_is_enabled): Check for upper limit.
645 (dbg_cnt): Adjust dumping.
646 (dbg_cnt_set_limit_by_index): Add new argument for high
648 (dbg_cnt_set_limit_by_name): Likewise.
649 (dbg_cnt_process_single_pair): Parse new format.
650 (dbg_cnt_process_opt): Use strtok.
651 (dbg_cnt_list_all_counters): Remove 'value' and add
653 * doc/invoke.texi: Document changes.
655 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
657 * doc/sourcebuild.texi (scalar_all_fma): Document.
658 * tree.def (FMA_EXPR): Delete.
659 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
660 * internal-fn.c (ternary_direct): New macro.
661 (expand_ternary_optab_fn): Likewise.
662 (direct_ternary_optab_supported_p): Likewise.
663 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
664 * builtins.c (fold_builtin_fma): Delete.
665 (fold_builtin_3): Don't call it.
666 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
667 * expr.c (expand_expr_real_2): Likewise.
668 * fold-const.c (operand_equal_p): Likewise.
669 (fold_ternary_loc): Likewise.
670 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
671 * gimple.c (DEFTREECODE): Likewise.
672 * gimplify.c (gimplify_expr): Likewise.
673 * optabs-tree.c (optab_for_tree_code): Likewise.
674 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
675 * tree-eh.c (operation_could_trap_p): Likewise.
676 (stmt_could_throw_1_p): Likewise.
677 * tree-inline.c (estimate_operator_cost): Likewise.
678 * tree-pretty-print.c (dump_generic_node): Likewise.
679 (op_code_prio): Likewise.
680 * tree-ssa-loop-im.c (stmt_cost): Likewise.
681 * tree-ssa-operands.c (get_expr_operands): Likewise.
682 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
683 * fold-const-call.h (fold_fma): Delete.
684 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
685 CFN_FNMA and CFN_FNMS.
687 * genmatch.c (combined_fn): New enum.
688 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
689 (commutative_op): New function.
690 (commutate): Use it. Handle more than 2 operands.
691 (dt_operand::gen_gimple_expr): Use commutative_op.
692 (parser::parse_expr): Allow :c to be used with non-binary
693 operators if the commutative operand is known.
694 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
695 CFN_FMS, CFN_FNMA and CFN_FNMS.
696 (backprop::process_assign_use): Remove FMA_EXPR handling.
697 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
698 (gen_hsa_fma): New function.
699 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
700 IFN_FNMA and IFN_FNMS.
701 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
702 * gimple-fold.h (follow_all_ssa_edges): Declare.
703 * gimple-fold.c (follow_all_ssa_edges): New function.
704 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
705 gimple_build interface and use follow_all_ssa_edges to fold the result.
706 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
707 instead of checking for optabs directly.
708 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
709 rather than FMA_EXPRs.
710 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
711 call to IFN_FMA instead of an FMA_EXPR.
713 2018-05-17 Jim Wilson <jimw@sifive.com>
715 * expr.c (do_tablejump): When converting index to Pmode, if we have a
716 sign extended promoted subreg, and the range does not have the sign bit
717 set, then do a sign extend.
719 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
720 test, check for sign extended subreg and/or constant operands, and
721 do a sign extend in that case.
723 2018-05-17 Steve Ellcey <sellcey@cavium.com>
725 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
726 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
728 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
729 Change logics_shift_reg to logics_shift_imm.
730 (thunderx2t99_fp_loadpair_basic): Delete.
731 (thunderx2t99_fp_storepair_basic): Delete.
732 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
733 (thunderx2t99_asimd_polynomial): Delete.
734 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
735 and neon_fp_mul_d_scalar_q.
736 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
737 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
738 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
739 (thunderx2t99_asimd_lut): Add missing tbl types.
740 (thunderx2t99_asimd_ext): Delete.
741 (thunderx2t99_asimd_load1_1_mult): Delete.
742 (thunderx2t99_asimd_load1_2_mult): Delete.
743 (thunderx2t99_asimd_load1_ldp): New.
744 (thunderx2t99_asimd_load1): New.
745 (thunderx2t99_asimd_load2): Add missing *load2* types.
746 (thunderx2t99_asimd_load3): New.
747 (thunderx2t99_asimd_load4): New.
748 (thunderx2t99_asimd_store1_1_mult): Delete.
749 (thunderx2t99_asimd_store1_2_mult): Delete.
750 (thunderx2t99_asimd_store2_mult): Delete.
751 (thunderx2t99_asimd_store2_onelane): Delete.
752 (thunderx2t99_asimd_store_stp): New.
753 (thunderx2t99_asimd_store1): New.
754 (thunderx2t99_asimd_store2): New.
755 (thunderx2t99_asimd_store3): New.
756 (thunderx2t99_asimd_store4): New.
758 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
760 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
761 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
763 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
764 Segher Boessenkool <segher@kernel.crashing.org>
767 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
770 2018-05-17 Richard Biener <rguenther@suse.de>
772 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
773 for pruning loop and prune defs feeding only already visited PHIs.
775 2018-05-17 Richard Biener <rguenther@suse.de>
777 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
779 2018-05-17 Bin Cheng <bin.cheng@arm.com>
780 Richard Biener <rguenther@suse.de>
782 PR tree-optimization/85793
783 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
784 for VMAT_ELEMENTWISE.
786 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
788 * internal-fn.h (lookup_internal_fn): Declare
789 * internal-fn.c (lookup_internal_fn): New function.
790 * gimple.c (gimple_build_call_from_tree): Handle calls to
792 * gimple-pretty-print.c (dump_gimple_call): Print "." before
793 internal function names.
794 * tree-pretty-print.c (dump_generic_node): Likewise.
795 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
797 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
799 * gimple-fold.h (gimple_build): Make the function forms take
800 combined_fn rather than built_in_function.
801 (gimple_simplify): Likewise.
802 * gimple-match-head.c (gimple_simplify): Likewise.
803 * gimple-fold.c (gimple_build): Likewise.
804 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
805 rather than gimple_build_call_internal.
806 (get_initial_defs_for_reduction): Likewise.
807 (vect_create_epilog_for_reduction): Likewise.
808 (vectorizable_live_operation): Likewise.
810 2018-05-17 Martin Liska <mliska@suse.cz>
812 * gimple-ssa-sprintf.c (format_directive): Do not use
813 space in between 'G_' and '('.
815 2018-05-17 Jakub Jelinek <jakub@redhat.com>
818 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
819 even if the mask is not all ones.
822 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
824 (ix86_gimple_fold_builtin): Likewise.
827 * config/i386/i386.c: Include tree-vector-builder.h.
828 (ix86_vector_shift_count): New function.
829 (ix86_fold_builtin): Fold shift builtins by scalar count.
830 (ix86_gimple_fold_builtin): Likewise.
832 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
833 _mm512_setzero): New intrinsics.
835 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
836 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
838 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
839 code generation for cases where splatting a value is not useful.
840 * simplify-rtx.c (simplify_ternary_operation): Simplify
841 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
842 mode to a vec_concat.
844 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
846 * config.gcc: Support "goldmont-plus".
847 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
849 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
850 PROCESSOR_GOLDMONT_PLUS.
851 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
852 (processor_target_table): Add "goldmont-plus".
853 (PTA_GOLDMONT_PLUS): Define.
854 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
855 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
856 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
857 (fold_builtin_cpu): Add "goldmont-plus".
858 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
859 (ix86_option_override_internal): Add "goldmont-plus".
860 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
861 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
862 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
863 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
865 2018-05-17 Richard Biener <rguenther@suse.de>
867 PR tree-optimization/85757
868 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
869 remove defs that only feed that PHI from further processing.
871 2018-05-16 Jim Wilson <jimw@sifive.com>
873 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
875 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
877 2018-05-16 Mark Wielaard <mark@klomp.org>
879 * dwarf2out.c (count_index_strings): New function.
880 (output_indirect_strings): Call count_index_strings and generate
881 header for dwarf_version >= 5.
883 2018-05-16 Mark Wielaard <mark@klomp.org>
885 * dwarf2out.c (dwarf_FORM): New function.
886 (set_indirect_string): Use dwarf_FORM.
887 (reset_indirect_string): Likewise.
888 (size_of_die): Likewise.
889 (value_format): Likewise.
890 (output_die): Likewise.
891 (add_skeleton_AT_string): Likewise.
892 (output_macinfo_op): Likewise.
893 (index_string): Likewise.
894 (output_index_string_offset): Likewise.
895 (output_index_string): Likewise.
896 (count_index_strings): Likewise.
898 2018-05-16 Carl Love <cel@us.ibm.com>
900 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
901 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
903 2018-05-16 Martin Jambor <mjambor@suse.cz>
905 * ipa-prop.c (ipa_free_all_edge_args): Remove.
906 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
908 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
910 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
911 (fnma<mode>4): Likewise.
912 (fms<mode>4): Likewise.
913 (fnms<mode>4): Likewise.
914 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
915 (aarch64_fnma<mode>4): Likewise.
916 (aarch64_fms<mode>4): Likewise.
917 (aarch64_fnms<mode>4): Likewise.
918 (aarch64_fnmadd<mode>4): Likewise.
920 2018-05-16 Jason Merrill <jason@redhat.com>
922 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
924 2018-05-16 Richard Biener <rguenther@suse.de>
926 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
927 (dump_stmt_cost): Declare.
928 (add_stmt_cost): Dump cost we add.
929 (add_stmt_costs): New function.
930 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
932 (vect_analyze_stmt): Adjust prototype.
933 (vectorizable_condition): Likewise.
934 (vectorizable_live_operation): Likewise.
935 (vectorizable_reduction): Likewise.
936 (vectorizable_induction): Likewise.
937 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
938 cost vector to pass to vectorizable_ and record afterwards.
939 (vect_model_reduction_cost): Take cost vector argument and adjust.
940 (vect_model_induction_cost): Likewise.
941 (vectorizable_reduction): Likewise.
942 (vectorizable_induction): Likewise.
943 (vectorizable_live_operation): Likewise.
944 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
945 SLP_TREE_NUMBER_OF_VEC_STMTS.
946 (vect_analyze_slp_cost_1): Remove.
947 (vect_analyze_slp_cost): Likewise.
948 (vect_slp_analyze_node_operations): Take visited args and
949 a target cost vector. Avoid processing already visited stmt sets.
950 (vect_slp_analyze_operations): Use a local cost vector to gather
951 costs and register those of non-discarded instances.
952 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
953 (vect_schedule_slp_instance): Remove copying of
954 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
956 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
957 adding cost. Record cost entry location.
958 (vect_prologue_cost_for_slp_op): Function to compute cost of
959 a constant or invariant generated for SLP vect in the prologue,
960 split out from vect_analyze_slp_cost_1.
961 (vect_model_simple_cost): Make static. Adjust for SLP costing.
962 (vect_model_promotion_demotion_cost): Likewise.
963 (vect_model_store_cost): Likewise, make static.
964 (vect_model_load_cost): Likewise.
965 (vectorizable_bswap): Add cost vector arg and adjust.
966 (vectorizable_call): Likewise.
967 (vectorizable_simd_clone_call): Likewise.
968 (vectorizable_conversion): Likewise.
969 (vectorizable_assignment): Likewise.
970 (vectorizable_shift): Likewise.
971 (vectorizable_operation): Likewise.
972 (vectorizable_store): Likewise.
973 (vectorizable_load): Likewise.
974 (vectorizable_condition): Likewise.
975 (vectorizable_comparison): Likewise.
976 (can_vectorize_live_stmts): Likewise.
977 (vect_analyze_stmt): Likewise.
978 (vect_transform_stmt): Adjust calls to vectorizable_*.
979 * tree-vectorizer.c: Include gimple-pretty-print.h.
980 (dump_stmt_cost): New function.
982 2018-05-16 Richard Biener <rguenther@suse.de>
984 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
985 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
986 * tree-ssa-dse.c: Include tree-ssa-loop.h.
987 (check_name): New callback.
988 (dse_classify_store): Track cycles via a visited bitmap of PHI
989 defs and simplify handling of in-loop and across loop dead stores
990 and properly fail for loop-variant refs. Handle byte-tracking with
991 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
994 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
996 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
997 (vect_get_mask_type_for_stmt): Likewise.
998 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1000 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1001 to determine the statement's vector type and the vector type that
1002 should be used for calculating nunits. Deal with cases in which
1003 the type has to be deferred.
1004 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1005 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1006 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1007 (vect_determine_vf_for_stmt): New functions, split out from...
1008 (vect_determine_vectorization_factor): ...here.
1009 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1010 (vect_get_mask_type_for_stmt): New functions, split out from
1011 vect_determine_vectorization_factor.
1013 2018-05-16 Richard Biener <rguenther@suse.de>
1015 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1016 verify the [VEC_]COND_EXPR embedded comparison.
1018 2018-05-15 Martin Sebor <msebor@redhat.com>
1020 PR tree-optimization/85753
1021 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1022 RECORD_TYPE in addition to ARRAY_TYPE.
1024 2018-05-15 Martin Sebor <msebor@redhat.com>
1027 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1029 2018-05-15 Richard Biener <rguenther@suse.de>
1031 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1032 add by_clobber_p one. Change algorithm to collect all defs
1033 representing uses we need to walk and try reducing them to
1034 a single one before failing.
1035 (dse_dom_walker::dse_optimize_stmt): Adjust.
1037 2018-05-13 Mark Wielaard <mark@klomp.org>
1039 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1040 (size_of_loc_descr): Likewise.
1041 (output_loc_operands): Likewise.
1042 (output_loc_operands_raw): Likewise.
1043 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1044 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1045 (hash_loc_operands): Likewise.
1046 (compare_loc_operands): Likewise.
1048 2018-05-14 Mark Wielaard <mark@klomp.org>
1050 * dwarf2out.c (count_index_addrs): New function.
1051 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1053 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1055 PR tree-optimization/83648
1056 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1057 return value as malloc candidate.
1059 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1062 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1063 as true in call to suggest_attribute.
1065 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1067 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1068 -mreadonly-in-sdata.
1070 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1072 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1074 (aarch64_crypto_aesd_fused): Likewise.
1076 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1078 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1079 (movsi_aarch64): Likewise.
1080 (load_pairsi): Likewise.
1081 (load_pairdi): Likewise.
1082 (store_pairsi): Likewise.
1083 (store_pairdi): Likewise.
1084 (load_pairsf): Likewise.
1085 (load_pairdf): Likewise.
1086 (store_pairsf): Likewise.
1087 (store_pairdf): Likewise.
1088 (zero_extend): Likewise.
1089 (trunc): Swap alternatives.
1090 (fcvt_target): Add '?' to prefer w over r.
1092 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1095 * config/i386/i386.md: Disallow non-commutative arithmetics in
1096 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1097 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1098 in the peephole2 before it.
1100 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1102 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1103 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1104 (ix86_handle_option): Handle -mcldemote.
1105 * config.gcc: New header.
1106 * config/i386/cldemoteintrin.h: New file.
1107 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1108 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1110 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1111 OPTION_MASK_ISA_CLDEMOTE.
1112 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1113 (ix86_valid_target_attribute_inner_p): Ditto.
1114 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1115 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1116 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1117 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1118 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1120 * config/i386/i386.opt: Add -mcldemote.
1121 * config/i386/x86intrin.h: New header.
1122 * doc/invoke.texi: Add -mcldemote.
1124 2018-05-14 Richard Biener <rguenther@suse.de>
1126 * doc/match-and-simplify.texi: Adjust :s documentation.
1128 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1130 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1131 intended memcpy size.
1132 (REORDER_45): Likewise.
1134 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1136 * sort.cc: New file.
1137 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1138 * vec.c (qsort_chk): Use gcc_qsort.
1139 * Makefile.in (OBJS-libcommon): Add sort.o.
1140 (build/sort.o): New target. Use it...
1141 (BUILD_RTL): ... here, and...
1142 (build/gencfn-macros): ... here, and...
1143 (build/genmatch): ... here.
1145 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1146 Chung-Ju Wu <jasonwucj@gmail.com>
1148 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1149 * config/nds32/nds32-graywolf.md: New file.
1150 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1151 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1153 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1154 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1155 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1156 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1157 * config/nds32/pipelines.md: Include n15 settings.
1159 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1160 Chung-Ju Wu <jasonwucj@gmail.com>
1162 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1163 * config/nds32/nds32-n13.md: New file.
1164 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1165 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1167 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1168 * config/nds32/nds32.md (pipeline_model): Add n13.
1169 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1170 * config/nds32/pipelines.md: Include n13 settings.
1172 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1173 Chung-Ju Wu <jasonwucj@gmail.com>
1175 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1176 * config/nds32/nds32-n10.md: New file.
1177 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1178 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1180 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1181 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1182 * config/nds32/nds32.md (pipeline_model): Add n10.
1183 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1184 * config/nds32/pipelines.md: Include n10 settings.
1186 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1187 Kito Cheng <kito.cheng@gmail.com>
1188 Chung-Ju Wu <jasonwucj@gmail.com>
1190 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1191 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1192 Add enum values for DSP extension instructions.
1193 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1195 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1196 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1198 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1199 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1200 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1201 * config/nds32/nds32-intrinsic.md: Likewise.
1202 * config/nds32/nds32_intrinsic.h: Likewise.
1203 * config/nds32/nds32-md-auxiliary.c: Likewise.
1204 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1205 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1206 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1207 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1208 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1209 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1210 TYPE_DMAC in switch statement.
1211 * config/nds32/nds32.c: New checking and implementation for DSP
1212 extension instructions.
1213 * config/nds32/nds32.h: Likewise.
1214 * config/nds32/nds32.md: Likewise.
1215 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1216 * config/nds32/predicates.md: Implement new predicates for DSP
1219 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1221 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1222 Reformat alternatives and attributes so it is easier to identify
1223 which constraints/attributes go with which instruction.
1224 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1225 (mov<mode>_softfloat32, FMOVE64): Likewise.
1226 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1227 (mov<mode>_softfloat64, FMOVE64): Likewise.
1229 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1231 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1233 (Basic PowerPC Built-in Functions): The new name of the
1234 subsection previously known as "PowerPC Built-in Functions".
1235 (Basic PowerPC Built-in Functions Available on all Configurations):
1237 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1238 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1239 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1240 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1242 2018-05-11 Martin Jambor <mjambor@suse.cz>
1245 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1248 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1251 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1253 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1255 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1256 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1257 (ix86_handle_option): Handle -mwaitpkg.
1258 * config.gcc: New header.
1259 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1260 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1261 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1263 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1264 OPTION_MASK_ISA_WAITPKG.
1265 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1266 (ix86_option_override_internal): Add PTA_WAITPKG.
1267 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1268 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1269 IX86_BUILTIN_TPAUSE.
1270 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1271 __builtin_ia32_umwait and __builtin_ia32_tpause.
1272 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1273 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1274 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1275 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1276 UNSPECV_TPAUSE): New.
1277 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1278 * config/i386/i386.opt: Add -mwaitpkg.
1279 * config/i386/waitpkgintrin.h: New file.
1280 * config/i386/x86intrin.h: New header.
1281 * doc/invoke.texi: Add -mwaitpkg.
1283 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1286 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1288 (cortex-m0): Use armv6s-m isa.
1289 (cortex-m0plus): Likewise.
1290 (cortex-m1): Likewise.
1291 (cortex-m0.small-multiply): Likewise.
1292 (cortex-m0plus.small-multiply): Likewise.
1293 (cortex-m1.small-multiply): Likewise.
1295 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1296 Jakub Jelinek <jakub@redhat.com>
1298 PR tree-optimization/85692
1299 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1300 source permute as well.
1302 2018-05-11 Martin Liska <mliska@suse.cz>
1305 * doc/extend.texi: Document LLVM style format for no_sanitize
1308 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1310 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1311 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1312 (mode_supports_vsx_dform_quad): Likewise.
1313 (mode_supports_vmx_dform): Move these functions to be next to the
1314 other mode_supports functions.
1315 (mode_supports_dq_form): Likewise.
1316 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1317 mode_supports_dq_form.
1318 (reg_offset_addressing_ok_p): Likewise.
1319 (offsettable_ok_by_alignment): Likewise.
1320 (rs6000_legitimate_offset_address_p): Likewise.
1321 (legitimate_lo_sum_address_p): Likewise.
1322 (rs6000_legitimize_address): Likewise.
1323 (rs6000_legitimize_reload_address): Likewise.
1324 (rs6000_secondary_reload_inner): Likewise.
1325 (rs6000_preferred_reload_class): Likewise.
1326 (rs6000_output_move_128bit): Likewise.
1328 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1330 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1331 Generate SImode target register for null target.
1332 <case IX86_BUILTIN_XGETBV>: Ditto.
1333 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1334 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1336 2018-05-10 Carl Love <cel@us.ibm.com>
1338 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1339 dcbtt and dcbtstt if operands[2] is 0.
1341 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1344 * config/i386/sse.md (usadv64qi): New expander.
1346 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1348 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1349 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1350 -maltivec=be support.
1351 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1352 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1353 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1354 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1355 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1356 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1357 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1358 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1359 altivec_vsumsws): Adjust.
1360 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1361 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1362 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1364 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1365 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1366 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1367 (altivec_lve<VI_char>x): Delete expand.
1368 (*altivec_lve<VI_char>x_internal): Rename to...
1369 (altivec_lve<VI_char>x): ... this.
1370 (altivec_lvxl_<mode>): Delete expand.
1371 (*altivec_lvxl_<mode>_internal): Rename to ...
1372 (altivec_lvxl_<mode>): ... this.
1373 (altivec_stvxl_<mode>): Delete expand.
1374 (*altivec_stvxl_<mode>_internal): Rename to ...
1375 (altivec_stvxl_<mode>): ... this.
1376 (altivec_stve<VI_char>x): Delete expand.
1377 (*altivec_stve<VI_char>x_internal): Rename to ...
1378 (altivec_stve<VI_char>x): ... this.
1379 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1380 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1381 reduc_plus_scal_<mode>): Adjust.
1382 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1384 (rs6000_cpu_cpp_builtins): Adjust.
1385 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1386 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1387 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1388 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1389 -maltivec=be support.
1390 (rs6000_split_vec_extract_var): Adjust.
1391 (rs6000_split_v4si_init): Adjust.
1392 (swap_selector_for_mode): Delete.
1393 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1394 altivec_expand_stvex_be): Delete.
1395 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1396 -maltivec=be support.
1397 (rs6000_gimple_fold_builtin): Ditto.
1398 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1400 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1401 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1402 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1403 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1404 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1405 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1406 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1407 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1408 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1409 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1410 anonymous split): Adjust.
1411 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1412 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1414 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1416 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1417 when --with-gxx-include-dir is also specified.
1418 * configure: Regenerate.
1420 2018-05-09 Jim Wilson <jimw@sifive.com>
1423 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1424 * config/riscv/t-withmultilib: New.
1425 * config/riscv/withmultilib.h: New.
1426 * doc/install.texi: Document RISC-V --with-multilib-list support.
1428 2018-05-09 Richard Biener <rguenther@suse.de>
1430 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1432 (vect_bb_vectorization_profitable_p): Adjust. Compute
1433 actual scalar cost using the cost vector and the add_stmt_cost
1436 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1438 PR rtl-optimization/85645
1439 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1440 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1442 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1444 PR rtl-optimization/85645
1445 * shrink-wrap.c (spread_components): Return a boolean saying if
1446 anything was changed.
1447 (try_shrink_wrapping_separate): Iterate spread_components until
1448 nothing changes anymore.
1450 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1452 PR rtl-optimization/85645
1453 * regrename.c (build_def_use): Also kill the chains that include the
1454 destination of a REG_CFA_REGISTER note.
1456 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1458 PR rtl-optimization/85645
1459 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1460 insn that has a REG_CFA_REGISTER note.
1462 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1464 * cfgexpand.c (expand_clobber): New function.
1465 (expand_gimple_stmt_1): Use it.
1466 * tree-vect-stmts.c (vect_clobber_variable): New function,
1468 (vectorizable_simd_clone_call): ...here.
1469 (vectorizable_store): Emit a clobber either side of an
1470 IFN_STORE_LANES sequence.
1471 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1473 2018-05-09 Tom de Vries <tom@codesourcery.com>
1476 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1477 (define_insn "trap_if_false"): Add exit after trap.
1479 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1481 PR rtl-optimization/85638
1482 * bb-reorder.c: Include common/common-target.h.
1483 (create_forwarder_block): New function extracted from...
1484 (fix_up_crossing_landing_pad): ...here. Rename into...
1485 (dw2_fix_up_crossing_landing_pad): ...this.
1486 (sjlj_fix_up_crossing_landing_pad): New function.
1487 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1488 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1489 from both partitions and exit the loop after one iteration.
1491 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1494 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1496 (Basic PowerPC Built-in Functions): The new name of the
1497 subsection previously known as "PowerPC Built-in Functions".
1498 (Basic PowerPC Built-in Functions Available on all Configurations):
1500 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1502 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1504 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1506 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1509 2018-05-08 Jim Wilson <jimw@sifive.com>
1511 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1512 (LD_EMUL_SUFFIX): New.
1513 (LINK_SPEC): Use it.
1515 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1517 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1519 (Basic PowerPC Built-in Functions): The new name of the
1520 subsection previously known as "PowerPC Built-in Functions".
1521 (Basic PowerPC Built-in Functions Available on all Configurations):
1523 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1525 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1527 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1529 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1532 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1535 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1536 after cmpelim optimization.
1538 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1540 * config.gcc: Support "goldmont".
1541 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1542 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1544 * config/i386/i386.c (m_GOLDMONT): Define.
1545 (processor_target_table): Add "goldmont".
1546 (PTA_GOLDMONT): Define.
1547 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1548 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1549 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1550 (fold_builtin_cpu): Add "goldmont".
1551 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1552 (ix86_option_override_internal): Add "goldmont".
1553 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1554 (processor_type): Add PROCESSOR_GOLDMONT.
1555 * config/i386/i386.md: Add CPU "glm".
1556 * config/i386/glm.md: New file.
1557 * config/i386/x86-tune.def: Add m_GOLDMONT.
1558 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1560 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1563 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1565 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1566 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1567 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1570 * config/i386/i386.c (ix86_fold_builtin): Handle
1571 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1574 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1575 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1577 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1580 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1581 (check_arch): Likewise.
1582 (check_fpu): Return the result rather than printing it.
1583 (end arch): Fix operator precedence.
1584 (end cpu): Likewise.
1585 (END): Print the result from check_fpu.
1587 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1588 Alan Hayward <alan.hayward@arm.com>
1589 David Sherwood <david.sherwood@arm.com>
1591 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1592 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1593 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1594 (*fcmuo<mode>_and): New patterns.
1596 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1598 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1599 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1600 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1601 (cmp_op, sve_imm_con): New code attributes.
1602 (SVE_COND_INT_CMP, imm_con): Delete.
1603 (cmp_op): Remove above unspecs from int attribute.
1604 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1606 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1607 comparison-specific unspecs.
1608 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1609 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1610 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1611 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1612 (*vec_fcm<cmp_op><mode>): Rename to...
1613 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1614 (*vec_fcmuo<mode>): Rename to...
1615 (*fcmuo<mode>): ...this and adjust likewise.
1616 (*pred_fcm<cmp_op><mode>): New pattern.
1617 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1618 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1620 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1622 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1623 (aarch64_emit_sve_predicated_cond): New function.
1624 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1625 (aarch64_emit_unspec_cond_or): Replace with...
1626 (aarch64_emit_sve_or_conds): ...this new function. Use
1627 aarch64_emit_sve_ptrue_op for the individual comparisons and
1628 aarch64_emit_binop to OR them together.
1629 (aarch64_emit_inverted_unspec_cond): Replace with...
1630 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1631 aarch64_emit_sve_ptrue_op for the comparison and
1632 aarch64_emit_unop to invert the result.
1633 (aarch64_expand_sve_vec_cmp_float): Update after the above
1634 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1636 2018-05-07 Nathan Sidwell <nathan@acm.org>
1638 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1639 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1640 (Backwards Compatibility): Likewise.
1642 2018-05-07 Luis Machado <luis.machado@linaro.org>
1646 2018-05-07 Luis Machado <luis.machado@linaro.org>
1648 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1649 <prefetch_dynamic_strides>: New const bool field.
1650 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1651 prefetch_dynamic_strides.
1652 (exynosm1_prefetch_tune): Likewise.
1653 (thunderxt88_prefetch_tune): Likewise.
1654 (thunderx_prefetch_tune): Likewise.
1655 (thunderx2t99_prefetch_tune): Likewise.
1656 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1657 (aarch64_override_options_internal): Update to set
1658 PARAM_PREFETCH_DYNAMIC_STRIDES.
1659 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1660 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1661 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1662 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1663 prefetch-dynamic-strides setting.
1665 2018-05-07 Luis Machado <luis.machado@linaro.org>
1667 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1668 <minimum_stride>: New const int field.
1669 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1670 minimum_stride field.
1671 (exynosm1_prefetch_tune): Likewise.
1672 (thunderxt88_prefetch_tune): Likewise.
1673 (thunderx_prefetch_tune): Likewise.
1674 (thunderx2t99_prefetch_tune): Likewise.
1675 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1676 (aarch64_override_options_internal): Update to set
1677 PARAM_PREFETCH_MINIMUM_STRIDE.
1678 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1679 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1680 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1681 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1682 stride is constant and is below the minimum stride threshold.
1684 2018-05-07 Luis Machado <luis.machado@linaro.org>
1686 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1689 2018-05-07 Luis Machado <luis.machado@linaro.org>
1691 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1692 <prefetch_dynamic_strides>: New const bool field.
1693 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1694 prefetch_dynamic_strides.
1695 (exynosm1_prefetch_tune): Likewise.
1696 (thunderxt88_prefetch_tune): Likewise.
1697 (thunderx_prefetch_tune): Likewise.
1698 (thunderx2t99_prefetch_tune): Likewise.
1699 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1700 (aarch64_override_options_internal): Update to set
1701 PARAM_PREFETCH_DYNAMIC_STRIDES.
1702 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1703 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1704 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1705 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1706 prefetch-dynamic-strides setting.
1708 2018-05-07 Luis Machado <luis.machado@linaro.org>
1710 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1711 <minimum_stride>: New const int field.
1712 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1713 minimum_stride field.
1714 (exynosm1_prefetch_tune): Likewise.
1715 (thunderxt88_prefetch_tune): Likewise.
1716 (thunderx_prefetch_tune): Likewise.
1717 (thunderx2t99_prefetch_tune): Likewise.
1718 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1719 (aarch64_override_options_internal): Update to set
1720 PARAM_PREFETCH_MINIMUM_STRIDE.
1721 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1722 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1723 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1724 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1725 stride is constant and is below the minimum stride threshold.
1727 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1730 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1731 the type is addressable. Don't force op into register if it has
1734 2018-05-05 Roland McGrath <mcgrathr@google.com>
1737 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1738 any section for which we don't know a specific type it should have,
1739 regardless of name. Previously this was done only for the exact
1740 names ".init_array", ".fini_array", and ".preinit_array".
1741 (default_elf_asm_named_section): Add comment about
1742 relationship with default_section_type_flags and SECTION_NOTYPE.
1743 (get_section): Don't consider it a type conflict if one side has
1744 SECTION_NOTYPE and the other doesn't, as long as neither has the
1745 SECTION_BSS et al used in the default_section_type_flags logic.
1747 2018-05-05 Tom de Vries <tom@codesourcery.com>
1750 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1751 (workaround_barsyncs): New function.
1752 (nvptx_reorg): Use workaround_barsyncs.
1753 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1754 (define_expand "nvptx_membar_cta"): New define_expand.
1755 (define_insn "*nvptx_membar_cta"): New insn.
1757 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1759 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1760 To improve optimization opportunities.
1761 * builtin-types.def: The new needed builtin types for the above.
1763 2018-05-04 Richard Biener <rguenther@suse.de>
1765 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1766 * gimple-ssa-store-merging.c
1767 (imm_store_chain_info::output_merged_store): Remove redundant create,
1768 release split_store vector contents on failure.
1769 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1770 scalar stmt vector on cache hit.
1772 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1774 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1776 * config.gcc (powerpc-xilinx-eabi*): Remove.
1777 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1779 (fusion_addis_mem_combo_load): Ditto.
1780 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1782 (rs6000_cpu_cpp_builtins): Ditto.
1783 * config/rs6000/rs6000-linux.c
1784 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1785 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1786 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1788 (rs6000_setup_reg_addr_masks): Ditto.
1789 (rs6000_init_hard_regno_mode_ok): Ditto.
1790 (rs6000_option_override_internal): Ditto.
1791 (legitimate_lo_sum_address_p): Ditto.
1792 (rs6000_legitimize_address): Ditto.
1793 (rs6000_legitimize_reload_address): Ditto.
1794 (rs6000_legitimate_address_p): Ditto.
1795 (abi_v4_pass_in_fpr): Ditto.
1796 (setup_incoming_varargs): Ditto.
1797 (rs6000_gimplify_va_arg): Ditto.
1798 (rs6000_split_multireg_move): Ditto.
1799 (rs6000_savres_strategy): Ditto.
1800 (rs6000_emit_prologue_components): Ditto.
1801 (rs6000_emit_epilogue_components): Ditto.
1802 (rs6000_emit_prologue): Ditto.
1803 (rs6000_emit_epilogue): Ditto.
1804 (rs6000_elf_file_end): Ditto.
1805 (rs6000_function_value): Ditto.
1806 (rs6000_libcall_value): Ditto.
1807 * config/rs6000/rs6000.h: Ditto.
1808 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1809 (TARGET_MINMAX): ... this. New.
1810 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1811 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1812 (*movsi_internal1_single): Delete.
1813 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1814 mfpu=, mxilinx-fpu): Delete.
1815 * config/rs6000/singlefp.h: Delete.
1816 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1817 * config/rs6000/t-rs6000: Ditto.
1818 * config/rs6000/t-xilinx: Delete.
1819 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1820 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1821 (VStype_simple): Delete.
1822 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1823 * config/rs6000/xfpu.h: Delete.
1824 * config/rs6000/xfpu.md: Delete.
1825 * config/rs6000/xilinx.h: Delete.
1826 * config/rs6000/xilinx.opt: Delete.
1827 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1828 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1830 2018-05-04 Tom de Vries <tom@codesourcery.com>
1833 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1836 2018-05-04 Richard Biener <rguenther@suse.de>
1839 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1840 (expand_complex_div_wide): Likewise.
1841 (expand_complex_operations_1): Likewise.
1842 (expand_complex_libcall): Preserve EH info of the original stmt.
1843 (tree_lower_complex): Handle removed blocks.
1844 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1845 on complex multiplication and division libcall builtins.
1847 2018-05-04 Richard Biener <rguenther@suse.de>
1850 * fold-const.c (negate_expr_p): Restrict negation of operand
1851 zero of a division to when we know that can happen without
1853 (fold_negate_expr_1): Likewise.
1855 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1858 * real.h (real_nextafter): Declare.
1859 * real.c (real_nextafter): New function.
1860 * fold-const-call.c (fold_const_nextafter): New function.
1861 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1862 CASE_CFN_NEXTTOWARD.
1863 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1864 even when arg1_mode is different from arg0_mode.
1866 2018-05-03 Nathan Sidwell <nathan@acm.org>
1868 * doc/extend.texi (Deprecated Features): Remove
1870 (Backwards Compatibility): Likewise.
1871 * doc/invoke.texi (C++ Language Options): Likewise.
1872 (C++ Dialect Options): Likewise.
1874 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1877 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1878 _mm512_mask_mullox_epi64): New intrinsics.
1880 2018-05-03 Tom de Vries <tom@codesourcery.com>
1883 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1884 dump files): Add offload-tree.
1886 2018-05-03 Richard Biener <rguenther@suse.de>
1888 PR tree-optimization/85615
1889 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1890 to loops not nested in BBs loop father to avoid creating multi-entry
1893 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1895 PR tree-optimization/70291
1896 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1897 arguments. Change return type to tree. Emit libcall as a new
1898 statement rather than replacing existing one when inplace_p is true.
1899 (expand_complex_multiplication_components): New function.
1900 (expand_complex_multiplication): Expand floating-point complex
1901 multiplication using the above.
1902 (expand_complex_division): Rename inner_type parameter to type.
1903 Update expand_complex_libcall call-site.
1904 (expand_complex_operations_1): Update expand_complex_multiplication
1905 and expand_complex_division call-sites.
1907 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1910 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1911 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1912 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1913 the highest significant bit of the shift count mask is clear. In
1914 check whether and[sq]i3 is needed verify that all significant bits
1915 of the shift count other than the highest are set.
1917 2018-05-02 Tom de Vries <tom@codesourcery.com>
1920 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1921 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1922 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1923 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1924 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1925 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1926 __builtin_goacc_parlevel_size.
1928 2018-05-02 Richard Biener <rguenther@suse.de>
1930 PR tree-optimization/85597
1931 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1932 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1935 2018-05-02 Tom de Vries <tom@codesourcery.com>
1938 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1939 dump files): Add ltrans-tree.
1941 2018-05-02 Tom de Vries <tom@codesourcery.com>
1944 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1945 dump files): Add wpa-ipa.
1947 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1949 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1950 powerpc*-*-linux*paired* target.
1951 * config/rs6000/750cl.h: Delete.
1952 * config/rs6000/paired.h: Delete.
1953 * config/rs6000/paired.md: Delete.
1954 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1956 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1957 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1958 comment. Remove paired float support.
1959 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1960 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1962 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1963 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1965 * config/rs6000/rs6000.c: Remove paired float support.
1966 (paired_expand_vector_init, paired_expand_vector_move,
1967 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1968 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1969 paired_expand_builtin, paired_expand_predicate_builtin,
1970 paired_init_builtins): Delete.
1971 * config/rs6000/rs6000.h: Remove paired float support.
1972 * config/rs6000/rs6000.md: Remove paired float support.
1973 (move_from_CR_ov_bit): Delete.
1974 * config/rs6000/rs6000.opt (mpaired): Delete.
1975 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1976 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1978 2018-05-02 Richard Biener <rguenther@suse.de>
1981 * gimplify.c (gimplify_save_expr): When in SSA form allow
1982 SAVE_EXPRs to compute to SSA vars.
1984 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1987 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1988 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1989 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1990 clobber operands[2], instead use a new pseudo. Formatting fixes.
1992 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1994 PR tree-optimization/85586
1995 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1996 exit early for statements in the same group if the accesses are
1999 2018-05-02 Tom de Vries <tom@codesourcery.com>
2002 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2005 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2007 PR tree-optimization/85143
2008 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2010 2018-05-01 Tom de Vries <tom@codesourcery.com>
2013 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2014 not found" error message.
2016 2018-05-01 Tom de Vries <tom@codesourcery.com>
2019 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2020 * vec.c (test_ordered_remove_if): New function.
2021 (vec_c_tests): Call test_ordered_remove_if.
2022 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2023 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2024 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2025 VEC_ORDERED_REMOVE_IF.
2027 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2029 PR tree-optimization/82665
2030 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2031 pointer subtraction where arguments come from a memchr call.
2033 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2035 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2036 --push-state --as-needed and --pop-state instead of --as-needed and
2037 --no-as-needed if ld supports it.
2038 * configure: Regenerated.
2041 * doc/install.texi2html: Replace _002d with - and _002a with * in
2042 generated html files using sed.
2044 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2047 * gcc-rich-location.c (blank_line_before_p): New function.
2048 (use_new_line): New function.
2049 (gcc_rich_location::add_fixit_insert_formatted): New function.
2050 * gcc-rich-location.h
2051 (gcc_rich_location::add_fixit_insert_formatted): New function.
2053 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2055 * selftest.c (assert_streq): Rename "expected" and "actual" to
2056 "val1" and "val2". Extend NULL-handling to cover both inputs
2057 symmetrically, while still requiring both to be non-NULL for a pass.
2058 * selftest.h (assert_streq): Rename "expected" and "actual" to
2060 (ASSERT_EQ): Likewise.
2061 (ASSERT_EQ_AT): Likewise.
2062 (ASSERT_KNOWN_EQ): Likewise.
2063 (ASSERT_KNOWN_EQ_AT): Likewise.
2064 (ASSERT_NE): Likewise.
2065 (ASSERT_MAYBE_NE): Likewise.
2066 (ASSERT_MAYBE_NE_AT): Likewise.
2067 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2068 the assertion to pass.
2069 (ASSERT_STREQ_AT): Likewise.
2071 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2073 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2074 interaction with -pie.
2076 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2078 * selftest.h: Fix alphabetization of per-source-file selftest
2081 2018-04-30 Jason Merrill <jason@redhat.com>
2083 PR c++/61982 - dead stores to destroyed objects.
2084 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2087 2018-04-30 Jason Merrill <jason@redhat.com>
2089 * tree.c (build_clobber): New.
2090 * tree.h: Declare it.
2091 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2093 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2095 * diagnostic-show-locus.c (layout::layout): Update for
2096 location_get_source_line returning a char_span.
2097 (struct char_span): Move to input.h.
2098 (struct correction): Update for fields in char_span becoming
2100 (struct source_line): Update for location_get_source_line
2101 returning a char_span.
2102 (layout::print_line): Likewise.
2103 * edit-context.c (edited_file::print_content): Likewise.
2104 (edited_file::print_diff_hunk): Likewise.
2105 (edited_file::print_run_of_changed_lines): Likewise.
2106 (edited_file::get_num_lines): Likewise.
2107 (edited_line::edited_line): Likewise.
2108 * final.c (asm_show_source): Likewise.
2109 * input.c (location_get_source_line): Convert return type
2110 from const char * to char_span, losing the final "line_len"
2112 (dump_location_info): Update for the above.
2113 (get_substring_ranges_for_loc): Likewise. Use a char_span
2114 when handling the literal within the line.
2115 (test_reading_source_line): Update for location_get_source_line
2116 returning a char_span.
2117 * input.h (class char_span): Move here from
2118 diagnostic-show-locus.c, converting from a struct to a class.
2119 Make data members private.
2120 (char_span::operator bool): New.
2121 (char_span::length): New.
2122 (char_span::get_buffer): New.
2123 (char_span::operator[]): New.
2124 (char_span::subspan): Make const.
2125 (char_span::xstrdup): New.
2126 (location_get_source_line): Convert return type from const char *
2127 to char_span, losing the final "line_size" param.
2129 2018-04-30 Jan Hubicka <jh@suse.cz>
2131 * lto-wrapper.c (ltrans_priorities): New static var.
2132 (cmp_priority): New.
2133 (run_gcc): Read priorities and if doing parallel build order
2134 the Makefile by them.
2136 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2138 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2140 2018-04-30 Richard Biener <rguenther@suse.de>
2142 * tree-cfg.c (verify_address): Remove base argument, add
2143 flag whether to check TREE_ADDRESSABLE and do that.
2144 (verify_expr): Remove.
2145 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2146 (verify_gimple_assign_single): Likewise.
2147 (verify_gimple_switch): Likewise.
2148 (verify_expr_location_1): Dereference tp once. Add (disabled)
2149 piece from verify_expr.
2150 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2152 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2154 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2156 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2158 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2159 (small_data_pattern): Likewise.
2160 (arc_rewrite_small_data): Likewise.
2161 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2162 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2163 (get_symbol_alignment): New function.
2164 (legitimate_small_data_address_p): Likewise.
2165 (legitimate_scaled_address): Update, call
2166 legitimate_small_data_address_p.
2167 (output_sdata): New static variable.
2168 (arc_print_operand): Update how we handle small data operands.
2169 (arc_print_operand_address): Likewise.
2170 (arc_legitimate_address_p): Update, use
2171 legitimate_small_data_address_p.
2172 (arc_rewrite_small_data_p): Remove.
2173 (arc_rewrite_small_data_1): Likewise.
2174 (arc_rewrite_small_data): Likewise.
2175 (small_data_pattern): Likewise.
2176 (compact_sda_memory_operand): Update to use
2177 legitimate_small_data_address_p and get_symbol_alignment.
2178 (prepare_move_operands): Don't rewite sdata pattern.
2179 (prepare_extend_operands): Remove.
2180 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2182 (zero_extendqisi2): Likewise.
2183 (zero_extendhisi2): Likewise.
2184 (extendqihi2): Likewise.
2185 (extendqisi2): Likewise.
2186 (extendhisi2): Likewise.
2190 * config/arc/constraints.md (Usd): Change it to memory constraint.
2192 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2194 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2195 as source of std instructions.
2196 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2197 allow 6-bit constants as source for store instructions.
2198 (movdi_insn): Update instruction pattern to allow 6-bit constants
2199 as source for store instructions.
2201 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2203 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2205 2018-04-30 Nathan Sidwell <nathan@acm.org>
2206 Sandra Loosemore <sandra@codesourcery.com>
2208 * dumpfile.c (dump_open): Allow '-' for stdout.
2209 * doc/invoke.texi (Developer Options): Document dump filename
2210 determination early. Document stdin/stdout selection.
2212 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2214 Microblaze Target: PIC data text relative
2216 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2217 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2219 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2220 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2221 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2222 New addressing mode for data-text relative position indepenedent code.
2223 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2224 'ADDRESS_SYMBOLIC_TXT_REL'.
2225 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2226 (microblaze_legitimate_pic_operand): Exclude function calls from
2227 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2228 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2230 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2231 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2232 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2233 for 'address + offset'.
2234 (microblaze_expand_prologue): Add new function prologue call for
2236 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2237 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2238 table in case of TARGET_PIC_DATA_TEXT_REL.
2239 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2240 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2241 Add new macros 'UNSPEC_TEXT',
2242 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2243 + exclude function calls from 'UNSPEC_PLT' in case of data text
2245 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2246 new target hook for generating address diff vector tables in case of
2248 * doc/tm.texi : Regenerate.
2249 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2250 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2251 of addr diff vector generation.
2252 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2253 target hook definition.
2254 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2255 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2256 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2257 option for data text relative.
2259 2018-04-30 Richard Biener <rguenther@suse.de>
2261 * tree-chrec.h (evolution_function_is_constant_p): Remove
2263 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2265 2018-04-30 Richard Biener <rguenther@suse.de>
2268 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2270 2018-04-30 Richard Biener <rguenther@suse.de>
2272 PR tree-optimization/28364
2273 PR tree-optimization/85275
2274 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2275 copying first exit test.
2277 2018-04-28 Mark Wielaard <mark@klomp.org>
2279 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2281 (dwarf_AT): Handle DW_AT_addr_base.
2282 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2284 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2287 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2288 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2289 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2290 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2292 2018-04-28 Richard Biener <rguenther@suse.de>
2294 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2295 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2296 to reflect use. Only add interesting stmts.
2298 2018-04-27 Martin Jambor <mjambor@suse.cz>
2301 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2302 the jump function allows for passing through aggregate values.
2304 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2306 * input.h (in_system_header_at): Convert from macro to inline
2308 (from_macro_expansion_at): Likewise.
2309 (from_macro_definition_at): Likewise.
2311 2018-04-27 Jeff Law <law@redhat.com>
2313 * config.gcc: Mark tile* targets as deprecated/obsolete.
2315 2018-04-27 Richard Biener <rguenther@suse.de>
2317 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2320 2018-04-27 Richard Biener <rguenther@suse.de>
2322 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2324 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2326 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2327 with Yd constraint. Set "preferred_for_speed" attribute from
2328 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2330 (*movdi_internal): Ditto.
2331 (movti_interunit splitters): Remove
2332 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2333 (movdi_interunit splitters): Ditto.
2334 * config/i386/constraints.md (Ye): Remove.
2335 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2337 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2340 * config/aarch64/constraints.md (Usg): Limit to 31.
2343 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2345 PR tree-optimization/85529
2346 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2347 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2348 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2349 zero extension or masking of the MSB bit.
2350 (optimize_range_tests): Add FIRST_BB argument, pass it through
2351 to optimize_range_tests_var_bound.
2352 (maybe_optimize_range_tests, reassociate_bb): Adjust
2353 optimize_range_tests callers.
2355 2018-04-26 Richard Biener <rguenther@suse.de>
2356 Jakub Jelinek <jakub@redhat.com>
2358 * cgraph.h (symbol_table): Just declare debug method here.
2359 * symtab.c (symbol_table::debug): Define.
2361 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2363 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2365 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2367 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2368 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2369 (*movdi_internal): Substitute Yi and Yj constraint with x
2370 and Ym and Yn constraint with y constraint. Update "isa"
2371 attribute and set "preferred_for_speed" attribute from
2372 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2373 (*movsi_internal): Ditto.
2374 (*movdf_internal): Ditto.
2375 (*movsf_internal): Ditto.
2376 (*zero_extendsidi2): Ditto.
2377 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2378 (sse2_loadld): Ditto.
2379 (*vec_extract<ssevecmodelower>_0): Ditto.
2380 (*vec_extractv4si_0_zext_sse4): Ditto.
2381 (vec_concatv2di): Ditto.
2382 (*vec_dup<mode>): Ditto.
2383 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2384 * config/i386/constraints.md (Yi): Remove.
2389 2018-04-26 Nathan Sidwell <nathan@acm.org>
2391 * dumpfile.c (dump_open): New.
2392 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2393 (dump_finish): Detect stdio/stderr by value not name.
2395 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2397 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2399 2018-04-26 Tom de Vries <tom@codesourcery.com>
2402 * config/nvptx/nvptx.c (verify_neutering_jumps)
2403 (verify_neutering_labels): New function
2404 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2406 2018-04-26 Tom de Vries <tom@codesourcery.com>
2409 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2410 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2413 2018-04-26 Richard Biener <rguenther@suse.de>
2414 Tom de Vries <tom@codesourcery.com>
2417 * lto-streamer-out.c (output_function): Fixup loops if required to match
2418 discovery done in the reader.
2420 2018-04-26 Richard Biener <rguenther@suse.de>
2422 PR tree-optimization/85116
2423 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2424 have a loop exit from the single latch predecessor. Remove
2425 case of header with just condition.
2426 (ch_base::copy_headers): Exclude infinite loops from any
2428 (pass_ch::execute): Record exits.
2430 2018-04-26 Richard Biener <rguenther@suse.de>
2432 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2433 prologue cost vector and pass it to vect_get_load_cost.
2434 (vect_get_peeling_costs_all_drs): Likewise.
2435 (vect_peeling_hash_get_lowest_cost): Likewise.
2436 (vect_enhance_data_refs_alignment): Likewise.
2438 2018-04-26 Richard Biener <rguenther@suse.de>
2441 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2442 checking of integer<->pointer conversions.
2443 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2444 sign-/zero-extending pointer types.
2445 (expand_omp_for_static_chunk): Likewise.
2447 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2448 Jean Lee <xiaoyur347@gmail.com>
2450 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2451 (TARGET_ASAN_SHADOW_OFFSET): Define.
2452 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2453 true for -fsanitize=address.
2455 2018-04-25 Mark Wielaard <mark@klomp.org>
2457 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2460 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2462 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2463 than "alu", remove explicit "memory" and "imm_disp" attributes.
2464 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2467 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2468 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2469 gen_lowpart_no_emit.
2471 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2474 * config/i386/i386.c (ix86_expand_builtin): Change memory
2475 operand to XI, extend p0 to Pmode.
2476 * config/i386/i386.md: Change unspec volatile and operand
2477 1 mode to XI, change operand 0 mode to P.
2479 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2481 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2482 GET_MODE_MASK before any checking.
2483 (nds32_can_use_bset_p): Likewise.
2484 (nds32_can_use_btgl_p): Likewise.
2486 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2488 * config/nds32/nds32-doubleword.md: New define_split pattern for
2489 illegal register number.
2491 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2493 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2495 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2497 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2499 2018-04-25 Richard Biener <rguenther@suse.de>
2501 * lto-streamer.h (LTO_major_version): Bump to 8.
2503 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2505 * BASE-VER: Set to 9.0.0.
2507 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2509 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2510 in __abskf2 and __powikf2.
2512 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2515 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2516 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2517 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2518 Use the above on operand 2. Reindent.
2519 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2521 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2524 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2526 * config/i386/i386.opt (mcet): Removed.
2527 * doc/install.texi: Remove -mcet documentation.
2528 * doc/invoke.texi: Likewise.
2530 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2533 * doc/install.texi: Remove -mcet from bootstrap-cet.
2535 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2538 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2539 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2543 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2544 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2545 containing a CONST_VECTOR.
2547 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2549 * doc/install.texi: Update newlib dependency for nvptx.
2551 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2554 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2555 instead of INTVAL when shifting x left.
2557 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2559 PR tree-optimization/85478
2560 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2561 vect_grouped_store_supported for single element vectors.
2563 2018-04-24 Richard Biener <rguenther@suse.de>
2566 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2567 load cost increase to the case of non-constant step.
2569 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2572 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2573 destination if any_malformed_asm.
2575 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2578 * expr.c (store_field): In the bitfield case, if the value comes from
2579 a function call and is returned in registers by means of a PARALLEL,
2580 do not change the mode of the temporary unless BLKmode and VOIDmode.
2582 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2584 PR rtl-optimization/85423
2585 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2586 dependencies to debug insns when the previous insn is non-debug.
2588 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2590 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2591 enums into a single definition.
2592 (fls): Fix predicates and printing.
2595 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2597 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2598 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2599 and short u6 immediate.
2600 (check_if_valid_sleep_operand): Remove.
2601 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2603 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2605 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2606 flag_always_save_lp condition.
2607 * config/nds32/nds32.opt (malways-save-lp): New option.
2609 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2611 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2612 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2613 * config/nds32/nds32.h
2614 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2615 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2617 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2619 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2620 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2622 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2623 Chung-Ju Wu <jasonwucj@gmail.com>
2625 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2627 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2628 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2630 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2632 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2634 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2636 * config/nds32/nds32-protos.h (nds32_data_alignment,
2637 nds32_local_alignment): Declare.
2638 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2639 nds32_local_alignment): New functions.
2640 (TARGET_CONSTANT_ALIGNMENT): Define.
2641 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2643 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2645 * config/nds32/nds32.c
2646 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2647 (TARGET_MODES_TIEABLE_P): Likewise.
2649 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2651 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2654 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2655 Chung-Ju Wu <jasonwucj@gmail.com>
2657 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2658 for unaligned access.
2659 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2661 * config/nds32/nds32-intrinsic.md: Likewise.
2662 * config/nds32/nds32_intrinsic.h: Likewise.
2663 * config/nds32/nds32.h (nds32_builtins): Likewise.
2664 * config/nds32/nds32.opt (munaligned-access): New option.
2665 * config/nds32/nds32.c (nds32_asm_file_start): Display
2666 flag_unaligned_access status.
2668 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2670 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2671 -mno-relax is present.
2672 * config/riscv/linux.h (LINK_SPEC): Ditto.
2674 2018-04-20 Martin Sebor <msebor@redhat.com>
2677 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2679 (gimple_fold_builtin_stxcpy_chk): Same.
2680 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2682 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2685 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2686 __powikf2 when long double is IEEE 128-bit.
2688 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2690 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2691 step to make sure stack always aligned.
2693 2018-04-20 Carl Love <cel@us.ibm.com>
2696 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2697 size check for arg0.
2699 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2700 Tom de Vries <tom@codesourcery.com>
2703 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2704 Emit insns for calls too.
2705 (nvptx_find_par): Always look for worker-level predecessor insn.
2706 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2708 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2709 (nvptx_process_pars): Propagate frames for calls.
2711 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2714 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2716 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2717 (ix86_handle_option): Don't handle OPT_mibt.
2718 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2720 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2722 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2723 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2724 (ix86_target_macros): Define __CET__ with flag_cf_protection
2725 for -fcf-protection.
2726 * config/i386/i386.c (isa2_opts): Remove -mibt.
2727 * config/i386/i386.h (TARGET_IBT): Removed.
2728 (TARGET_IBT_P): Likewise.
2729 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2730 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2731 * config/i386/i386.opt (mcet): Update help message.
2734 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2735 -mcet as an alias for -mshstk.
2737 2018-04-20 Richard Biener <rguenther@suse.de>
2740 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2741 complexity by forcing a single use of the multiply operand.
2743 2018-04-20 Martin Jambor <mjambor@suse.cz>
2746 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2747 recursion dependency to only apply to non-clones.
2749 2018-04-20 Martin Jambor <mjambor@suse.cz>
2752 * ipa-cp.c (create_specialized_node): Check that clones of
2753 self-recursive edges exist during IPA-CP.
2755 2018-04-19 Toon Moene <toon@moene.org>
2757 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2760 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2762 PR tree-optimization/85467
2763 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2764 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2765 VECTOR_CST element to type.
2767 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2770 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2771 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2772 (builtin_longjmp): Likewise.
2773 (save_stack_nonlocal): New pattern.
2774 (restore_stack_nonlocal): Likewise.
2776 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2779 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2780 Replace ASM_OUTPUT_LABEL with fprintf.
2782 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2785 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2786 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2787 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2788 define __IBT__ and __SHSTK__ for -fcf-protection.
2789 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2791 (ix86_trampoline_init): Likewise.
2792 (x86_output_mi_thunk): Likewise.
2793 (ix86_notrack_prefixed_insn_p): Likewise.
2794 (ix86_option_override_internal): Don't disallow -fcf-protection.
2795 * config/i386/i386.md (rdssp<mode>): Also enable for
2797 (incssp<mode>): Likewise.
2798 (nop_endbr): Likewise.
2799 * config/i386/i386.opt (mcet): Change help message to built-in
2803 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2804 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2805 enable CET built-in functions.
2807 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2809 * common/config/i386/i386-common.c
2810 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2811 OPTION_MASK_ISA_MOVDIRI_UNSET,
2812 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2813 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2814 * config.gcc (movdirintrin.h): New header.
2815 * config/i386/cpuid.h (bit_MOVDIRI,
2816 bit_MOVDIR64B): New bits.
2817 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2819 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2820 (VOID, PVOID, PCVOID)): New function types.
2821 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2822 __builtin_ia32_directstoreu_u64,
2823 __builtin_ia32_movdir64b): New builtins.
2824 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2825 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2827 (ix86_valid_target_attribute_inner_p): Ditto.
2828 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2829 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2830 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2831 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2832 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2833 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2834 (movdiri<mode>, movdir64b_<mode>): New.
2835 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2836 * config/i386/immintrin.h: Include movdirintrin.h.
2837 * config/i386/movdirintrin.h: New file.
2838 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2840 2018-04-19 Richard Biener <rguenther@suse.de>
2843 * cfg.c (clear_bb_flags): When loop state says we have
2844 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2846 2018-04-19 Richard Biener <rguenther@suse.de>
2848 PR tree-optimization/84737
2849 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2850 copying restrict info.
2851 (vect_setup_realignment): Use it.
2852 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2853 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2854 the first DR to all generated stores.
2855 (vectorizable_load): Likewise for loads.
2857 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2859 PR tree-optimization/85446
2860 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2861 the integral and pointer types to have the same precision.
2863 * doc/install.texi: Document --disable-cet being the default and
2866 2018-04-18 Martin Liska <mliska@suse.cz>
2868 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2871 2018-04-18 Martin Liska <mliska@suse.cz>
2874 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2877 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2878 arguments if they are comparable.
2880 2018-04-18 Martin Liska <mliska@suse.cz>
2883 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2886 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2889 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2892 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2893 ENDBR after calling __morestack.
2895 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2898 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2899 by using gcc_base_ver to generate a gcc_driver_version, and use
2900 it when generating GCC_DRIVER_NAME.
2901 * configure: Regenerate.
2903 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2906 * config.gcc: Obsolete powerpc*-*-*spe*.
2908 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2911 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2912 (stabstr_D): Change type of unum from unsigned int to
2913 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2916 2018-04-17 Jim Wilson <jimw@sifive.com>
2919 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2920 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2921 Set arg_pointer_offset after using pretend_args_size.
2923 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2925 PR rtl-optimization/85431
2926 * dse.c (record_store): Ignore zero width stores.
2929 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2930 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2931 __builtin_stack_restore rather than after it.
2932 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2933 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2934 argument instead of virtual_dynamic_stack_rtx.
2936 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2938 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2940 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2941 Add note to error message to explain internal mapping of overloaded
2942 built-in function name to non-overloaded built-in function name.
2943 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2946 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2949 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2950 where the inputs overlap with the output.
2952 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2954 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2955 (=v, v) alternative and explicit "memory" attribute.
2956 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2957 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2959 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2960 "sselog1" type instead of "sselog".
2961 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2962 "sselog". Remove explicit "memory" attribute.
2963 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2964 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2966 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2967 "sselog1" type instead of "sselog". Remove explicit "memory"
2969 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2970 use "sselog1" type instead of "sselog". Remove explicit "memory"
2972 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2973 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2975 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2976 "sselog1" type instead of "sselog". Remove explicit "memory"
2978 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2979 use "sselog1" type instead of "sselog". Remove explicit "memory"
2983 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2986 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2989 2018-04-17 Martin Jambor <mjambor@suse.cz>
2992 * ipa-cp.c (create_specialized_node): Call
2993 expand_all_artificial_thunks if necessary.
2995 2018-04-17 Martin Liska <mliska@suse.cz>
2998 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2999 in message, remote space in between '_G' and '('.
3001 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3004 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3005 avx512f_vmcmp<mode>3<round_saeonly_name>,
3006 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3007 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3008 avx512f_rndscale<mode><round_saeonly_name>,
3009 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3010 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3011 Use %<iptr>2 instead of %2 for -masm=intel.
3012 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3013 avx512f_vcvttss2usi<round_saeonly_name>,
3014 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3016 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3017 avx512f_vcvttsd2usi<round_saeonly_name>,
3018 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3019 Use %q1 instead of %1 for -masm=intel.
3020 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3021 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3022 of %3 for -masm=intel.
3023 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3025 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3027 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3029 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3030 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3032 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3034 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3035 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3036 %g1 and one with %0 and %1.
3037 (avx512er_vmrcp28<mode><round_saeonly_name>,
3038 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3040 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3041 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3042 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3043 of %0 and %{%4%} for -masm=intel.
3044 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3045 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3046 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3047 order of %0 and %{%5%}%{z%} for -masm=intel.
3049 2018-04-17 Jan Hubicka <jh@suse.cz>
3052 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3054 2018-04-17 Martin Liska <mliska@suse.cz>
3057 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3058 for target_clone error message. Make default implementation
3059 clone to be a local declaration.
3060 (separate_attrs): Add new argument and check for an empty
3062 (expand_target_clones): Handle it.
3063 (ipa_target_clone): Make redirection just for target_clones
3066 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3067 Tom de Vries <tom@codesourcery.com>
3070 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3071 tiled basic blocks without omp continue statements.
3073 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3076 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3077 vec_extract expression as having side effects to make sure it gets
3080 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3083 * config/i386/i386.c (get_builtin_code_for_version): Check
3086 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3089 * gcc/config.gcc: Support "skylake".
3090 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3092 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3093 (processor_target_table): Add "skylake".
3094 (ix86_option_override_internal): Add "skylake".
3095 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3096 PROCESSOR_CANNONLAKE.
3097 (get_builtin_code_for_version): Fix priority for
3098 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3099 PROCESSOR_SKYLAKE-AVX512.
3100 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3101 (processor_type): Add PROCESSOR_SKYLAKE.
3103 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3104 Jason Merrill <jason@redhat.com>
3107 * convert.c (convert_to_integer_1): Use direct recursion for
3108 enumeral types and types with a precision less than the number
3109 of bits in their mode.
3111 2018-04-16 Julia Koval <julia.koval@intel.com>
3114 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3115 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3117 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3120 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3121 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3122 and -mno-direct-move.
3124 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3127 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3128 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3129 Ensure negative shifts result in {0}.
3131 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3133 PR rtl-optimization/79916
3134 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3135 regs (if any) to define how to gnerate SD moves when LRA is in
3138 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3140 PR rtl-optimization/85393
3141 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3142 * except.c (expand_dw2_landing_pad_for_region): Make static.
3143 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3144 a label and unconditional jump to old_bb, rather than
3145 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3148 PR rtl-optimization/85376
3149 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3150 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3151 instead of a specific value.
3153 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3154 Bin Cheng <bin.cheng@arm.com>
3156 PR tree-optimization/82965
3157 PR tree-optimization/83991
3158 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3159 by_profile_only parameter.
3160 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3161 information if the loop was predicted to iterate too many times.
3162 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3164 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3167 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3170 2018-04-13 Martin Liska <mliska@suse.cz>
3171 Jakub Jelinek <jakub@redhat.com>
3174 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3175 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3176 * builtins.c (expand_builtin_memory_copy_args): Use
3177 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3178 handle dest_addr == pc_rtx.
3180 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3183 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3184 asked to not generate direct moves.
3185 (fix_trunc<mode>si2_stfiwx): Similar.
3186 (fix_trunc<mode>si2_internal): Similar.
3188 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3191 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3192 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3193 lookup if dest in some wider mode is known to be const0_rtx and
3194 if so, record permanent equivalence for it to be ZERO_EXTEND of
3195 the narrower mode destination.
3197 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3199 * lto-streamer-out.c (output_function): Revert 259346.
3200 * omp-expand.c (expand_oacc_for): Likewise.
3202 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3204 PR rtl-optimization/85354
3205 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3206 * sel-sched.c (sel_global_init): ... here.
3208 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3211 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3212 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3213 mode for PE-COFF targets.
3214 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3215 (i386_pe_asm_lto_end): Likewise.
3216 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3217 (TARGET_ASM_LTO_END): Likewise.
3218 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3219 (i386_pe_asm_lto_start): New function.
3220 (i386_pe_asm_lto_end): Likewise.
3222 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3223 Richard Biener <rguenther@suse.de>
3226 * lto-streamer-out.c (output_function): Fix CFG loop state before
3228 * omp-expand.c (expand_oacc_for): Handle calls to internal
3229 functions like regular functions.
3231 2018-04-12 Richard Biener <rguenther@suse.de>
3234 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3235 for the early LTO debug to properly generate references to it
3236 during DIE emission. Do not re-use that for the skeleton for
3238 (dwarf2out_early_finish): Likewise.
3240 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3243 * config/i386/sse.md
3244 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3245 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3246 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3247 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3248 and output is a reg, avoid creating invalid lowpart subreg, but
3249 instead split into a 512-bit move. Don't split if not AVX512VL,
3250 input is xmm16+ reg and output is a mem.
3251 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3252 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3253 xmm16+ reg and output is a mem.
3255 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3257 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3258 also for flag_dwarf2_cfi_asm.
3260 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3262 PR rtl-optimization/85342
3263 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3264 a bool scalar var inside of the loop instead. Don't try to update
3265 recog_data.operand after failed apply_change_group.
3267 2018-04-12 Tom de Vries <tom@codesourcery.com>
3270 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3271 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3272 array with flexible array member as array without given dimension.
3273 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3274 argument for undefined param to true.
3276 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3279 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3280 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3281 from PowerPC section.
3282 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3283 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3284 help text that is too long.
3285 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3286 help text that is too long.
3287 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3288 help text that is too long.
3290 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3292 * config/alpha/alpha.md (stack_probe_internal): Rename
3293 from "probe_stack". Update all callers.
3295 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3297 PR rtl-optimization/84566
3298 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3299 sched_macro_fuse_insns.
3301 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3304 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3305 (compute_block_dependences): ... from here.
3307 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3309 PR tree-optimization/85331
3310 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3311 from int to HOST_WIDE_INT.
3313 2018-04-11 Martin Jambor <mjambor@suse.cz>
3316 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3317 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3318 not the same as the source val.
3319 (cgraph_edge_brings_value_p): New parameter.
3320 (gather_edges_for_value): Pass destination value to
3321 cgraph_edge_brings_value_p.
3322 (perhaps_add_new_callers): Likewise.
3323 (get_info_about_necessary_edges): Likewise and exclude values brought
3324 only by self-recursive edges.
3325 (create_specialized_node): Redirect only clones of self-calling edges.
3326 (+self_recursive_pass_through_p): New function.
3327 (find_more_scalar_values_for_callers_subset): Use it.
3328 (find_aggregate_values_for_callers_subset): Likewise.
3329 (known_aggs_to_agg_replacement_list): Removed.
3330 (decide_whether_version_node): Re-calculate known constants for all
3331 remaining context clones.
3333 2018-04-11 Richard Biener <rguenther@suse.de>
3336 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3337 from early DWARF output.
3338 (dwarf2out_early_finish): Output line info unconditionally into
3339 early DWARF and add reference to it.
3341 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3344 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3345 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3346 other than V2DFmode using iptr mode attribute.
3347 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3349 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3351 PR rtl-optimization/84659
3352 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3354 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3357 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3359 (output_loc_list): Pass address of a dummy size variable even in the
3360 locview handling loop.
3361 (index_location_lists): Add comment on why skip_loc_list_entry can't
3364 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3367 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3370 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3373 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3374 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3375 and -mstring-compare-inline-limit.
3377 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3380 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3381 for stack clash protection in a register whenever we need it to be in
3384 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3386 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3387 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3389 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3392 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3394 (mlong-double-): Ditto.
3395 * config/rs6000/sysv4.opt (msdata=): Ditto.
3396 (mtls-size=): Ditto.
3398 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3400 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3401 erroneous entries for
3402 "vector int vec_ldl (int, long int *)", and
3403 "vector unsigned int vec_ldl (int, unsigned long int *)".
3404 Add comments and entries for
3405 "vector bool char vec_ldl (int, bool char *)",
3406 "vector bool short vec_ldl (int, bool short *)",
3407 "vector bool int vec_ldl (int, bool int *)",
3408 "vector bool long long vec_ldl (int, bool long long *)",
3409 "vector pixel vec_ldl (int, pixel *)",
3410 "vector long long vec_ldl (int, long long *)",
3411 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3412 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3413 type tree bool_long_long_type_node and correct definition of
3414 bool_V2DI_type_node to make reference to this new type tree.
3415 (rs6000_mangle_type): Replace erroneous reference to
3416 bool_long_type_node with bool_long_long_type_node.
3417 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3418 comments to emphasize sign distinctions for char and int types and
3419 replace RS6000_BTI_bool_long constant with
3420 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3421 use of RS6000_BTI_pixel.
3422 (bool_long_type_node): Remove this macro definition.
3423 (bool_long_long_type_node): New macro definition
3425 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3427 PR rtl-optimization/85300
3428 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3429 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3430 simplify_unary_operation fails.
3432 2018-04-10 Martin Liska <mliska@suse.cz>
3434 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3435 cgraph_edge and ipa_ref.
3437 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3441 * config/i386/sse.md
3442 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3443 computation of the VEC_MERGE selector from mask.
3444 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3445 Fix decoding of the VEC_MERGE selector into mask.
3447 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3449 PR tree-optimization/85286
3450 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3452 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3454 * final.c (final_1): Set insn_last_address as well as
3455 insn_current_address.
3457 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3460 * explow.c (emit_stack_probe): Call validize_mem on memory location
3461 before passing it to gen_probe_stack. Create address operand and
3462 legitimize it for the probe_stack_address case.
3464 2018-04-09 Jan Hubicka <jh@suse.cz>
3467 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3468 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3469 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3471 (free_lang_data): Rebuild type inheritance graph.
3473 2018-04-09 Martin Sebor <msebor@redhat.com>
3475 * invoke.texi (-finline-small-functions): Mention other optimization
3477 (-findirect-inlining, -fpartial-inlining): Same.
3478 (-finline-functions-called-once): Same.
3479 (-freorder-blocks-and-partition): Same.
3481 2018-04-09 Jan Hubicka <jh@suse.cz>
3484 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3485 jumps; choose last target that matches the criteria (i.e.
3486 no partition changes for non-crossing jumps).
3487 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3488 support for redirecting crossing jumps to non-crossing.
3490 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3492 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3493 also for naked functions.
3495 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3497 * config/arc/arc.md (add_shift): New pattern.
3498 (add_shift2): Likewise.
3499 (sub_shift): Likewise.
3500 (sub_shift_cmp0_noout): Likewise.
3501 (compare_si_ashiftsi): Likewise.
3502 (xbfu_cmp0_noout): New combine pattern.
3503 (xbfu_cmp0"): Likewise.
3504 (movsi_set_cc_insn): Place the predicable variant first.
3505 (commutative_binary_cmp0_noout): Remove clobber.
3506 (commutative_binary_cmp0): New pattern.
3507 (noncommutative_binary_cmp0): Likewise.
3508 (noncommutative_binary_cmp0_noout): Likewise.
3509 (noncommutative_binary_comparison_result_used): Removed.
3510 (rsub_cmp0): New pattern.
3511 (rsub_cmp0_noout): Likewise.
3512 (extzvsi): Changed, keep only meaningful variants.
3513 (SQH, SEZ): New iterators.
3514 (SQH_postfix): New mode attribute.
3515 (SEZ_prefix): New code attribute.
3516 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3517 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3518 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3520 (noncommutative_operator): Check the availability of barrel
3523 2018-04-09 Richard Biener <rguenther@suse.de>
3525 PR tree-optimization/85284
3526 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3527 Only use the niter constraining form of simple_iv when the exit
3530 2018-04-09 Tom de Vries <tom@codesourcery.com>
3533 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3534 (define_expand "*memory_barrier"): New define_expand.
3535 (define_insn "memory_barrier"): New insn.
3537 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3539 PR rtl-optimization/80463
3540 PR rtl-optimization/83972
3541 PR rtl-optimization/83480
3543 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3544 correct producer for the insn.
3545 (tidy_control_flow): Fixup seqnos in case of debug insns.
3547 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3549 PR rtl-optimization/83913
3551 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3552 different sched-times when merging exprs.
3554 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3556 PR rtl-optimization/83962
3558 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3559 tidy_fallthru_edge and tidy_control_flow.
3561 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3563 PR rtl-optimization/83530
3565 * sel-sched.c (force_next_insn): New global variable.
3566 (remove_insn_for_debug): When force_next_insn is true, also leave only
3567 next insn in the ready list.
3568 (sel_sched_region): When the region wasn't scheduled, make another pass
3569 over it with force_next_insn set to 1.
3571 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3573 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3575 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3576 for interrupt control.
3577 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3578 functions for interrupt control.
3579 * config/nds32/nds32-intrinsic.md: Likewise.
3580 * config/nds32/nds32_intrinsic.h: Likewise.
3581 * config/nds32/nds32.h (nds32_builtins): Likewise.
3583 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3585 * config/nds32/nds32.c (nds32_init_machine_status,
3586 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3587 strict_aligned_p field.
3588 (nds32_expand_to_rtl_hook): New function.
3589 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3590 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3592 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3593 Chung-Ju Wu <jasonwucj@gmail.com>
3595 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3596 * config/nds32/nds32-n7.md: New file.
3597 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3598 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3600 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3601 * config/nds32/nds32.md (pipeline_model): Add n7.
3602 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3603 * config/nds32/pipelines.md: Include n7 settings.
3605 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3606 Chung-Ju Wu <jasonwucj@gmail.com>
3608 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3609 * config/nds32/nds32-e8.md: New file.
3610 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3611 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3613 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3614 * config/nds32/nds32.md (pipeline_model): Add e8.
3615 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3616 * config/nds32/pipelines.md: Include e8 settings.
3618 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3619 Chung-Ju Wu <jasonwucj@gmail.com>
3621 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3622 * config/nds32/nds32-n8.md: New file.
3623 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3624 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3626 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3627 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3628 * config/nds32/nds32.md (pipeline_model): Add n8.
3629 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3630 * config/nds32/pipelines.md: Include n8 settings.
3632 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3633 Chung-Ju Wu <jasonwucj@gmail.com>
3635 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3636 * config/nds32/nds32-n9-2r1w.md: New file.
3637 * config/nds32/nds32-n9-3r2w.md: New file.
3638 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3639 nds32_register_ports): New or modify for cpu n9.
3640 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3642 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3643 * config/nds32/nds32-utils.c: New file.
3644 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3645 TARGET_MUL_SLOW): Define.
3646 * config/nds32/nds32.md (pipeline_model): New attribute.
3647 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3648 New options that support cpu n9.
3649 * config/nds32/pipelines.md: Include n9 settings.
3650 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3652 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3654 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3655 information if necessary.
3656 (output_cond_branch_compare_zero): Likewise.
3657 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3658 (nds32_target_alignment): Refine for alignment.
3659 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3660 (FUNCTION_BOUNDARY): Modify.
3661 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3663 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3665 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3667 * config/nds32/constants.md (unspec_volatile_element): Add values for
3668 TLB operation and data prefetch.
3669 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3670 functions for TLB operation and data prefetch.
3671 * config/nds32/nds32-intrinsic.md: Likewise.
3672 * config/nds32/nds32_intrinsic.h: Likewise.
3673 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3674 (nds32_print_operand): Likewise.
3675 * config/nds32/nds32.h (nds32_builtins): Likewise.
3677 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3678 Andrew Pinski <pinsika@gcc.gnu.org>
3681 * match.pd: Use constant_boolean_node of correct type instead of
3682 boolean_true_node or boolean_false_node for simplifying
3683 pointer comparisons to zero.
3685 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3687 PR tree-optimization/80021
3688 * tree.c (verify_type_variant): Make error call in verify_variant_match
3689 translatable and remove final full stop.
3691 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3693 * config/nds32/constants.md (unspec_volatile_element): Add
3694 UNSPEC_VOLATILE_EH_RETURN.
3695 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3696 nds32_output_stack_pop): Support dwarf exception handling process.
3697 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3698 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3699 exception handling process.
3700 (nds32_compute_stack_frame): Likewise.
3701 (nds32_return_addr_rtx): Likewise.
3702 (nds32_initial_elimination_offset): Likewise.
3703 (nds32_expand_prologue): Likewise.
3704 (nds32_expand_epilogue): Likewise.
3705 (nds32_dynamic_chain_address): New function.
3706 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3708 (DYNAMIC_CHAIN_ADDRESS): Define.
3709 (EH_RETURN_DATA_REGNO): Define.
3710 (EH_RETURN_STACKADJ_RTX): Define.
3711 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3712 patterns for dwarf exception handling.
3714 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3716 * config/nds32/nds32.h: Clean up obsolete macros.
3718 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3720 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3721 Add enum values for particular instructions.
3722 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3723 particular intrinsic functions.
3724 * config/nds32/nds32-intrinsic.md: Likewise.
3725 * config/nds32/nds32_intrinsic.h: Likewise.
3726 * config/nds32/nds32.h (nds32_builtins): Likewise.
3727 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3728 (btst, ave): New patterns for particular instructions.
3730 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3732 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3733 Add enum values for atomic load/store and memory sync.
3734 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3736 * config/nds32/nds32-intrinsic.md: Likewise.
3737 * config/nds32/nds32_intrinsic.h: Likewise.
3738 * config/nds32/nds32.h (nds32_builtins): Likewise.
3740 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3742 PR tree-optimization/85257
3743 * fold-const.c (native_encode_vector): If not all elts could fit
3744 and off is -1, return 0 rather than offset.
3745 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3746 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3747 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3748 adjust buffer in native_interpret_expr call.
3750 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3752 * config/nds32/constants.md (unspec_volatile_element): Add cache
3753 control enum values.
3754 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3755 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3756 * config/nds32/nds32.c (nds32_cctl_names): New.
3757 (nds32_print_operand): Handle cache control register names.
3758 * config/nds32/nds32.h (nds32_builtins): New enum values.
3759 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3761 * config/nds32/nds32.md (type): Add mmu.
3762 * config/nds32/pipelines.md (simple_insn): Add mmu.
3764 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3766 * config/nds32/nds32.md (type): Remove call.
3767 * config/nds32/pipelines.md (simple_insn): Likewise.
3769 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3771 * config/nds32/constants.md (unspec_volatile_element): Add
3772 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3773 UNSPEC_VOLATILE_FMFCFG.
3774 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3775 description for fmfcfg and fmfcsr.
3776 (bdesc_1arg): Add fmtcsr.
3777 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3778 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3779 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3780 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3781 unspec_fmfcfg): New patterns.
3782 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3783 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3784 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3785 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3786 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3787 __nds32__fmfcfg): Define.
3789 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3791 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3792 intrinsic register names.
3793 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3794 intrinsic register enum values and macros.
3796 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3798 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3799 for load/store addressing form.
3800 (nds32_print_operand_address): Likewise.
3802 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3805 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3806 based on LABEL_REF. Remove useless assertion.
3807 (pic_address_needs_scratch): Fix formatting.
3808 (sparc_legitimize_pic_address): Minor tweaks.
3809 (sparc_delegitimize_address): Adjust assertion accordingly.
3810 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3811 into symbolic_operand.
3812 (movsi_high_pic_label_ref): Likewise.
3813 (movsi_lo_sum_pic_label_ref): Likewise.
3814 (movdi_pic_label_ref): Likewise.
3815 (movdi_high_pic_label_ref): Likewise.
3816 (movdi_lo_sum_pic_label_ref): Likewise.
3818 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3820 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3821 custom LIB_SPEC setup.
3823 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3824 Kito Cheng <kito.cheng@gmail.com>
3826 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3827 * config/riscv/freebsd.h: New.
3829 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3831 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3832 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3835 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3836 Kito Cheng <kito.cheng@gmail.com>
3838 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3839 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3840 * config/nds32/nds32-protos.h (nds32_output_call,
3841 nds32_output_return): Declare.
3842 * config/nds32/nds32.md: Refine all the call and return patterns.
3844 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3847 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3848 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3850 PR rtl-optimization/84872
3851 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3852 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3855 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3857 * expr.c (copy_blkmode_to_reg): Revert 254862.
3858 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3860 2018-04-06 Richard Biener <rguenther@suse.de>
3863 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3864 after seeing a component reference with an adjacent field. Treat
3865 refs to arrays at struct end of external decls similar to
3866 refs to unconstrained commons.
3868 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3871 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3872 look through SAVE_EXPRs with non-side-effects argument. Adjust
3874 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3877 2018-04-06 Richard Biener <rguenther@suse.de>
3880 * alias.c (find_base_term): New wrapper around find_base_term
3881 unwinding CSELIB_VAL_PTR changes.
3882 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3885 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3887 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3889 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3890 constant definitions.
3891 ("nop"): lr 0,0 -> nopr r0
3892 ("nop_lr0", "nop_lr1"): New insn definitions.
3894 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3896 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3897 NDS32_V3PUSH_AVAILABLE_P macro.
3899 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3900 Chung-Ju Wu <jasonwucj@gmail.com>
3902 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3903 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3904 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3905 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3906 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3907 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3908 * config/nds32/constraints.md: New constraints and checking for hard
3909 float configuration.
3910 * config/nds32/iterators.md: New mode iterator and attribute for hard
3911 float configuration.
3912 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3914 * config/nds32/nds32-fpu.md: New file.
3915 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3916 deal with hard float code generation.
3917 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3919 (abi_type, float_reg_number): New enum type.
3920 * config/nds32/nds32-predicates.c: New predicates for hard float.
3921 * config/nds32/nds32-protos.h: Declare functions for hard float.
3922 * config/nds32/nds32.c: Implementation for hard float configuration.
3923 * config/nds32/nds32.h: Definitions for hard float configuration.
3924 * config/nds32/nds32.md: Include hard float machine description and
3925 modify patterns for hard float configuration.
3926 * config/nds32/nds32.opt: New options for hard float configuration.
3927 * config/nds32/predicates.md: New predicates for hard float
3930 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3932 * common/config/nds32/nds32-common.c
3933 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3935 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3938 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3939 CONSTRUCTOR_ELT (ctor, ...)->value.
3941 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3944 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3946 2018-04-05 Tom de Vries <tom@codesourcery.com>
3949 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3952 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3953 Kito Cheng <kito.cheng@gmail.com>
3955 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3956 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3957 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3958 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3960 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3961 Kito Cheng <kito.cheng@gmail.com>
3963 * config/nds32/constraints.md (Ufe): New memory constraint.
3964 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3965 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3966 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3968 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3969 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3971 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3973 * config/nds32/nds32.md: Use optimize_size in the condition for
3974 alu-shift instructions.
3976 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3978 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3980 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3982 * config/nds32/nds32.md (negsi2): Refine pattern.
3984 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3985 Chung-Ju Wu <jasonwucj@gmail.com>
3987 * config/nds32/iterators.md (shift_rotate): New code iterator.
3988 (shift): New code attribute.
3989 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3990 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3991 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3992 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3993 bit-wise operations.
3994 (andsi3, *andsi3): Ditto.
3995 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3996 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3997 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3998 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3999 nds32_ior_operand, nds32_xor_operand): New predicates.
4001 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4003 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4004 (addsi3, subsi3): ... this.
4006 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4008 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4010 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4012 * config/nds32/nds32.md: Adjust indention.
4014 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4016 * config/nds32/nds32.md (feature): New attribute.
4018 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4020 * config/nds32/nds32.md (subtype): New attribute.
4022 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4025 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4026 expansion to perform a bitwise AND of the argument followed by a
4027 boolean negation of the result.
4029 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4031 PR rtl-optimization/84878
4032 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4033 the basic block. Assert the use reference is not artificial and that
4034 it has an associated insn.
4036 2018-04-04 Michael Matz <matz@suse.de>
4038 * builtins.c (compute_objsize): Pass correct operand
4039 to array_at_struct_end_p.
4041 2018-04-04 Richard Biener <rguenther@suse.de>
4044 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4045 from contexts for DINFO_LEVEL_TERSE and below.
4047 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4049 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4050 resiter_operand condition.
4051 * config/nds32/nds32.md (*move<mode>): Ditto.
4053 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4054 Monk Chiang <sh.chiang04@gmail.com>
4056 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4058 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4060 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4062 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4063 Kito Cheng <kito.cheng@gmail.com>
4065 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4066 nds32_cond_code_str, output_cond_branch,
4067 output_cond_branch_compare_zero, nds32_expand_cbranch,
4068 nds32_expand_cstore, nds32_expand_movcc,
4069 nds32_output_cbranchsi4_equality_zero,
4070 nds32_output_cbranchsi4_equality_reg,
4071 nds32_output_cbranchsi4_equality_reg_or_const_int,
4072 nds32_output_cbranchsi4_greater_less_zero: New functions.
4073 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4074 nds32_expand_cstore, nds32_expand_movcc,
4075 nds32_output_cbranchsi4_equality_zero,
4076 nds32_output_cbranchsi4_equality_reg,
4077 nds32_output_cbranchsi4_equality_reg_or_const_int,
4078 nds32_output_cbranchsi4_greater_less_zero): Declare.
4079 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4080 nds32_rimm11s_operand): New predicates.
4081 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4082 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4085 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4087 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4088 * config/nds32/nds32.md: Ditto.
4089 * config/nds32/pipelines.md: Ditto.
4091 2018-04-04 Richard Biener <rguenther@suse.de>
4093 PR tree-optimization/85168
4094 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4095 propagating abnormals.
4097 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4099 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4101 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4102 Kito Cheng <kito.cheng@gmail.com>
4104 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4105 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4106 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4107 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4108 * config/nds32/nds32.md (sibcall_internal): New.
4109 (sibcall_register): Remove.
4110 (sibcall_immediate): Remove.
4111 (sibcall_value_internal): New.
4112 (sibcall_value_register): Remove.
4113 (sibcall_value_immediate): Remove.
4114 * config/nds32/predicates.md (nds32_general_register_operand): New.
4115 (nds32_call_address_operand): New.
4117 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4119 PR rtl-optimization/85167
4120 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4121 bb_defs if *split_p, instead preinitialize it to NULL.
4123 PR tree-optimization/85156
4124 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4125 evaluating the argument multiple times.
4127 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4129 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4131 (_mm_cvtpd_ps): Likewise.
4132 (_mm_cvttpd_epi32): Likewise.
4133 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4134 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4135 vector, pixel, and bool following altivec.h include.
4137 2018-04-03 Martin Sebor <msebor@redhat.com>
4139 * doc/extend.texi (Common Function Attributes): Clarify.
4140 (const attribute): Likewise.
4141 (pure attribute): Likewise.
4143 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4146 * config/i386/i386.c (ix86_expand_vector_set): Use
4147 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4149 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4151 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4152 instructions when changing rounding bits to preserve precision bits
4153 in the x87 control word.
4155 2018-04-03 Martin Liska <mliska@suse.cz>
4157 PR tree-optimization/82491
4158 * rtl.h (strip_offset_and_add): Replace += suboffset with
4159 poly_uint64 () + suboffset.
4161 2018-03-29 Martin Liska <mliska@suse.cz>
4162 Martin Jambor <mjambor@suse.cz>
4165 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4166 param_type is not an integral or pointer type.
4168 2018-04-03 Richard Biener <rguenther@suse.de>
4170 * sese.h (recompute_all_dominators): Remove.
4172 2018-04-02 Martin Sebor <msebor@redhat.com>
4174 * doc/invoke.texi (-Wrestrict): Fix typos.
4176 2018-04-02 Jim Wilson <jimw@sifive.com>
4178 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4179 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4180 (<optab>di3, <optab>si3_extend): Likewise.
4181 (<optab>si3_mask, <optab>si3_mask_1): New.
4182 (<optab>di3_mask, <optab>di3_mask_1): New.
4183 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4184 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4185 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4187 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4189 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4192 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4194 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4195 (nds32_canonicalize_comparison): New function.
4197 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4198 Kito Cheng <kito.cheng@gmail.com>
4199 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4201 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4202 * config/nds32/constants.md (unspec_volatile_element): Add
4203 UNSPEC_VOLATILE_RELAX_GROUP.
4204 * config/nds32/nds32-relax-opt.c: New file.
4205 * config/nds32/nds32-predicates.c
4206 (nds32_symbol_load_store_p): New function.
4207 * config/nds32/nds32-protos.h
4208 (nds32_symbol_load_store_p): Declare function.
4209 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4210 * config/nds32/nds32.c
4211 (nds32_register_pass): New function to register pass.
4212 (nds32_register_passes): New function to register passes.
4213 * config/nds32/nds32.md (relax_group): New pattern.
4214 * config/nds32/nds32.opt (mrelax-hint): New option.
4215 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4217 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4219 * config/nds32/t-nds32: Modify files dependency.
4221 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4223 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4224 (PROFILE_HOOK): Define its implementation.
4226 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4228 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4229 type and 32-bit size.
4231 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4234 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4235 (V_128_256): New mode iterator.
4236 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4237 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4238 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4240 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4241 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4243 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4246 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4247 NaN inputs correctly.
4249 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4252 * config/rs6000/vsx.md (??r): New mode attribute.
4253 (*vsx_mov<mode>_64bit): Use it.
4254 (*vsx_mov<mode>_32bit): Likewise.
4256 2018-03-30 Martin Sebor <msebor@redhat.com>
4258 PR tree-optimization/84818
4259 * builtins.c (check_access): Use warning_n.
4261 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4264 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4266 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4269 2018-03-30 Julia Koval <julia.koval@intel.com>
4272 * x86-tune.def (movx, partial_reg_dependency): Enable for
4275 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4278 * lra-constraints.c (process_alt_operands): Move setting
4279 this_alternative_matches below.
4281 2018-03-29 Martin Liska <mliska@suse.cz>
4284 * doc/invoke.texi: Document how LTO works with debug info.
4285 Describe auto-load support of binutils. Mention 'x86-64'
4286 as valid option value of -march option.
4288 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4290 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4293 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4294 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4295 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4298 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4301 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4302 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4303 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4304 for RS6000_BTM_POWERPC64.
4305 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4306 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4307 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4312 2018-03-28 Carl Love <cel@us.ibm.com>
4315 2017-09-27 Carl Love <cel@us.ibm.com>
4317 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4318 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4319 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4322 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4324 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4325 instead of __vector bool.
4326 (_mm_max_pu8): Likewise.
4327 (_mm_min_pi16): Likewise.
4329 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4332 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4333 (DIVWEUO): Likewise.
4335 (DIVDEUO): Likewise.
4336 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4337 DIVWEUO and DIVDEUO.
4338 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4339 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4340 (div_extend): Likewise.
4341 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4343 (__builtin_divweuo): Likewise.
4344 (__builtin_divdeo): Likewise.
4345 (__builtin_divdeuo): Likewise.
4347 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4350 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4351 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4353 PR tree-optimization/82004
4354 * gimple-match-head.c (optimize_pow_to_exp): New function.
4355 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4356 Don't fold to exp if optimize_pow_to_exp is false.
4358 2018-03-28 Martin Liska <mliska@suse.cz>
4361 * calls.c (initialize_argument_information): Fix trailing space.
4362 * common.opt: Fix typo and provide better explanation for
4363 -fsanitize-coverage option.
4364 * config/i386/i386.opt: Fix typo.
4366 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4367 Martin Liska <mliska@suse.cz>
4370 * gimplify.c (asan_poison_variable): Don't do the check for
4371 gimplify_omp_ctxp here.
4372 (gimplify_decl_expr): Do it here.
4373 (gimplify_target_expr): Likewise.
4375 2018-03-28 Martin Liska <mliska@suse.cz>
4378 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4379 chkp_type_bounds_count if MPX is not enabled.
4381 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4383 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4385 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4388 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4389 function to create the function decl for complex long double
4390 multiply and divide for -mabi=ieeelongdouble.
4391 (init_float128_ieee): Call it.
4393 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4396 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4397 -fcf-protection=branch -mibt.
4398 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4400 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4403 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4405 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4408 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4409 extern array declarations.
4411 2018-03-27 Richard Biener <rguenther@suse.de>
4414 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4415 explicit single_use checks.
4417 2018-03-27 Richard Biener <rguenther@suse.de>
4419 PR tree-optimization/85082
4420 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4423 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4425 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4426 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4427 Turn on fasynchronous-unwind-tables and funwind-tables.
4429 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4432 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4433 (*bmi_blsr_<mode>_ccz): Ditto.
4435 2018-03-26 Tom de Vries <tom@codesourcery.com>
4437 PR tree-optimization/85063
4438 * omp-general.c (offloading_function_p): New function. Factor out
4440 * omp-offload.c (pass_omp_target_link::gate): ... here.
4441 * omp-general.h (offloading_function_p): Declare.
4442 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4443 with attribute omp declare target for offloading functions.
4445 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4447 PR tree-optimization/84005
4448 * tree-data-ref.h (get_base_for_alignment): Declare.
4449 * tree-data-ref.c (get_base_for_alignment_1): New function.
4450 (get_base_for_alignment): Likewise.
4451 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4452 get_base_for_alignment to find a suitable base object, instead
4453 of always using drb->base_address.
4455 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4458 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4459 known size by default.
4461 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4464 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4465 and non BLKmode operands.
4467 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4470 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4471 Clean up attributes.
4473 2018-03-23 Richard Biener <rguenther@suse.de>
4476 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4477 we are going to emit early debug for LTO.
4479 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4482 * function.c (match_asm_constraints_1): Don't optimize if input
4483 doesn't satisfy general_operand predicate for output's mode.
4486 * alias.c (write_dependence_p): Don't require for x_canonicalized
4487 non-VOIDmode if x has VOIDmode.
4490 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4491 just don't try to optimize it rather than assert it never happens.
4493 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4495 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4496 macro expansions for definition of ST_INTERNAL_<mode> and
4497 LD_INTERNAL_<mode> builtins.
4498 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4500 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4502 (altivec_expand_st_builtin): Likewise.
4503 (altivec_expand_builtin): Remove calls to deleted functions.
4504 (rs6000_address_for_altivec): Delete this function.
4505 * config/rs6000/vector.md: Remove expands for
4506 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4508 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4511 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4512 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4513 re-computing once computed.
4514 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4515 (arm_init_machine_status): Initialize
4516 machine->static_chain_stack_bytes.
4518 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4521 * doc/extend.texi: Add four new prototypes for vec_ld.
4522 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4523 definitions for more logical presentation.
4524 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4525 entries for V1TI variants of __builtin_altivec_ld builtin.
4526 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4527 handling of V1TI variant of LVX icode pattern.
4528 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4529 (rs6000_gimple_fold_builtin): Likewise.
4530 (altivec_init_builtins): Add code to define
4531 __builtin_altivec_lvx_v1ti function.
4533 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4536 * function.c (match_asm_constraints_1): Don't do the optimization
4537 if input isn't a REG, SUBREG, MEM or constant.
4539 2018-03-22 Tom de Vries <tom@codesourcery.com>
4541 PR tree-optimization/84956
4542 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4543 bb_has_abnormal_pred.
4545 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4548 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4549 DECL_INITIAL (decl) to decl at the end.
4550 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4553 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4555 * doc/extend.texi (__builtin_tgmath): Document when complex
4556 integer types are treated as _Complex _Float64.
4558 2018-03-21 Tom de Vries <tom@codesourcery.com>
4560 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4562 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4564 PR tree-optimization/84960
4565 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4566 if it is ENTRY block, move them into single succ of ENTRY in that case.
4568 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4570 PR tree-optimization/84811
4571 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4572 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4573 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4574 (known_subrange_p): Update accordingly. Cast each value involved
4575 in the size comparison, rather than casting the result of the
4578 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4580 PR tree-optimization/84982
4581 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4582 by flipping the least significant bit rather than all bits from
4583 bitpos to bitpos + bitsize - 1.
4585 2018-03-21 Nathan Sidwell <nathan@acm.org>
4587 * doc/extend.texi (Deprecated Features): Remove mention of
4588 long-deleted deprecations.
4590 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4593 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4594 * configure: Regenerate.
4596 2018-03-21 Tom de Vries <tom@codesourcery.com>
4598 PR tree-optimization/83126
4599 * tree-parloops.c (num_phis): New function.
4600 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4602 2018-03-21 Nathan Sidwell <nathan@acm.org>
4604 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4605 mention anon-struct/union members and trailing attributes.
4607 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4609 PR tree-optimization/84969
4610 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4611 builtin memset partitions if they set different rhs values.
4613 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4615 PR rtl-optimization/84989
4616 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4617 VEC_DUPLICATE with scalar result mode.
4619 2018-03-21 Martin Liska <mliska@suse.cz>
4622 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4623 not intended return statement.
4625 2018-03-21 Martin Liska <mliska@suse.cz>
4628 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4629 (chkp_find_bound_slots_1): Limit number of iterations.
4631 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4634 * Minor grammar fixes for x86 options.
4636 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4639 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4640 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4642 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4645 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4646 (altivec_lvx_<mode>_1op): Likewise.
4647 (altivec_stvx_<mode>_2op): Likewise.
4648 (altivec_stvx_<mode>_1op): Likewise.
4649 (altivec_lvx_<VM2:mode>): New define_expand.
4650 (altivec_stvx_<VM2:mode>): Likewise.
4651 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4652 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4653 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4654 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4655 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4656 (rs6000_gen_lvx): Likewise.
4657 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4658 (altivec_expand_stv_builtin): Likewise.
4659 (altivec_expand_builtin): Likewise.
4660 * config/rs6000/vector.md: Likewise.
4662 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4665 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4668 2018-03-20 Richard Biener <rguenther@suse.de>
4671 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4672 sign-conversions as zero, fall back to standard scalar_stmt
4675 2018-03-20 Martin Liska <mliska@suse.cz>
4678 * predict.c (rebuild_frequencies): Handle case when we have
4679 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4681 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4684 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4685 flag_section_anchors.
4686 * varasm.c (use_blocks_for_decl_p): Remove hack for
4687 dw2_force_const_mem.
4690 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4692 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4693 be created, use lowpart_subreg of operands[0] rather than operands[0]
4695 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4696 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4697 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4698 and n constraint instead of aarch64_shift_imm_di and Usd.
4699 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4700 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4702 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4705 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4706 to favor GPR over NEON registers.
4707 (<shift>di3_neon): Likewise.
4709 2018-03-20 Tom de Vries <tom@codesourcery.com>
4712 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4713 (nvptx_process_pars): Emit bar.sync asap and alap.
4715 2018-03-20 Tom de Vries <tom@codesourcery.com>
4718 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4719 seen_label if seen_label is already set.
4721 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4724 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4725 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4726 Use 1U instead of 1. Formatting fixes.
4729 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4730 instead of TREE_TYPE (s1) for the return value.
4732 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4734 PR tree-optimization/84946
4735 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4736 bitsize + bitsize in poly_uint64 rather than poly_int64.
4739 * dwarf2asm.c: Include fold-const.c.
4740 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4741 of decl rather than decl itself.
4743 PR rtl-optimization/84643
4744 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4746 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4749 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4750 calling assemble_variable.
4752 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4755 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4756 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4758 2018-03-19 Jim Wilson <jimw@sifive.com>
4761 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4762 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4763 (riscv_first_stack_step): Likewise.
4764 (riscv_option_override): Use STACK_BOUNDARY instead of
4766 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4768 (BIGGEST_ALIGNMENT): Set to 128.
4769 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4770 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4773 2018-03-19 Richard Biener <rguenther@suse.de>
4775 PR tree-optimization/84933
4776 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4777 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4779 2018-03-19 Richard Biener <rguenther@suse.de>
4781 PR tree-optimization/84859
4782 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4783 (cond_if_else_store_replacement): Perform sinking operation on
4784 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4785 Generalize what a BB with a single eligible store is.
4787 2018-03-19 Richard Biener <rguenther@suse.de>
4789 PR tree-optimization/84929
4790 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4791 chrec_is_positive against non-chrec arg.
4793 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4796 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4798 2018-03-18 Martin Liska <mliska@suse.cz>
4800 PR rtl-optimization/84635
4801 * regrename.c (build_def_use): Use matches_mode only when
4804 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4806 PR tree-optimization/84913
4807 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4808 vectorize chains of COND_EXPRs.
4810 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4812 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4814 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4816 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4818 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4820 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4822 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4823 Kito Cheng <kito.cheng@gmail.com>
4825 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4826 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4827 (nds32_adjust_reg_alloc_order): New function.
4828 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4830 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4832 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4833 nds32_print_operand, nds32_print_operand_address): Use
4834 HOST_WIDE_INT_PRINT_DEC instead.
4836 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4838 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4840 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4843 * config/i386/i386.c (initial_ix86_tune_features,
4844 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4846 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4847 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4848 rather than 1u << ix86_tune. Formatting fix.
4849 (ix86_option_override_internal): Change ix86_arch_mask from
4850 unsigned int to unsigned HOST_WIDE_INT, initialize to
4851 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4852 (ix86_function_specific_restore): Likewise.
4854 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4857 * postreload.c (reload_combine_recognize_pattern): Perform
4858 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4859 truncate_int_for_mode the result for the destination's mode.
4862 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4863 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4865 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4868 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4871 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4872 regno_allocno_class_array and sorted_pseudos.
4873 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4874 insns where regno is used.
4876 2018-03-16 Martin Liska <mliska@suse.cz>
4879 * multiple_target.c (create_dispatcher_calls): Redirect
4880 reference in the symbol table.
4882 2018-03-16 Martin Liska <mliska@suse.cz>
4885 * multiple_target.c (create_dispatcher_calls): Redirect also
4888 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4892 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4893 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4894 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4896 2018-03-16 Julia Koval <julia.koval@intel.com>
4898 * doc/invoke.texi (Skylake Server): Add CLWB.
4899 Cannonlake): Remove CLWB.
4901 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4903 PR tree-optimization/84841
4904 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4906 (FLOAT_ONE_CONST_TYPE): Define.
4907 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4908 (sort_by_operand_rank): Put entries with higher constant_type last
4909 rather than first to match comments.
4911 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4913 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4916 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4919 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4920 check and use address of target always.
4922 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4925 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4926 (indirect_thunk_bnd_needed): Likewise.
4927 (indirect_thunks_used): Likewise.
4928 (indirect_thunks_bnd_used): Likewise.
4929 (indirect_return_needed): New.
4930 (indirect_return_bnd_needed): Likewise.
4931 (output_indirect_thunk_function): Add a bool argument for
4933 (output_indirect_thunk_function): Don't generate alias for
4934 function return thunk.
4935 (ix86_code_end): Call output_indirect_thunk_function to generate
4936 function return thunks.
4937 (ix86_output_function_return): Set indirect_return_bnd_needed
4938 and indirect_return_needed instead of indirect_thunk_bnd_needed
4939 and indirect_thunk_needed.
4941 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4943 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4944 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4945 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4947 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4948 Paul Hua <paul.hua.gm@gmail.com>
4951 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4953 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4955 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4956 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4959 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4962 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4963 instead of GET_MODE_SIZE when comparing Units.
4965 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4968 * varasm.c (hash_section): Return an unchangeble hash value
4969 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4970 Return !aarch64_can_use_per_function_literal_pools_p ().
4972 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4975 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4976 rather than address of cmode as last argument to prepare_cmp_insn.
4978 2018-03-15 Julia Koval <julia.koval@intel.com>
4980 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4981 F_AVX512VNNI, F_AVX512BITALG): New.
4983 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4986 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4987 insn for floating-point loads and stores.
4989 2018-03-14 Carl Love <cel@us.ibm.com>
4991 * config/rs6000/rs6000-c.c: Add macro definitions for
4992 ALTIVEC_BUILTIN_VEC_PERMXOR.
4993 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4994 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4995 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4996 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4998 * config/doc/extend.texi: Add prototypes for vec_permxor.
5000 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5003 * diagnostic-show-locus.c (class layout_point): Convert m_line
5004 from int to linenum_type.
5005 (line_span::comparator): Use linenum "compare" function when
5006 comparing line numbers.
5007 (test_line_span): New function.
5008 (layout_range::contains_point): Convert param "row" from int to
5010 (layout_range::intersects_line_p): Likewise.
5011 (layout::will_show_line_p): Likewise.
5012 (layout::print_source_line): Likewise.
5013 (layout::should_print_annotation_line_p): Likewise.
5014 (layout::print_annotation_line): Likewise.
5015 (layout::print_leading_fixits): Likewise.
5016 (layout::annotation_line_showed_range_p): Likewise.
5017 (struct line_corrections): Likewise for field m_row.
5018 (line_corrections::line_corrections): Likewise for param "row".
5019 (layout::print_trailing_fixits): Likewise.
5020 (layout::get_state_at_point): Likewise.
5021 (layout::get_x_bound_for_row): Likewise.
5022 (layout::print_line): Likewise.
5023 (diagnostic_show_locus): Likewise for locals "last_line" and
5025 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5026 * input.c (selftest::test_linenum_comparisons): New function.
5027 (selftest::input_c_tests): Call it.
5028 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5029 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5030 * selftest.h (ASSERT_GT): New macro.
5031 (ASSERT_GT_AT): New macro.
5032 (ASSERT_LT): New macro.
5033 (ASSERT_LT_AT): New macro.
5035 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5037 PR rtl-optimization/84780
5038 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5040 2018-03-14 Martin Liska <mliska@suse.cz>
5042 * tree.c (record_node_allocation_statistics): Use
5043 get_stats_node_kind.
5044 (get_stats_node_kind): New function extracted from
5045 record_node_allocation_statistics.
5046 (free_node): Use get_stats_node_kind.
5048 2018-03-14 Richard Biener <rguenther@suse.de>
5050 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5051 that the value-set of ANTIC_IN doesn't grow.
5054 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5056 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5057 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5059 2018-03-14 Julia Koval <julia.koval@intel.com>
5061 * config.gcc (icelake-client, icelake-server): New.
5063 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5064 (initial_ix86_arch_features): Ditto.
5065 (PTA_SKYLAKE): Add SGX.
5066 (PTA_ICELAKE): Remove.
5067 (PTA_ICELAKE_CLIENT): New.
5068 (PTA_ICELAKE_SERVER): New.
5069 (ix86_option_override_internal): Split up icelake on icelake client and
5071 (get_builtin_code_for_version): Ditto.
5072 (fold_builtin_cpu): Ditto.
5073 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5074 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5075 * config/i386/i386.h (processor_type): Ditto.
5076 * doc/invoke.texi: Ditto.
5078 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5081 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5082 INTEGER_CST offset, add it together with bitpos / 8 and
5083 sign extend based on POINTER_SIZE.
5087 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5090 * config/i386/constraints.md (Yc): New register constraint.
5091 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5092 Use Yc constraint for alternative 2 of operand 0. Remove
5093 preferred_for_speed attribute.
5095 2018-03-14 Richard Biener <rguenther@suse.de>
5097 PR tree-optimization/84830
5098 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5099 with the old one to avoid oscillations.
5101 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5104 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5106 (assign_by_spills): Return a flag of reload assignment failure.
5107 Do not process the reload assignment failures. Do not spill other
5108 reload pseudos if they has the same reg class. Update n if
5110 (lra_assign): Add a return arg. Set up from the result of
5111 assign_by_spills call.
5112 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5113 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5114 usage_insns if it is not NULL.
5115 (spill_hard_reg_in_range): New function.
5116 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5117 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5118 function prototypes.
5119 (lra_assign): Change prototype.
5120 * lra.c (lra): Add code to deal with fails by splitting hard reg
5123 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5125 * config/riscv/riscv.opt (mrelax): New option.
5126 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5127 "norelax" when riscv_mrelax is disabled.
5128 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5130 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5133 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5134 reassociation for int modes.
5136 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5138 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5139 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5141 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5142 * config/aarch64/aarch64-sve.md
5143 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5144 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5145 (*extend<mode><Vwide>2): Rename to...
5146 (aarch64_sve_extend<mode><Vwide>2): ...this.
5147 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5148 renaming the old pattern to...
5149 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5151 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5152 define_expand, renaming the old pattern to...
5153 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5154 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5155 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5156 account when deciding which SVE instruction the optab should use.
5157 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5159 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5161 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5162 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5163 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5164 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5165 (tlsdesc_small_<mode>): Turn a define_expand and use
5166 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5167 (tlsdesc_small_advsimd_<mode>): ...this.
5168 (tlsdesc_small_sve_<mode>): New pattern.
5170 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5172 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5173 (UNSPEC_UMUL_HIGHPART): New constants.
5174 (MUL_HIGHPART): New int iteraor.
5175 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5176 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5178 (*<su>mul<mode>3_highpart): New define_insn.
5180 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5183 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5186 2018-03-13 Martin Liska <mliska@suse.cz>
5189 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5191 (sem_item_optimizer::~sem_item_optimizer): Release it.
5192 (sem_item_optimizer::merge_classes): Register variable aliases.
5193 (sem_item_optimizer::fixup_pt_set): New function.
5194 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5195 * ipa-icf.h: Declare new variables and functions.
5197 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5200 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5201 integer_pow2p@2 and test integer_pow2p in condition.
5202 (A < 0 ? C : 0): Similarly for @1.
5205 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5206 characters starting at p contain '\0' character, don't look beyond
5210 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5211 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5214 * reg-stack.c (change_stack): Change update_end var from int to
5215 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5216 also call set_block_for_insn on the newly added insns and rescan.
5219 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5220 on the last operand.
5223 * tree.c (stabilize_reference_1): Return save_expr (e) for
5224 STATEMENT_LIST even if it doesn't have side-effects.
5226 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5228 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5230 2018-03-12 Renlin Li <renlin.li@arm.com>
5232 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5233 aarch64_output_scalar_simd_mov_immediate.
5235 2018-03-12 Martin Sebor <msebor@redhat.com>
5237 PR tree-optimization/83456
5238 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5239 for perfectly overlapping calls to memcpy.
5240 (gimple_fold_builtin_memory_chk): Same.
5241 (gimple_fold_builtin_strcpy): Handle no-warning.
5242 (gimple_fold_builtin_stxcpy_chk): Same.
5243 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5245 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5247 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5248 parameter. Use it for SFmode.
5249 (rs6000_function_arg_advance_1): Adjust.
5250 (rs6000_function_arg): Adjust.
5251 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5253 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5255 PR rtl-optimization/84169
5256 PR rtl-optimization/84780
5257 * combine.c (can_combine_p): Check for a 2-insn combination whether
5258 the destination register is used between the two insns, too.
5260 2018-03-12 Richard Biener <rguenther@suse.de>
5262 PR tree-optimization/84803
5263 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5264 for refs DR analysis didn't process.
5266 2018-03-12 Richard Biener <rguenther@suse.de>
5268 PR tree-optimization/84777
5269 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5270 force-vectorize loops ignore whether we are optimizing for size.
5272 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5274 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5275 (TARGET_MD_ASM_ADJUST): Define.
5277 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5278 Kito Cheng <kito.cheng@gmail.com>
5279 Chung-Ju Wu <jasonwucj@gmail.com>
5281 * config/nds32/nds32.c (nds32_compute_stack_frame,
5282 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5283 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5284 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5285 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5286 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5287 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5288 * config/nds32/nds32.md (prologue, epilogue): Use macro
5289 NDS32_V3PUSH_AVAILABLE_P to do checking.
5291 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5294 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5295 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5296 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5297 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5298 addition of most attributes on !orig_type_die or the attribute not
5299 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5301 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5302 Chung-Ju Wu <jasonwucj@gmail.com>
5304 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5306 * config/nds32/nds32.opt (mvh): New option.
5308 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5309 Chung-Ju Wu <jasonwucj@gmail.com>
5311 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5313 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5314 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5317 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5318 Chung-Ju Wu <jasonwucj@gmail.com>
5320 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5322 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5323 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5325 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5326 Kito Cheng <kito.cheng@gmail.com>
5327 Chung-Ju Wu <jasonwucj@gmail.com>
5329 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5330 UNSPEC_FFMISM and UNSPEC_FLMISM.
5331 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5332 for ffb, ffmism and flmism.
5333 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5334 (unspec_ffmism): Ditto.
5335 (unspec_flmism): Ditto.
5336 (nds32_expand_builtin_impl): Check if string extension is available.
5337 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5338 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5340 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5343 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5346 * lra-assigns.c (assign_by_spills): Return a flag of reload
5347 assignment failure. Do not process the reload assignment
5348 failures. Do not spill other reload pseudos if they has the same
5350 (lra_assign): Add a return arg. Set up from the result of
5351 assign_by_spills call.
5352 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5353 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5354 usage_insns if it is not NULL.
5355 (spill_hard_reg_in_range): New function.
5356 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5357 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5358 function prototypes.
5359 (lra_assign): Change prototype.
5360 * lra.c (lra): Add code to deal with fails by splitting hard reg
5363 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5366 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5368 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5371 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5372 (dw_val_node): Add val_symbolic_view.
5373 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5374 (symview_upper_bound): New.
5375 (new_line_info_table): Initialize symviews_since_reset.
5376 (dwarf2out_source_line): Count symviews_since_reset and set
5377 symview_upper_bound.
5378 (dw_val_equal_p): Handle symview.
5379 (add_AT_symview): New.
5380 (print_dw_val): Handle symview.
5381 (attr_checksum, attr_checksum_ordered): Likewise.
5382 (same_dw_val_p, size_of_die): Likewise.
5383 (value_format, output_die): Likewise.
5384 (add_high_low_attributes): Use add_AT_symview for entry_view.
5385 (dwarf2out_finish): Reset symview_upper_bound, clear
5388 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5391 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5392 Add strict argument and use it.
5393 (rs6000_split_multireg_move): Update for new strict argument.
5394 (mem_operand_gpr): Disallow all non-offsettable addresses.
5395 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5397 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5400 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5401 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5402 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5405 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5406 decl, use remap_type if we want to use the type.
5408 2018-03-09 Martin Sebor <msebor@redhat.com>
5410 PR tree-optimization/84526
5411 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5413 (builtin_access::generic_overlap): Be prepared to handle non-array
5416 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5418 PR rtl-optimization/84682
5419 * lra-constraints.c (process_address_1): Check is_address flag
5420 for address constraints.
5421 (process_alt_operands): Likewise.
5422 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5423 preprocess_constraints.
5424 * recog.h (preprocess_constraints): Add oploc parameter.
5426 * recog.c (preprocess_constraints): Test address_operand for
5427 CT_ADDRESS constraints.
5429 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5432 * lra-assigns.c (assign_by_spills): Return a flag of reload
5433 assignment failure. Do not process the reload assignment
5434 failures. Do not spill other reload pseudos if they has the same
5436 (lra_assign): Add a return arg. Set up from the result of
5437 assign_by_spills call.
5438 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5439 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5440 usage_insns if it is not NULL.
5441 (spill_hard_reg_in_range): New function.
5442 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5443 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5444 function prototypes.
5445 (lra_assign): Change prototype.
5446 * lra.c (lra): Add code to deal with fails by splitting hard reg
5449 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5452 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5453 Accept complain bool parameter. Only emit errors if it is true.
5454 (arm_parse_cpu_option_name): Likewise.
5455 (arm_target_thumb_only): Adjust callers of the above.
5456 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5457 prototype to take a default true bool parameter.
5458 (arm_parse_arch_option_name): Likewise.
5460 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5461 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5465 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5466 * configure: Regenerate.
5467 * configure.ac ("linker --version-script option"): New.
5468 ("linker soname option"): New.
5470 2018-03-09 Richard Biener <rguenther@suse.de>
5472 PR tree-optimization/84775
5473 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5474 immediate uses of predicate stmts and mark them modified.
5477 PR tree-optimization/84178
5478 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5480 (version_loop_for_if_conversion): Delay update_ssa call.
5481 (tree_if_conversion): Delay update_ssa until after predicate
5484 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5487 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5488 when the function accesses prior frames.
5490 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5493 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5494 gen_llsym, otherwise call maybe_gen_llsym.
5497 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5498 has ',' character inside of it.
5500 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5503 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5504 as clobbering CC_REGNUM.
5506 2018-03-08 Richard Biener <rguenther@suse.de>
5509 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5510 (follow_copies_to_constant): Do not follow SSA names registered
5513 2018-03-08 Richard Biener <rguenther@suse.de>
5515 PR tree-optimization/84178
5516 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5518 (version_loop_for_if_conversion): Delay update_ssa call.
5519 (tree_if_conversion): Delay update_ssa until after predicate
5522 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5524 PR tree-optimization/84178
5525 * tree-if-conv.c (release_bb_predicate): Remove the
5526 the assertion that the stmts have NULL use_ops.
5527 Discard the statements, asserting that they haven't
5528 yet been added to a BB.
5530 2018-03-08 Richard Biener <rguenther@suse.de>
5532 PR tree-optimization/84746
5533 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5534 (phi_translate): Pass in destination ANTIC_OUT set.
5535 (phi_translate_1): Likewise. For a simplified result lookup
5536 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5537 (phi_translate_set): Adjust.
5538 (do_pre_regular_insertion): Likewise.
5539 (do_pre_partial_partial_insertion): Likewise.
5541 2018-03-08 Martin Liska <mliska@suse.cz>
5543 PR gcov-profile/84735
5544 * doc/gcov.texi: Document usage of profile files.
5545 * gcov-io.h: Document changes in the format.
5547 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5551 * dwarf2out.c (struct dw_line_info_table): Update comments for
5553 (FORCE_RESET_NEXT_VIEW): New.
5554 (FORCE_RESETTING_VIEW_P): New.
5555 (RESETTING_VIEW_P): Check for -1 too.
5556 (ZERO_VIEW_P): Likewise.
5557 (new_line_info_table): Force-reset next view.
5558 (dwarf2out_begin_function): Likewise.
5559 (dwarf2out_source_line): Simplify zero_view_p initialization.
5560 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5561 view directly. Omit view when omitting .loc at line 0.
5563 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5565 PR tree-optimization/84740
5566 * tree-switch-conversion.c (process_switch): Call build_constructors
5567 only if info.phi_count is non-zero.
5569 PR tree-optimization/84739
5570 * tree-tailcall.c (find_tail_calls): Check call arguments against
5571 DECL_ARGUMENTS (current_function_decl) rather than
5572 DECL_ARGUMENTS (func) when checking for tail recursion.
5574 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5576 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5577 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5578 Volker Reichelt's entry and add entries for people that perform
5579 GCC fuzzy testing and report numerous bugs.
5581 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5584 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5585 readonly data in sdata, if that is disabled.
5586 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5587 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5588 -mreadonly-in-sdata option.
5590 2018-03-07 Martin Sebor <msebor@redhat.com>
5592 PR tree-optimization/84468
5593 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5594 basic block when looking for nul assignment.
5596 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5599 * except.h (output_function_exception_table): Adjust prototype.
5600 * except.c (output_function_exception_table): Remove FNNAME parameter
5601 and add SECTION parameter. Ouput one part of the table at a time.
5602 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5603 the first part of the exception table and emit unwind directives.
5604 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5605 (i386_pe_seh_cold_init): Likewise.
5606 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5607 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5608 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5609 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5610 * config/i386/winnt.c: Include except.h.
5611 (struct seh_frame_state): Add reg_offset, after_prologue and
5612 in_cold_section fields.
5613 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5614 (i386_pe_seh_cold_init): New function.
5615 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5616 to seh->in_cold_section.
5617 (seh_emit_push): Record the offset of the push.
5618 (seh_emit_save): Record the offet of the save.
5619 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5620 Test seh->after_prologue to disregard the epilogue.
5621 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5622 (i386_pe_end_cold_function): New function.
5624 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5627 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5628 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5631 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5633 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5634 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5637 * multiple_target.c: Include tree-inline.h and intl.h.
5638 (expand_target_clones): Diagnose and fail if node->definition and
5639 !tree_versionable_function_p (node->decl).
5641 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5643 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5645 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5646 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5647 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5649 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5652 * combine.c (try_combine): Use reg_or_subregno instead of handling
5653 just paradoxical SUBREGs and REGs.
5655 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5657 * config/arc/arc.c (arc_finalize_pic): Remove function.
5658 (arc_must_save_register): We use single base PIC register, remove
5659 checks to save/restore the PIC register.
5660 (arc_expand_prologue): Likewise.
5661 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5663 (arc_verify_short): Likewise.
5664 (arc_attr_type): Likewise.
5665 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5666 (walk_stores): Likewise.
5667 (arc_address_cost): Make it static.
5668 (arc_verify_short): Likewise.
5669 (branch_dest): Likewise.
5670 (arc_attr_type): Likewise.
5671 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5672 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5673 (arc_final_prescan_insn): Remove inserting the nops due to
5674 hardware hazards. It is done in reorg step.
5675 (insn_length_variant_t): Remove.
5676 (insn_length_parameters_t): Likewise.
5677 (arc_insn_length_parameters): Likewise.
5678 (arc_get_insn_variants): Likewise.
5679 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5681 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5684 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5687 PR tree-optimization/84687
5688 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5690 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5692 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5694 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5695 Rename to ppc_speculation_barrier.
5696 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5697 __builtin_ppc_speculation_barrier.
5699 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5702 * combine.c (combine_simplify_rtx): Don't try to simplify if
5703 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5706 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5708 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5709 to 32 bytes when compiling for POWER9.
5711 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5714 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5715 regparm >= 3 with no arg reg available also for calls with
5716 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5719 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5721 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5723 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5726 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5728 2018-03-05 Richard Biener <rguenther@suse.de>
5730 PR tree-optimization/84486
5731 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5732 When inserting a __builtin_assume_aligned call set the LHS
5733 SSA name alignment info accordingly.
5735 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5737 PR tree-optimization/84114
5738 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5739 Avoid reassociation of FLOAT_MODE addition.
5741 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5743 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5744 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5745 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5746 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5747 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5748 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5749 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5751 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5752 __builtin_ia32_wbinvd): New builtins.
5753 (SPECIAL_ARGS2): New.
5754 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5755 (SPECIAL_ARGS2): New.
5756 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5757 (ix86_valid_target_attribute_inner_p): Ditto.
5758 (ix86_init_mmx_sse_builtins): Add special_args2.
5759 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5760 TARGET_WBNOINVD_P): New.
5761 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5762 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5763 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5764 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5765 * config/i386/pconfigintrin.h: New file.
5766 * config/i386/wbnoinvdintrin.h: Ditto.
5767 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5768 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5770 2018-03-05 Richard Biener <rguenther@suse.de>
5772 PR tree-optimization/84670
5773 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5775 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5776 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5777 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5778 doesn't grow if all successors (recursively) were visited at least
5781 2018-03-05 Richard Biener <rguenther@suse.de>
5783 PR tree-optimization/84650
5784 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5785 if executed in the loop pipeline.
5787 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5789 * doc/configfiles.texi (Configuration Files): Move info about
5790 conditionalizing $target-protos.h to...
5791 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5792 differs from $target-protos.h.
5794 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5795 Chung-Ju Wu <jasonwucj@gmail.com>
5797 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5798 * config/nds32/nds32-multiple.md (setmemsi): Define.
5799 * config/nds32/nds32-memory-manipulation.c
5800 (nds32_gen_dup_4_byte_to_word_value): New.
5801 (emit_setmem_word_loop): New.
5802 (emit_setmem_byte_loop): New.
5803 (nds32_expand_setmem_loop): New.
5804 (nds32_expand_setmem_loop_v3m): New.
5805 (nds32_expand_setmem_unroll): New.
5806 (nds32_expand_setmem): New.
5808 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5809 Chung-Ju Wu <jasonwucj@gmail.com>
5811 * config/nds32/nds32-memory-manipulation.c
5812 (nds32_emit_load_store): New.
5813 (nds32_emit_post_inc_load_store): New.
5814 (nds32_emit_mem_move): New.
5815 (nds32_emit_mem_move_block): New.
5816 (nds32_expand_movmemsi_loop_unknown_size): New.
5817 (nds32_expand_movmemsi_loop_known_size): New.
5818 (nds32_expand_movmemsi_loop): New.
5819 (nds32_expand_movmemsi_unroll): New.
5820 (nds32_expand_movmemqi): Rename ...
5821 (nds32_expand_movmemsi): ... to this.
5822 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5823 (movmemsi): ... to this.
5824 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5825 (nds32_expand_movmemsi): ... to this.
5827 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5828 Monk Chiang <sh.chiang04@gmail.com>
5829 Chung-Ju Wu <jasonwucj@gmail.com>
5831 * config/nds32/nds32-protos.h
5832 (nds32_expand_load_multiple): New arguments.
5833 (nds32_expand_store_multiple): Ditto.
5834 (nds32_valid_multiple_load_store): Rename ...
5835 (nds32_valid_multiple_load_store_p): ... to this.
5836 * config/nds32/nds32-memory-manipulation.c
5837 (nds32_expand_load_multiple): Refine implementation.
5838 (nds32_expand_store_multiple): Ditto.
5839 * config/nds32/nds32-multiple.md
5840 (load_multiple): Update nds32_expand_load_multiple interface.
5841 (store_multiple): Update nds32_expand_store_multiple interface.
5842 * config/nds32/nds32-predicates.c
5843 (nds32_valid_multiple_load_store): Rename ...
5844 (nds32_valid_multiple_load_store_p): ... to this and refine
5846 * config/nds32/predicates.md
5847 (nds32_load_multiple_and_update_address_operation): New predicate.
5848 (nds32_store_multiple_and_update_address_operation): New predicate.
5850 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5851 Chung-Ju Wu <jasonwucj@gmail.com>
5853 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5854 (combo): New attribute.
5855 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5857 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5859 * config/nds32/nds32.opt: Change -mcmodel= default value.
5861 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5862 Monk Chiang <sh.chiang04@gmail.com>
5863 Chung-Ju Wu <jasonwucj@gmail.com>
5865 * config/nds32/constants.md (unspec_element): New enum.
5866 * config/nds32/constraints.md (Umw): New constraint.
5867 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5868 * config/nds32/nds32-intrinsic.md: Likewise.
5869 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5870 (nds32_valid_smw_lwm_base_p): New.
5871 (nds32_output_smw_single_word): New.
5872 (nds32_output_lmw_single_word): New.
5873 (nds32_expand_unaligned_load): New.
5874 (nds32_expand_unaligned_store): New.
5875 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5876 (nds32_output_smw_single_word): Declare.
5877 (nds32_output_lmw_single_word): Declare.
5878 (nds32_expand_unaligned_load): Declare.
5879 (nds32_expand_unaligned_store): Declare.
5880 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5881 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5882 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5883 NDS32_BUILTIN_UASTORE_DW.
5884 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5887 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5888 Kito Cheng <kito.cheng@gmail.com>
5889 Chung-Ju Wu <jasonwucj@gmail.com>
5891 * config/nds32/nds32-intrinsic.c
5892 (nds32_expand_builtin_null_ftype_reg): Delete.
5893 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5894 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5895 (nds32_read_argument): New.
5896 (nds32_legitimize_target): Ditto.
5897 (nds32_legitimize_argument): Ditto.
5898 (nds32_check_constant_argument): Ditto.
5899 (nds32_expand_unop_builtin): Ditto.
5900 (nds32_expand_unopimm_builtin): Ditto.
5901 (nds32_expand_binop_builtin): Ditto.
5902 (nds32_builtin_decl_impl): Ditto.
5903 (builtin_description): Ditto.
5904 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5905 (nds32_init_builtins_impl): Ditto.
5906 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5907 (nds32_builtin_decl): New.
5908 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5909 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5911 2018-03-02 Jeff Law <law@redhat.com>
5913 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5914 (redundant_insn, fill_simple_delay_slots): Likewise.
5915 (fill_slots_from_thread): Likewise.
5916 * resource.c (mark_referenced_resources): Likewise.
5917 (mark_set_resources, find_dead_or_set_registers): Likewise.
5919 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5921 * substring-locations.h (format_warning_va): Formatting fix for
5923 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5925 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5926 * substring-locations.c: Include intl.h.
5927 (format_warning_va): Turned into small wrapper around
5928 format_warning_n_va, renamed to ...
5929 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5930 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5932 (format_warning_at_substring_n): New function.
5933 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5934 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5935 format_warning_at_substring with just a shorter name instead of
5936 const function pointer.
5937 (fmtwarn_n): New function.
5938 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5939 appropriate, get rid of all the fmtstr temporaries, move conditionals
5940 with G_() wrapped string literals directly into fmtwarn arguments,
5941 cast dir.len to (int), formatting fixes.
5943 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5945 * doc/invoke.texi: Remove "Cilk Plus" references.
5947 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5948 Richard Biener <rguenther@suse.de>
5951 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5952 for error or warning attributes if CALL_FROM_THUNK_P is set.
5955 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5958 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5959 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5962 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5963 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5965 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5966 instead of -1U in last predictors element's probability member.
5968 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5971 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5972 arguments if they are comparable.
5974 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5976 PR tree-optimization/84634
5977 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5978 masks and masked_loop_p with a single loop_masks, making sure it's
5979 null for bb vectorization.
5981 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5983 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5984 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5985 loop->force_vectorize to check whether there is no alias.
5987 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5990 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5992 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5994 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5995 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5996 instead of a loop around prev_real_insn.
5997 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6001 * config/i386/i386.c (ix86_print_operand): Use conditional
6002 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6005 2018-03-02 Richard Biener <rguenther@suse.de>
6007 PR tree-optimization/84427
6008 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6009 (bitmap_set_subtract_values): Rewrite to handle multiple
6012 (prune_clobbered_mems): Likewise.
6013 (phi_translate): Take edge instead of pred/phiblock.
6014 (phi_translate_1): Likewise.
6015 (phi_translate_set): Likewise. Insert all translated
6016 exprs for a value into the set, keeping possibly multiple
6017 expressions per value.
6018 (compute_antic_aux): Adjust for phi_translate changes.
6019 When intersecting union the expressions and prune those
6020 not in the final value set, keeping possibly multiple
6021 expressions per value. Do not use value-insertion
6022 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6023 all expressions. Add verification that the value-sets
6024 only shrink during iteration.
6025 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6026 (do_pre_regular_insertion): Likewise.
6027 (do_pre_partial_partial_insertion): Likewise.
6029 2018-03-02 Richard Biener <rguenther@suse.de>
6032 * config/darwin.c (saved_debug_info_level): New static global.
6033 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6034 (darwin_asm_lto_end): Restore debug info generation settings.
6036 2018-03-01 Martin Liska <mliska@suse.cz>
6039 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6042 2018-03-01 Richard Biener <rguenther@suse.de>
6045 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6046 type annotation with LTO when debug was disabled at compile-time.
6048 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6050 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6052 (mips_final_postscan_insn): Likewise.
6054 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6056 PR rtl-optimization/84528
6057 * alias.c (init_alias_target): Add commentary.
6058 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6059 a unique base value if the frame pointer is not eliminated
6060 to the stack pointer.
6062 2018-03-01 Tom de Vries <tom@codesourcery.com>
6064 PR rtl-optimization/83327
6065 * lra-int.h (hard_regs_spilled_into): Declare.
6066 * lra.c (hard_regs_spilled_into): Define.
6067 (init_reg_info): Init hard_regs_spilled_into.
6068 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6069 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6070 (process_bb_lives): Handle hard_regs_spilled_into.
6071 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6072 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6074 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6076 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6077 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6078 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6079 * config/rs6000/aix72.h: New file.
6081 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6083 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6084 instead of warning_at with conditional singular and plural messages
6088 * stor-layout.c (update_alignment_for_field): For
6089 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6090 && !DECL_PACKED (field), do the alignment update, just use
6091 only desired_align instead of MAX (type_align, desired_align)
6093 (place_field): Don't do known_align < desired_align handling
6094 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6095 is non-NULL, instead do it after rli->prev_field handling and
6096 only if not within a bitfield word. For DECL_PACKED (field)
6097 use type_align of BITS_PER_UNIT.
6099 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6101 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6102 superfluous parentheses and trailing spaces.
6104 2018-02-28 Richard Biener <rguenther@suse.de>
6106 PR tree-optimization/84584
6107 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6108 SCoPs with fake exit edge.
6110 2018-02-28 Martin Liska <mliska@suse.cz>
6113 * timevar.c (timer::print): Fix format to properly print 100%
6116 2018-02-28 Richard Biener <rguenther@suse.de>
6119 * genmatch.c (capture_info::walk_match): Do not mark
6120 captured expressions without operands as expr_p given
6121 they act more like predicates and should be subject to
6122 "lost tail" side-effect preserving.
6124 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6126 PR rtl-optimization/81611
6127 * auto-inc-dec.c (attempt_change): Move dead note from
6128 mem_insn if it's the next use of regno
6129 (find_address): Take address use of reg holding
6130 non-incremented value. Add parm to limit search to the named
6132 (merge_in_block): Attempt to use a mem insn that is the next
6133 use of the original regno.
6135 2018-02-27 Martin Sebor <msebor@redhat.com>
6138 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6139 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6141 2018-02-27 Martin Sebor <msebor@redhat.com>
6143 PR translation/84207
6144 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6145 n argument to unsigned HOST_WIDE_INT.
6146 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6147 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6148 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6149 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6151 2018-02-27 Richard Biener <rguenther@suse.de>
6153 PR tree-optimization/84512
6154 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6155 Do not use the estimate returned from record_stmt_cost for
6156 the scalar iteration cost but sum properly using add_stmt_cost.
6158 2018-02-27 Richard Biener <rguenther@suse.de>
6160 PR tree-optimization/84466
6161 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6162 Adjust last change to less strictly validate use operands.
6164 2018-02-27 Martin Liska <mliska@suse.cz>
6166 PR gcov-profile/84548
6167 * gcov.c (process_file): Allow partial overlap and consider it
6168 also as group functions.
6169 (output_lines): Properly calculate range of lines for a group.
6171 2018-02-27 Martin Liska <mliska@suse.cz>
6173 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6174 'ggc' suffixes. Change first column width.
6175 (timer::print): Fix formatting of the column.
6177 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6179 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6180 preserve inline entry blocks for the sake of debug inline
6181 entry point markers alone.
6182 (remove_unused_locals): Suggest in comments a better place to
6183 force the preservation of inline entry blocks that are
6184 otherwise unused, but do not preserve them.
6186 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6188 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6190 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6193 * config/i386/constraints.md (Bs): Replace
6194 ix86_indirect_branch_register with
6195 TARGET_INDIRECT_BRANCH_REGISTER.
6197 * config/i386/i386.md (indirect_jump): Likewise.
6198 (tablejump): Likewise.
6199 (*sibcall_memory): Likewise.
6200 (*sibcall_value_memory): Likewise.
6201 Peepholes of indirect call and jump via memory: Likewise.
6202 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6203 (*sibcall_value_GOT_32): Likewise.
6204 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6205 (GOT_memory_operand): Likewise.
6206 (call_insn_operand): Likewise.
6207 (sibcall_insn_operand): Likewise.
6208 (GOT32_symbol_operand): Likewise.
6209 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6211 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6213 PR rtl-optimization/83496
6214 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6215 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6216 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6217 redundant insn, if any.
6218 (relax_delay_slots): Likewise.
6219 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6221 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6223 PR tree-optimization/83965
6224 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6225 that grouped statements are part of a reduction chain. Return
6226 true if the statement is not marked as a reduction itself but
6228 (vect_recog_dot_prod_pattern): Don't check whether the statement
6229 is part of a group here.
6230 (vect_recog_sad_pattern): Likewise.
6231 (vect_recog_widen_sum_pattern): Likewise.
6233 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6236 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6238 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6241 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6243 (ix86_output_indirect_function_return): New prototype.
6244 (ix86_split_simple_return_pop_internal): Likewise.
6245 * config/i386/i386.c (indirect_return_via_cx): New.
6246 (indirect_return_via_cx_bnd): Likewise.
6247 (indirect_thunk_name): Handle return va CX_REG.
6248 (output_indirect_thunk_function): Create alias for
6249 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6250 (ix86_output_indirect_jmp): Remove the bool argument.
6251 (ix86_output_indirect_function_return): New function.
6252 (ix86_split_simple_return_pop_internal): Likewise.
6253 * config/i386/i386.md (*indirect_jump): Don't pass false
6254 to ix86_output_indirect_jmp.
6255 (*tablejump_1): Likewise.
6256 (simple_return_pop_internal): Change it to define_insn_and_split.
6257 Call ix86_split_simple_return_pop_internal to split it for
6259 (simple_return_indirect_internal): Call
6260 ix86_output_indirect_function_return instead of
6261 ix86_output_indirect_jmp.
6263 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6266 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6267 memset and value initialization afterwards.
6269 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6271 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6273 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6276 * common/config/aarch64/aarch64-common.c
6277 (aarch_option_optimization_table[]): Switch
6278 off fomit-frame-pointer
6280 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6281 Chung-Ju Wu <jasonwucj@gmail.com>
6283 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6285 (store_multiple): Ditto.
6287 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6289 * config.gcc: Add --with-cpu support for nds32 target.
6290 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6291 * config/nds32/nds32.opt: Add -mcpu= option.
6293 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6295 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6296 isel=yes): Warn for these deprecated options.
6298 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6300 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6301 ISA_2_5_MASKS_EMBEDDED.
6303 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6305 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6306 p->max as pointers rather than using iterative_hash_expr.
6308 2018-02-23 Carl Love <cel@us.ibm.com>
6310 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6311 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6313 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6314 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6315 P8V_BUILTIN_VEC_VUNSIGNED2.
6317 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6320 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6321 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6323 * lra-constraints.c (curr_insn_transform): Set up
6324 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6326 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6327 * lra-eliminations.c (spill_pseudos): Ditto.
6328 (process_insn_for_elimination): Ditto.
6329 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6330 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6331 LRA_NON_CLOBBERED_ALT.
6333 2018-02-22 Martin Sebor <msebor@redhat.com>
6335 PR tree-optimization/84480
6336 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6337 to maybe_diag_stxncpy_trunc. Call it.
6338 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6339 from gimple_fold_builtin_strcpy. Print inlining stack.
6340 (handle_builtin_stxncpy): Print inlining stack.
6341 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6343 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6346 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6347 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6348 and -fcheck-pointer-bounds are used together.
6349 (indirect_thunk_prefix): New enum.
6350 (indirect_thunk_need_prefix): New function.
6351 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6352 "_nt" instead of "_bnd" for NOTRACK prefix.
6353 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6354 (output_indirect_thunk_function): Likewise.
6356 (ix86_code_end): Update output_indirect_thunk_function calls.
6357 (ix86_output_indirect_branch_via_reg): Replace
6358 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6359 (ix86_output_indirect_branch_via_push): Likewise.
6360 (ix86_output_function_return): Likewise.
6361 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6362 incompatible with -fcf-protection=branch and
6363 -fcheck-pointer-bounds.
6365 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6368 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6369 Change gcc_assert call to output_operand_lossage.
6371 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6373 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6375 2018-02-22 DJ Delorie <dj@redhat.com>
6376 Sebastian Perta <sebastian.perta@renesas.com>
6377 Oleg Endo <olegendo@gcc.gnu.org>
6379 * config/rx/rx.c (rx_rtx_costs): New function.
6380 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6382 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6384 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6386 2018-02-22 Martin Liska <mliska@suse.cz>
6389 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6390 Add "native" as a possible value.
6392 2018-02-22 Martin Liska <mliska@suse.cz>
6395 * config/i386/i386.c (ix86_option_override_internal):
6396 Add "native" as a possible value for -march and -mtune.
6398 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6401 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6402 to all type variants.
6404 PR tree-optimization/84503
6405 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6406 width as info->bitpos + info->bitsize - start.
6407 (merged_store_group::merge_overlapping): Simplify width computation.
6408 (check_no_overlap): New function.
6409 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6410 start + width and last_order of the group, fail if check_no_overlap
6412 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6413 to group if check_no_overlap fails.
6415 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6417 * config/rs6000/altivec.md: Delete contraint arguments to
6418 define_expand, define_split, and define_peephole2, and in
6419 define_insn_and_split if always unused.
6420 * config/rs6000/darwin.md: Ditto.
6421 * config/rs6000/dfp.md: Ditto.
6422 * config/rs6000/rs6000.md: Ditto.
6423 * config/rs6000/sync.md: Ditto.
6424 * config/rs6000/vector.md: Ditto.
6425 * config/rs6000/vsx.md: Ditto.
6427 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6429 * config/rs6000/altivec.md: Write output control strings as braced
6430 blocks instead of double-quoted strings.
6431 * config/rs6000/darwin.md: Ditto.
6432 * config/rs6000/rs6000.md: Ditto.
6433 * config/rs6000/vector.md: Ditto.
6434 * config/rs6000/vsx.md: Ditto.
6436 2018-02-21 Jason Merrill <jason@redhat.com>
6438 PR c++/84314 - ICE with templates and fastcall attribute.
6439 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6441 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6443 * ipa-cp.c (determine_versionability): Fix comment typos.
6445 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6448 * ipa-cp.c (determine_versionability): Do not version functions caling
6451 2018-02-21 Martin Liska <mliska@suse.cz>
6454 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6455 Add "native" as a possible value.
6456 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6457 the macro when native cpu detection is available.
6459 2018-02-21 Martin Liska <mliska@suse.cz>
6462 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6463 Add "native" as a possible value.
6464 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6465 when native cpu detection is available.
6467 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6468 Martin Sebor <msebor@redhat.com>
6470 PR tree-optimization/84478
6471 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6473 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6474 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6475 support which is conservatively correct, for 2 only stay conservative
6476 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6477 argument to the 2 argument get_range_strlen, adjust 6 arg
6478 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6480 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6481 (gimple_fold_builtin_strlen): Pass true as last argument to
6484 2018-02-20 Martin Sebor <msebor@redhat.com>
6487 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6488 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6489 (builtin_memref::builtin_memref): Factor out parts into
6490 set_base_and_offset and call it.
6492 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6495 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6496 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6497 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6498 search at the associated MODE_INT.
6500 2018-02-20 Jeff Law <law@redhat.com>
6503 PR tree-optimization/81592
6505 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6506 for range data rather than using global data.
6507 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6508 range data rather than using global data.
6509 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6510 pass it to children as needed.
6511 (struct directive::fmtresult): Similarly.
6512 (struct directive::set_width): Similarly.
6513 (struct directive::set_precision): Similarly.
6514 (format_integer, format_directive, parse_directive): Similarly.
6515 (format_none): Accept unnamed vr_values parameter.
6516 (format_percent, format_floating, format_character): Similarly.
6517 (format_string, format_plain): Similarly.
6518 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6519 the EVRP range analyzer for range data rather than using global data.
6520 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6521 gimple-ssa-evrp-analyze.h
6522 (class sprintf_dom_walker): Add after_dom_children member function.
6523 Add evrp_range_analyzer member.
6524 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6525 range analyzer as needed.
6526 (sprintf_dom_walker::after_dom_children): New member function.
6527 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6529 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6530 (evrp_range_analyzer::pop_to_marker): Likewise.
6532 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6534 PR tree-optimization/84419
6535 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6536 with the required type if its current type is compatible but
6539 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6542 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6543 after vectorization.
6545 2018-02-20 Martin Liska <mliska@suse.cz>
6548 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6549 possible values if we don't have a hint.
6551 2018-02-20 Martin Liska <mliska@suse.cz>
6555 * final.c (shorten_branches): Build align_tab array with one
6557 * opts.c (finish_options): Add alignment option limit check.
6558 (MAX_CODE_ALIGN): Likewise.
6559 (MAX_CODE_ALIGN_VALUE): Likewise.
6560 * doc/invoke.texi: Document maximum allowed option value for
6561 all -falign-* options.
6563 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6566 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6567 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6568 * var-tracking.c (emit_note_insn_var_location): Remove all references
6569 to NOTE_INSN_CALL_ARG_LOCATION.
6570 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6571 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6572 Use copy_rtx_if_shared.
6573 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6574 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6575 (dwarf2out_var_location): Remove handling of
6576 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6578 * final.c (final_scan_insn): Remove all references to
6579 NOTE_INSN_CALL_ARG_LOCATION.
6580 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6581 before dumping final insns.
6582 * except.c (emit_note_eh_region_end): Remove all references to
6583 NOTE_INSN_CALL_ARG_LOCATION.
6584 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6585 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6586 * config/arc/arc.c (hwloop_optimize): Likewise.
6587 * config/arm/arm.c (create_fix_barrier): Likewise.
6588 * config/s390/s390.c (s390_chunkify_start): Likewise.
6589 * config/sh/sh.c (find_barrier): Likewise.
6590 * config/i386/i386.c (rest_of_insert_endbranch,
6591 ix86_seh_fixup_eh_fallthru): Likewise.
6592 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6593 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6594 * config/frv/frv.c (frv_function_prologue): Likewise.
6595 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6597 (note_outside_basic_block_p): Remove all references to
6598 NOTE_INSN_CALL_ARG_LOCATION.
6599 * gengtype.c (adjust_field_rtx_def): Likewise.
6600 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6602 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6603 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6606 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6609 PR tree-optimization/84452
6610 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6611 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6614 2018-02-19 Martin Liska <mliska@suse.cz>
6617 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6619 2018-02-19 Martin Liska <mliska@suse.cz>
6620 Richard Sandiford <richard.sandiford@linaro.org>
6622 PR tree-optimization/82491
6623 * gimple-fold.c (get_base_constructor): Make earlier bail out
6626 2018-02-19 Carl Love <cel@us.ibm.com>
6628 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6629 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6631 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6632 P8V_BUILTIN_VEC_NEG.
6634 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6636 * config/rl78/rl78.md (movdf): New define expand.
6638 2018-02-19 Martin Liska <mliska@suse.cz>
6641 * doc/invoke.texi: Fix typo.
6642 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6644 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6646 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6647 handle rs6000_single_float and rs6000_double_float specially for
6650 2018-02-16 Jeff Law <law@redhat.com>
6652 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6655 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6658 * ipa-inline.c (inline_small_functions): Fix a typo.
6660 2018-02-16 Nathan Sidwell <nathan@acm.org>
6662 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6664 2018-02-16 Carl Love <cel@us.ibm.com>
6666 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6667 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6668 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6669 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6670 expansion to P8V_BUILTIN_VEC_FLOAT2.
6672 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6674 PR rtl-optimization/70023
6675 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6676 src_regno into account.
6678 2018-02-16 Carl Love <cel@us.ibm.com>
6680 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6681 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6682 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6683 * config/rs6000/rs6000.c: Remove case statements for
6684 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6685 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6686 and P9V_BUILTIN_VEC_VINSERT4B.
6687 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6688 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6689 * config/rs6000/vsx.md:
6690 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6693 2018-02-16 Carl Love <cel@us.ibm.com>
6695 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6697 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6699 * config/rs6000/rs6000-c.c: Add the definitions for
6700 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6701 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6702 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6703 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6704 definition for insert4b and define insn *insert3b_internal.
6705 * doc/extend.texi: Add documentation for vec_extract4b.
6707 2018-02-16 Nathan Sidwell <nathan@acm.org>
6709 * doc/extend.texi (Backwards Compatibility): Mention friend
6710 injection. Note for-scope is deprecated.
6711 * doc/invoke.texi (-ffriend-injection): Deprecate.
6713 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6715 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6716 that moved to I2, also allow destinations that are a paradoxical
6717 subreg (instead of a normal reg).
6719 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6722 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6725 2018-02-16 Richard Biener <rguenther@suse.de>
6727 PR tree-optimization/84037
6728 PR tree-optimization/84016
6730 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6731 Adjust vec_construct for the fact we need additional higher latency
6732 128bit inserts for AVX256 and AVX512 vector builds.
6733 (ix86_add_stmt_cost): Scale vector construction cost for
6736 2018-02-16 Richard Biener <rguenther@suse.de>
6738 PR tree-optimization/84417
6739 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6740 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6741 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6743 2018-02-16 Martin Liska <mliska@suse.cz>
6746 * internal-fn.def (ASAN_CHECK): Set proper flags.
6747 (ASAN_MARK): Likewise.
6749 2018-02-16 Julia Koval <julia.koval@intel.com>
6751 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6752 from PTA_CANNONLAKE.
6754 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6757 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6758 Use ++iter rather than iter++ for std::list iterators.
6759 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6760 defer deleting them until all nodes in the forest are processed. Do
6761 free even leaf nodes. Change to_process into auto_vec.
6764 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6765 * vec.h (vec_default_construct): Use memset instead of placement new
6766 if BROKEN_VALUE_INITIALIZATION is defined.
6767 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6768 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6771 PR rtl-optimization/83723
6772 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6773 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6774 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6776 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6778 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6780 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6782 PR rtl-optimization/81443
6783 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6784 from inner REGs to paradoxical SUBREGs.
6786 2018-02-16 Richard Biener <rguenther@suse.de>
6788 PR tree-optimization/84399
6789 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6790 For operands we can analyze at their definition make sure we can
6791 analyze them at each use as well.
6793 2018-02-16 Richard Biener <rguenther@suse.de>
6795 PR tree-optimization/84190
6796 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6797 volatile accesses if the decl isn't volatile.
6799 2018-02-15 Jason Merrill <jason@redhat.com>
6801 PR c++/84314 - ICE with templates and fastcall attribute.
6802 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6803 TYPE_CANONICAL on an existing type.
6805 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6807 PR tree-optimization/84383
6808 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6809 dstoff nor call operand_equal_p if dstbase is NULL.
6811 PR tree-optimization/84334
6812 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6813 also a CONSTANT_CLASS_P, punt.
6815 2018-02-14 Jim Wilson <jimw@sifive.com>
6817 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6818 first SMALL_OPERAND check. New local min_second_step. Move assert
6819 to where locals are set. Add TARGET_RVC support.
6820 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6822 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6824 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6826 2018-02-14 Martin Sebor <msebor@redhat.com>
6828 PR tree-optimization/83698
6829 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6830 arrays constrain the offset range to their bounds.
6831 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6832 (builtin_access::overlap): Avoid setting the size of overlap if it's
6834 (maybe_diag_overlap): Also consider arrays when deciding what values
6835 of offsets to include in diagnostics.
6837 2018-02-14 Martin Sebor <msebor@redhat.com>
6840 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6841 that correspond to the kind of a declaration.
6843 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6846 * config/pa/pa.md: Load address of PIC label using the linkage table
6847 if the label is nonlocal.
6849 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6851 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6852 warning message if user requests -maltivec=be.
6853 * doc/invoke.texi: Document deprecation of -maltivec=be.
6855 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6858 * config/rs6000/rs6000-c.c: Update definitions for
6859 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6860 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6862 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6865 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6866 add _get_ssp intrinsics. Remove argument from
6867 __builtin_ia32_rdssp[d|q].
6868 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6869 * config/i386/i386-builtin.def: Remove argument from
6870 __builtin_ia32_rdssp[d|q].
6871 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6872 ix86_expand_special_args_builtin for _rdssp[d|q].
6873 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6874 Clear register before usage.
6875 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6876 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6878 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6880 PR tree-optimization/84357
6881 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6882 operand 1 of an ARRAY_REF too.
6884 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6887 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6888 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6890 (set_of_reg): New struct.
6891 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6892 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6893 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6895 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6896 Split into bitclr, bitset, bitinvert patterns if appropriate.
6897 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6898 use rx_fuse_in_memory_bitop.
6899 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6900 to named insn, correct maximum insn length.
6902 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6905 * machmode.def: Define a complex mode for PARTIAL_INT.
6906 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6908 * doc/rtl.texi: Document CSPImode.
6909 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6911 (msp430_hard_regno_nregs_with_padding): Likewise.
6913 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6916 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6918 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6920 PR rtl-optimization/84169
6921 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6922 we generated a parallel as new i3 and we split that to new i2 and i3
6923 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6924 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6925 those to i2, not i1. Partially rewrite this scan code.
6927 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6930 * stor-layout.c (place_field): For variable length fields, adjust
6931 offset_align afterwards not just based on the field's alignment,
6932 but also on the size.
6935 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6936 of exps and logs in the use_exp2 case.
6938 2018-02-13 Jeff Law <law@redhat.com>
6940 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6943 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6946 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6950 * common.opt (gas-loc-support, gas-locview-support): New.
6951 (ginline-points, ginternal-reset-location-views): New.
6952 * doc/invoke.texi: Document them. Use @itemx where intended.
6953 (gvariable-location-views): Adjust.
6954 * target.def (reset_location_view): New.
6955 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6956 (TARGET_RESET_LOCATION_VIEW): New.
6957 * doc/tm.texi: Rebuilt.
6958 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6959 (dwarf2out_default_as_locview_support): New.
6960 (output_asm_line_debug_info): Use option variables.
6961 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6962 (output_loc_list): Likewise.
6963 (add_high_low_attributes): Check option variables.
6964 Don't output entry view attribute in strict mode.
6965 (gen_inlined_subroutine_die): Check option variables.
6966 (dwarf2out_inline_entry): Likewise.
6967 (init_sections_and_labels): Likewise.
6968 (dwarf2out_early_finish): Likewise.
6969 (maybe_reset_location_view): New, from...
6970 (dwarf2out_var_location): ... here. Call it.
6971 * debug.h (dwarf2out_default_as_loc_support): Declare.
6972 (dwarf2out_default_as_locview_support): Declare.
6973 * hooks.c (hook_int_rtx_insn_0): New.
6974 * hooks.h (hook_int_rtx_insn_0): Declare.
6975 * toplev.c (process_options): Take -gas-loc-support and
6976 -gas-locview-support from dwarf2out. Enable
6977 -gvariable-location-views by default only with locview
6978 assembler support. Enable -ginternal-reset-location-views by
6979 default only if the target defines the corresponding hook.
6980 Enable -ginline-points by default if location views are
6981 enabled; force it disabled if statement frontiers are
6983 * tree-inline.c (expand_call_inline): Check option variables.
6984 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6986 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6988 PR tree-optimization/84321
6989 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6990 handling. Also check whether the anti-range contains any values
6991 that satisfy the mask; switch to a VR_RANGE if not.
6993 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6996 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6998 2018-02-13 Martin Jambor <mjambor@suse.cz>
7001 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7002 of call statements, also set location of a load to a temporary.
7004 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7006 * config/rl78/rl78.c (add_vector_labels): New function.
7007 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7008 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7009 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7010 which checks that no arguments are passed.
7011 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7012 * doc/extend.texi: Documentation for the new attribute.
7014 2018-02-13 Andreas Schwab <schwab@suse.de>
7016 * config/riscv/linux.h (CPP_SPEC): Define.
7018 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7021 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7022 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7023 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7024 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7025 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7026 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7027 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7028 temporarily for AES and PCLMUL builtins.
7030 PR tree-optimization/84339
7031 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7032 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7036 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7037 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7038 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7040 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7042 * omp-simd-clone.h: New file.
7043 * omp-simd-clone.c: Include omp-simd-clone.h.
7044 (expand_simd_clones): No longer static.
7045 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7046 cgraph.h and omp-simd-clone.h.
7047 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7048 (vect_recog_widen_shift_pattern): Formatting fix.
7049 (vect_pattern_recog_1): Don't check optab for calls.
7052 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7053 operands[2] into a REG before using gen_lowpart on it.
7055 2018-02-12 Jeff Law <law@redhat.com>
7058 * config/sh/sh.c (find_barrier): Consider a sibling call
7061 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7062 successfully back substituting a reg.
7064 2018-02-12 Richard Biener <rguenther@suse.de>
7066 PR tree-optimization/84037
7067 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7068 parameter, move visited init to caller.
7069 (vect_slp_analyze_operations): Separate cost from validity
7070 check, initialize visited once for all instances.
7071 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7073 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7075 (vect_model_promotion_demotion_cost): Likewise.
7076 (vectorizable_bswap): Guard cost modeling with !slp_node
7077 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7079 (vectorizable_call): Likewise.
7080 (vectorizable_conversion): Likewise.
7081 (vectorizable_assignment): Likewise.
7082 (vectorizable_shift): Likewise.
7083 (vectorizable_operation): Likewise.
7084 (vectorizable_store): Likewise.
7085 (vectorizable_load): Likewise.
7086 (vectorizable_condition): Likewise.
7087 (vectorizable_comparison): Likewise.
7089 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7092 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7093 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7094 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7096 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7099 * params.def (inline-min-speedup): Increase from 8 to 15.
7100 (max-inline-insns-auto): Decrease from 40 to 30.
7101 * ipa-split.c (consider_split): Add some buffer for function to
7102 be considered inlining candidate.
7103 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7106 2018-02-12 Richard Biener <rguenther@suse.de>
7108 PR tree-optimization/84037
7109 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7110 matched stmts if we cannot swap the non-matched ones.
7112 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7114 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7115 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7116 _mm_maskz_scalef_round_ss): New intrinsics.
7117 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7118 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7119 __builtin_ia32_scalefss_round): Remove.
7120 (__builtin_ia32_scalefsd_mask_round,
7121 __builtin_ia32_scalefss_mask_round): New intrinsics.
7122 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7123 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7124 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7125 "<round_constraint>")): Changed to ...
7126 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7127 "<round_scalar_constraint>")): ... this.
7128 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7129 %0, %1, %2<round_op3>}"): Changed to ...
7130 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7131 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7132 %2<round_scalar_mask_op3>}"): ... this.
7133 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7135 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7137 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7138 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7139 (_mm_maskz_sqrt_round_ss): New intrinsics.
7140 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7141 (__builtin_ia32_sqrtsd_mask_round)
7142 (__builtin_ia32_sqrtss_mask_round): New builtins.
7143 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7144 (__builtin_ia32_sqrtss_round): Remove.
7145 (__builtin_ia32_sqrtsd_mask_round)
7146 (__builtin_ia32_sqrtss_mask_round): New builtins.
7147 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7148 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7149 ((match_operand:VF_128 1 "vector_operand"
7150 "xBm,<round_constraint>")): Changed to ...
7151 ((match_operand:VF_128 1 "vector_operand"
7152 "xBm,<round_scalar_constraint>")): ... this.
7153 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7154 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7155 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7156 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7157 %<iptr>1<round_scalar_mask_op3>}): ... this.
7158 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7159 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7161 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7164 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7165 Cast vec_cmpeq result to correct type.
7166 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7167 Cast vec_cmpgt result to correct type.
7169 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7171 * final.c (final_scan_insn_1): Renamed from...
7172 (final_scan_insn): ... this. New wrapper, to recover
7173 seen from the outermost call in recursive ones.
7174 * config/sparc/sparc.c (output_return): Drop seen from call.
7175 (output_sibcall): Likewise.
7176 * config/visium/visium.c (output_branch): Likewise.
7178 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7180 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7183 2018-02-10 Alan Modra <amodra@gmail.com>
7186 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7187 Specify LR as an input.
7189 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7192 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7193 remove_member_access_dummy_vars): New functions.
7194 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7195 lower_omp_1, execute_lower_omp): Use them.
7197 PR rtl-optimization/84308
7198 * shrink-wrap.c (spread_components): Release todo vector.
7200 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7202 PR rtl-optimization/57193
7203 * ira-color.c (struct allocno_color_data): Add member
7204 conflict_allocno_hard_prefs.
7205 (update_conflict_allocno_hard_prefs): New.
7206 (bucket_allocno_compare_func): Add a preference based on
7207 conflict_allocno_hard_prefs.
7208 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7209 (color_allocnos): Remove a dead code. Initiate
7210 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7212 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7215 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7216 constraint from =wa to wa. Avoid a subreg on the output operand,
7217 instead use a pseudo and subreg it in a move.
7218 (p9_xxbrd_<mode>): Changed to ...
7219 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7220 (p9_xxbrd_v2df): New expander.
7221 (p9_xxbrw_<mode>): Changed to ...
7222 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7223 (p9_xxbrw_v4sf): New expander.
7225 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7227 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7229 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7232 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7233 multiply in 32-bit mode.
7234 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7235 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7238 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7240 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7241 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7242 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7243 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7245 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7248 * dwarf2out.c (is_trivial_indirect_ref): New function.
7249 (dwarf2out_late_global_decl): Do not generate a location
7250 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7251 and that are not defined in the current unit.
7253 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7255 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7256 instead of a libcall for UNORDERED.
7258 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7261 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7262 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7264 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7267 * config/s390/s390.c (s390_set_current_function): Invoke
7268 s390_indirect_branch_settings also if fndecl didn't change.
7270 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7272 * config/rs6000/rs6000.md (blockage): Set length to zero.
7274 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7276 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7278 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7281 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7282 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7286 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7287 PARALLEL incoming that failed vt_get_decl_and_offset check.
7290 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7291 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7293 (get_variable_section): For decls in named .bss* sections pass true as
7294 second argument to bss_initializer_p.
7296 2018-02-09 Marek Polacek <polacek@redhat.com>
7297 Jakub Jelinek <jakub@redhat.com>
7300 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7301 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7302 Sync some changes from cxx_fold_indirect_ref.
7304 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7306 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7308 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7309 (BLOCK_INLINE_ENTRY_LABEL): New.
7310 (dwarf2out_var_location): Disregard inline entry markers.
7311 (inline_entry_data): New struct.
7312 (inline_entry_data_hasher): New hashtable type.
7313 (inline_entry_data_hasher::hash): New.
7314 (inline_entry_data_hasher::equal): New.
7315 (inline_entry_data_table): New variable.
7316 (add_high_low_attributes): Add DW_AT_entry_pc and
7317 DW_AT_GNU_entry_view attributes if a pending entry is found
7318 in inline_entry_data_table. Add old entry_pc attribute only
7319 if debug nonbinding markers are disabled.
7320 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7321 markers are enabled.
7322 (block_within_block_p, dwarf2out_inline_entry): New.
7323 (dwarf2out_finish): Check that no entries remained in
7324 inline_entry_data_table.
7325 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7326 (final_scan_insn, notice_source_line): Likewise.
7327 (rest_of_clean_state): Skip inline entry markers.
7328 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7330 * gimple.c (gimple_build_debug_inline_entry): New.
7331 * gimple.h (enum gimple_debug_subcode): Add
7332 GIMPLE_DEBUG_INLINE_ENTRY.
7333 (gimple_build_debug_inline_entry): Declare.
7334 (gimple_debug_inline_entry_p): New.
7335 (gimple_debug_nonbind_marker_p): Adjust.
7336 * insn-notes.def (INLINE_ENTRY): New.
7337 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7338 inline entry marker notes.
7339 (print_insn): Likewise.
7340 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7341 (INSN_DEBUG_MARKER_KIND): Likewise.
7342 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7343 * tree-inline.c (expand_call_inline): Build and insert
7344 debug_inline_entry stmt.
7345 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7346 inline entry blocks early, if nonbind markers are enabled.
7347 (dump_scope_block): Dump fragment info.
7348 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7349 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7350 (gimple_build_debug_inline_entry): New.
7351 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7352 Enable/disable inline entry points too.
7353 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7354 (DEBUG_INSN): Describe inline entry markers.
7356 * common.opt (gvariable-location-views): New.
7357 (gvariable-location-views=incompat5): New.
7358 * config.in: Rebuilt.
7359 * configure: Rebuilt.
7360 * configure.ac: Test assembler for view support.
7361 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7362 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7363 * dwarf2out.c (var_loc_view): New typedef.
7364 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7365 (dwarf2out_locviews_in_attribute): New.
7366 (dwarf2out_locviews_in_loclist): New.
7367 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7368 (enum dw_line_info_opcode): Add LI_adv_address.
7369 (struct dw_line_info_table): Add view.
7370 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7371 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7372 (zero_view_p): New variable.
7373 (ZERO_VIEW_P): New macro.
7374 (output_asm_line_debug_info): New.
7375 (struct var_loc_node): Add view.
7376 (add_AT_view_list, AT_loc_list): New.
7377 (add_var_loc_to_decl): Add view param. Test it against last.
7378 (new_loc_list): Add view params. Record them.
7379 (AT_loc_list_ptr): Handle loc and view lists.
7380 (view_list_to_loc_list_val_node): New.
7381 (print_dw_val): Handle dw_val_class_view_list.
7382 (size_of_die): Likewise.
7383 (value_format): Likewise.
7384 (loc_list_has_views): New.
7385 (gen_llsym): Set vl_symbol too.
7386 (maybe_gen_llsym, skip_loc_list_entry): New.
7387 (dwarf2out_maybe_output_loclist_view_pair): New.
7388 (output_loc_list): Output view list or entries too.
7389 (output_view_list_offset): New.
7390 (output_die): Handle dw_val_class_view_list.
7391 (output_dwarf_version): New.
7392 (output_compilation_unit_header): Use it.
7393 (output_skeleton_debug_sections): Likewise.
7394 (output_rnglists, output_line_info): Likewise.
7395 (output_pubnames, output_aranges): Update version comments.
7396 (output_one_line_info_table): Output view numbers in asm comments.
7397 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7398 Call maybe_gen_llsym.
7399 (loc_list_from_tree_1): Adjust.
7400 (add_AT_location_description): Create view list attribute if
7401 needed, check it's absent otherwise.
7402 (convert_cfa_to_fb_loc_list): Adjust.
7403 (maybe_emit_file): Call output_asm_line_debug_info for test.
7404 (dwarf2out_var_location): Reset views as needed. Precompute
7405 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7406 attribute. Set view.
7407 (new_line_info_table): Reset next view.
7408 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7409 (dwarf2out_source_line): Likewise. Output view resets and labels to
7410 the assembler, or select appropriate line info opcodes.
7411 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7412 (optimize_string_length): Catch it. Adjust.
7413 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7414 dw_val_class_view_list, and remove it if no longer needed.
7415 (hash_loc_list): Hash view numbers.
7416 (loc_list_hasher::equal): Compare them.
7417 (optimize_location_lists): Check whether a view list symbol is
7418 needed, and whether the locview attribute is present, and
7419 whether they match. Remove the locview attribute if no longer
7421 (index_location_lists): Call skip_loc_list_entry for test.
7422 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7423 Use output_dwarf_version.
7424 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7425 (struct dw_val_node): Add val_view_list.
7426 * final.c (SEEN_NEXT_VIEW): New.
7427 (set_next_view_needed): New.
7428 (clear_next_view_needed): New.
7429 (maybe_output_next_view): New.
7430 (final_start_function): Rename to...
7431 (final_start_function_1): ... this. Take pointer to FIRST,
7432 add SEEN parameter. Emit param bindings in the initial view.
7433 (final_start_function): Reintroduce SEEN-less interface.
7434 (final): Rename to...
7435 (final_1): ... this. Take SEEN parameter. Output final pending
7436 next view at the end.
7437 (final): Reintroduce seen-less interface.
7438 (final_scan_insn): Output pending next view before switching
7439 sections or ending a block. Mark the next view as needed when
7440 outputting variable locations. Notify debug backend of section
7441 changes, and of location view changes.
7442 (rest_of_handle_final): Adjust.
7443 * toplev.c (process_options): Autodetect value for debug variable
7444 location views option. Warn on incompat5 without -gdwarf-5.
7445 * doc/invoke.texi (gvariable-location-views): New.
7446 (gvariable-location-views=incompat5): New.
7447 (gno-variable-location-views): New.
7449 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7451 PR tree-optimization/84136
7452 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7453 that the result of find_edge is non-NULL.
7455 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7458 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7459 storing integer register in SImode. Fix cost of 256 and 512
7460 byte aligned SSE register store.
7462 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7464 * config/i386/i386.c (ix86_multiplication_cost): Fix
7465 multiplication cost for TARGET_AVX512DQ.
7467 2018-02-08 Marek Polacek <polacek@redhat.com>
7469 PR tree-optimization/84238
7470 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7473 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7475 PR tree-optimization/84265
7476 * tree-vect-stmts.c (vectorizable_store): Don't treat
7477 VMAT_CONTIGUOUS accesses as grouped.
7478 (vectorizable_load): Likewise.
7480 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7482 PR tree-optimization/81635
7483 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7484 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7485 (test_round_for_mask): New functions.
7486 (wide_int_cc_tests): Call test_round_for_mask.
7487 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7488 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7489 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7490 range returned by get_range_info.
7492 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7495 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7496 * symtab.c: Include builtins.h
7497 (symtab_node::output_to_lto_symbol_table_p): Move here
7498 from lto-streamer-out.c:output_symbol_p.
7499 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7500 (output_symbol_p): Move all logic to symtab.c
7501 (produce_symtab): Update.
7503 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7505 * config/s390/s390-opts.h (enum indirect_branch): Define.
7506 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7507 (s390_indirect_branch_via_thunk)
7508 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7509 (enum s390_indirect_branch_type): Define.
7510 * config/s390/s390.c (struct s390_frame_layout, struct
7511 machine_function): Remove.
7512 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7513 (indirect_branch_table_label_no, indirect_branch_table_name):
7515 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7516 (enum s390_indirect_branch_option): Define.
7517 (s390_return_addr_from_memory): New function.
7518 (s390_handle_string_attribute): New function.
7519 (s390_attribute_table): Add new attribute handler.
7520 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7521 (s390_indirect_branch_via_thunk): New function.
7522 (s390_indirect_branch_via_inline_thunk): New function.
7523 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7524 sibling call optimization for non z10 compiles.
7525 (s390_emit_call): Force indirect branch target to be a single
7526 register. Add r1 clobber for non-z10 compiles.
7527 (s390_emit_epilogue): Emit return jump via return_use expander.
7528 (s390_reorg): Handle JUMP_INSNs as execute targets.
7529 (s390_option_override_internal): Perform validity checks for the
7530 new command line options.
7531 (s390_indirect_branch_attrvalue): New function.
7532 (s390_indirect_branch_settings): New function.
7533 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7534 (s390_output_indirect_thunk_function): New function.
7535 (s390_code_end): Implement target hook.
7536 (s390_case_values_threshold): Implement target hook.
7537 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7539 * config/s390/s390.h (struct s390_frame_layout)
7540 (struct machine_function): Move here from s390.c.
7541 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7542 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7543 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7544 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7545 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7546 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7547 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7548 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7549 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7550 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7551 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7552 (mnemonic attribute): Add values which aren't recognized
7554 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7555 pattern for branch conversion. Fix mnemonic attribute.
7556 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7557 indirect branch via thunk if requested.
7558 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7559 ("*indirect_jump"): Disable for branch conversion using out of
7561 ("indirect_jump_via_thunk<mode>_z10")
7562 ("indirect_jump_via_thunk<mode>")
7563 ("indirect_jump_via_inlinethunk<mode>_z10")
7564 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7565 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7566 ("casesi_jump_via_inlinethunk<mode>_z10")
7567 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7568 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7569 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7570 ("*indirect2_jump"): Disable for branch conversion.
7571 ("casesi_jump"): Turn into expander and expand patterns for branch
7573 ("return_use"): New expander.
7574 ("*return"): Emit return via thunk and rename it to ...
7575 ("*return<mode>"): ... this one.
7576 * config/s390/s390.opt: Add new options and and enum for the
7579 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7581 * lra-constraints.c (match_reload): Unconditionally use
7582 gen_lowpart_SUBREG, rather than selecting between that
7583 and equivalent gen_rtx_SUBREG code.
7585 2018-02-08 Richard Biener <rguenther@suse.de>
7587 PR tree-optimization/84233
7588 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7589 changed flag instead of boguously re-using phi_inserted.
7591 2018-02-08 Martin Jambor <mjambor@suse.cz>
7593 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7594 static local variables.
7596 2018-02-08 Richard Biener <rguenther@suse.de>
7598 PR tree-optimization/84278
7599 * tree-vect-stmts.c (vectorizable_store): When looking for
7600 smaller vector types to perform grouped strided loads/stores
7601 make sure the mode is supported by the target.
7602 (vectorizable_load): Likewise.
7604 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7606 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7607 Increase LDP/STP opportunities by adding adjacent callee-saves.
7609 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7611 PR rtl-optimization/84068
7612 PR rtl-optimization/83459
7613 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7615 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7617 PR tree-optimization/84224
7618 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7619 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7622 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7625 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7626 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7627 count, remove one USE.
7629 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7631 * doc/install.texi (Configuration): Document the
7632 --with-long-double-format={ibm,ieee} PowerPC configuration
7636 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7637 Convert from define_expand to be define_insn_and_split. Rework
7638 float/double/_Float128 conversions to QI/HI/SImode to work with
7639 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7640 conversions to QI/HImode types did a store and then a load to
7641 truncate the value. For conversions to VSX registers, don't split
7642 the insn, instead emit the code directly. Use the code iterator
7643 any_fix to combine signed and unsigned conversions.
7644 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7645 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7646 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7647 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7648 (fix_<mode>di2_hw): Likewise.
7649 (fixuns_<mode>di2_hw): Likewise.
7650 (fix_<mode>si2_hw): Likewise.
7651 (fixuns_<mode>si2_hw): Likewise.
7652 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7653 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7654 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7655 fix<uns>_trunc<SFDF:mode>si2_p8.
7656 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7658 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7659 (fix<uns>_<mode>_mem): Likewise.
7660 (fctiw<u>z_<mode>_mem): Likewise.
7661 (fix<uns>_<mode>_mem): Likewise.
7662 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7663 the register allocator from doing a direct move to the GPRs to do
7664 a store, and instead use the ISA 3.0 store byte/half-word from
7665 vector register instruction. For IEEE 128-bit floating point,
7666 also optimize stores of 32-bit ints.
7667 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7669 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7671 * genextract.c (push_pathstr_operand): New function to support
7673 (walk_rtx): Call push_pathstr_operand.
7674 (print_path): Support [a-zA-Z].
7676 2018-02-07 Richard Biener <rguenther@suse.de>
7678 PR tree-optimization/84037
7679 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7680 (cse_and_gimplify_to_preheader): Declare.
7681 (vect_get_place_in_interleaving_chain): Likewise.
7682 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7684 (_loop_vec_info::~_loop_vec_info): Delete it.
7685 (cse_and_gimplify_to_preheader): New function.
7686 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7687 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7688 (vectorizable_load): Likewise. For grouped stores always base
7689 the IV on the first element.
7690 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7691 condition before gimplifying.
7693 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7695 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7696 *DIV_EXPR and *MOD_EXPR.
7698 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7701 * config/i386/i386.c (ix86_option_override_internal): Mask out
7702 the CF_SET bit when checking -fcf-protection.
7704 2018-02-07 Tom de Vries <tom@codesourcery.com>
7707 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7710 2018-02-07 Richard Biener <rguenther@suse.de>
7712 PR tree-optimization/84204
7713 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7716 PR tree-optimization/84205
7717 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7718 special-case isl_ast_op_zdiv_r.
7720 PR tree-optimization/84223
7721 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7722 Only add conditions from within the region.
7723 (gather_bbs::after_dom_children): Adjust.
7725 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7728 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7729 * config/avr/avr.md: Only post-reload split REG-REG moves if
7730 either register is GENERAL_REG_P.
7732 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7734 PR tree-optimization/84235
7735 * tree-ssa-scopedtables.c
7736 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7737 if the subtraction is performed in floating point type where NaNs are
7738 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7739 build 1. Formatting fix.
7741 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7744 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7745 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7746 and skip it regardless of bb boundaries. Use CALL_P macro,
7747 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7748 unnecessarily, formatting fix.
7750 2018-02-06 Michael Collison <michael.collison@arm.com>
7752 * config/arm/thumb2.md:
7753 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7754 (*thumb_mov_notscc): Ditto.
7756 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7759 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7760 unsigned_fix, not "s".
7762 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7764 * configure.ac (gcc_fn_eh_frame_ro): New function.
7765 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7766 correct .eh_frame permissions.
7767 * configure: Regenerate.
7769 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7771 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7774 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7776 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7777 Display warning message for -mno-speculate-indirect-jumps.
7779 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7781 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7783 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7785 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7787 PR tree-optimization/84225
7788 * tree-eh.c (find_trapping_overflow): Only call
7789 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7791 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7794 * config/i386/i386.c: Reimplement the check of possible options
7795 -mibt/-mshstk conbination. Change error messages.
7796 * doc/invoke.texi: Fix a typo: remove extra '='.
7798 2018-02-06 Marek Polacek <polacek@redhat.com>
7800 PR tree-optimization/84228
7801 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7803 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7806 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7807 emitted arch directives.
7808 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7809 __ARM_FEATURE_COPROC before changing architectures.
7811 2018-02-06 Richard Biener <rguenther@suse.de>
7813 * config/i386/i386.c (print_reg): Fix typo.
7814 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7816 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7818 * configure: Regenerate.
7820 2018-02-05 Martin Sebor <msebor@redhat.com>
7822 PR tree-optimization/83369
7823 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7826 2018-02-05 Martin Liska <mliska@suse.cz>
7828 * doc/invoke.texi: Cherry-pick upstream r323995.
7830 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7832 * ira.c (ira_init_register_move_cost): Adjust comment.
7834 2018-02-05 Martin Liska <mliska@suse.cz>
7836 PR gcov-profile/84137
7837 * doc/gcov.texi: Fix typo in documentation.
7839 2018-02-05 Martin Liska <mliska@suse.cz>
7841 PR gcov-profile/83879
7842 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7843 using dlopen functionality.
7845 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7847 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7848 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7849 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7850 _mm_maskz_range_round_ss): New intrinsics.
7851 (__builtin_ia32_rangesd128_round)
7852 (__builtin_ia32_rangess128_round): Remove.
7853 (__builtin_ia32_rangesd128_mask_round,
7854 __builtin_ia32_rangess128_mask_round): New builtins.
7855 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7856 __builtin_ia32_rangess128_round): Remove.
7857 (__builtin_ia32_rangesd128_mask_round,
7858 __builtin_ia32_rangess128_mask_round): New builtins.
7859 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7860 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7861 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7862 "<round_saeonly_constraint>")): Changed to ...
7863 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7864 "<round_saeonly_scalar_constraint>")): ... this.
7865 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7866 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7867 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7868 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7869 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7871 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7873 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7875 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7876 Remove all values except native, 8540 and 8548.
7878 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7880 * config/i386/i386.c (ix86_output_function_return): Pass
7881 INVALID_REGNUM, instead of -1, as invalid register number to
7882 indirect_thunk_name and output_indirect_thunk.
7884 2018-02-02 Julia Koval <julia.koval@intel.com>
7886 * config.gcc: Add -march=icelake.
7887 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7888 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7889 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7890 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7891 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7892 (processor_target_table): Add icelake.
7893 (ix86_option_override_internal): Handle new PTAs.
7894 (get_builtin_code_for_version): Handle icelake.
7895 (M_INTEL_COREI7_ICELAKE): New.
7896 (fold_builtin_cpu): Handle icelake.
7897 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7898 * doc/invoke.texi: Add -march=icelake.
7900 2018-02-02 Julia Koval <julia.koval@intel.com>
7902 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7903 to wide_int_bitmask.
7904 * wide-int-bitmask.h: New.
7906 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7909 * config/i386/i386.md: Replace Pmode with word_mode in
7910 builtin_setjmp_setup and builtin_longjmp to support x32.
7912 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7916 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7918 (rs6000_supported_cpu_names): New static variable.
7919 (linux_cpu_translation_table): Likewise.
7920 (elf_platform) <cpu>: Define new static variable and use it.
7921 Translate kernel AT_PLATFORM name to canonical name if needed.
7922 Error if platform name is unknown.
7924 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7927 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7929 2018-02-01 Jeff Law <law@redhat.com>
7932 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7933 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7934 the scratch if RELEASE_VIA_POP is false.
7935 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7936 If we have to save a temporary register, decrement SIZE appropriately.
7937 Pass new arguments to release_scratch_register_on_entry.
7938 (ix86_adjust_stack_and_probe): Likewise.
7939 (ix86_emit_probe_stack_range): Pass new arguments to
7940 release_scratch_register_on_entry.
7942 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7944 PR rtl-optimization/84157
7945 * combine.c (change_zero_ext): Use REG_P predicate in
7946 front of HARD_REGISTER_P predicate.
7948 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7950 * config/avr/avr.c (avr_option_override): Move disabling of
7951 -fdelete-null-pointer-checks to...
7952 * common/config/avr/avr-common.c (avr_option_optimization_table):
7955 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7957 PR tree-optimization/81635
7958 * tree-data-ref.c (split_constant_offset_1): For types that
7959 wrap on overflow, try to use range info to prove that wrapping
7962 2018-02-01 Renlin Li <renlin.li@arm.com>
7965 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7967 (aarch64_register_move_cost): Likewise.
7968 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7970 (REG_CLASS_NAMES): Likewise.
7971 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7972 TAILCALL_ADDR_REGS. Remove IP registers.
7973 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7975 2018-02-01 Richard Biener <rguenther@suse.de>
7977 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7978 for specifying RPO order and allow NULL for that.
7979 * domwalk.c (dom_walker::dom_walker): Likewise.
7980 (dom_walker::walk): Handle NULL RPO order.
7981 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7983 (rewrite_update_dom_walker): Likewise.
7984 (mark_def_dom_walker): Likewise.
7986 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7988 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7989 (aarch64_maybe_expand_sve_subreg_move): Declare.
7990 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7991 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7993 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7994 that are semantically a reverse operation.
7995 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7996 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7997 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7999 (aarch64_can_change_mode_class): For big-endian, forbid changes
8000 between two SVE modes if they have different element sizes.
8002 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8004 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8005 the TImode handling for big-endian targets.
8007 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8009 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8010 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8012 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8013 Remove BSWAP handing for big-endian targets and use the form of
8014 LD1RQ appropariate for the mode.
8016 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8018 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8019 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8022 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8025 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8026 check for operands that need to go through aarch64_sve_reload_be.
8028 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8030 PR tree-optimization/81661
8031 PR tree-optimization/84117
8032 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8033 * tree-eh.c: Include gimplify.h.
8034 (find_trapping_overflow, replace_trapping_overflow,
8035 rewrite_to_non_trapping_overflow): New functions.
8036 * tree-vect-loop.c: Include tree-eh.h.
8037 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8038 * tree-data-ref.c: Include tree-eh.h.
8039 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8041 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8043 PR rtl-optimization/84123
8044 * combine.c (change_zero_ext): Check if hard register satisfies
8045 can_change_dest_mode before calling gen_lowpart_SUBREG.
8047 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8050 * ira.c (ira_init_register_move_cost): Remove assert.
8052 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8054 PR rtl-optimization/84071
8055 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8056 * doc/tm.texi: Regenerate.
8058 2018-01-31 Richard Biener <rguenther@suse.de>
8060 PR tree-optimization/84132
8061 * tree-data-ref.c (analyze_miv_subscript): Properly
8062 check whether evolution_function_is_affine_multivariate_p
8063 before calling gcd_of_steps_may_divide_p.
8065 2018-01-31 Julia Koval <julia.koval@intel.com>
8068 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8069 * config/i386/i386.md (rdpid_rex64) New.
8070 (rdpid): Make 32bit only.
8072 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8075 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8076 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8078 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8081 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8083 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8085 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8087 PR rtl-optimization/84071
8088 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8089 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8091 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8093 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8094 (arc_attribute_table): Add 'aux' attribute.
8095 (arc_in_small_data_p): Consider aux like variables.
8096 (arc_is_aux_reg_p): New function.
8097 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8098 (arc_get_aux_arg): New function.
8099 (prepare_move_operands): Handle aux-register access.
8100 (arc_handle_aux_attribute): New function.
8101 * doc/extend.texi (ARC Variable attributes): Add subsection.
8103 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8105 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8106 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8107 (arc_attribute_table): Add 'uncached' attribute.
8108 (arc_print_operand): Print '.di' flag for uncached memory
8110 (arc_in_small_data_p): Do not consider for small data the uncached
8112 (arc_is_uncached_mem_p): New function.
8113 * config/arc/predicates.md (compact_store_memory_operand): Check
8114 for uncached memory accesses.
8115 (nonvol_nonimm_operand): Likewise.
8116 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8118 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8121 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8122 falign-loops=): Add Optimization flag.
8124 2018-01-30 Jeff Law <law@redhat.com>
8127 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8128 INT_REGISTERS_SAVED. Check it prior to calling
8129 get_scratch_register_on_entry.
8130 (ix86_adjust_stack_and_probe): Similarly.
8131 (ix86_emit_probe_stack_range): Similarly.
8132 (ix86_expand_prologue): Corresponding changes.
8134 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8137 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8138 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8140 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8143 * lra-constraints.c (curr_insn_transform): Process AND in the
8146 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8148 PR rtl-optimization/83986
8149 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8150 dependence against last_pending_memory_flush in addition to
8153 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8155 PR tree-optimization/81611
8156 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8159 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8162 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8165 2018-01-30 Richard Biener <rguenther@suse.de>
8166 Jakub Jelinek <jakub@redhat.com>
8168 PR tree-optimization/84111
8169 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8170 inner loops added during recursion, as they don't have up-to-date
8173 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8176 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8177 (can_inline_edge_by_limits_p): ... here.
8178 (can_early_inline_edge_p, check_callers,
8179 update_caller_keys, update_callee_keys, recursive_inlining,
8180 add_new_edges_to_heap, speculation_useful_p,
8181 inline_small_functions,
8182 inline_small_functions, flatten_function,
8183 inline_to_all_callers_1): Update.
8185 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8187 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8190 2018-01-30 Richard Biener <rguenther@suse.de>
8192 PR tree-optimization/83008
8193 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8194 invariant and constant vector uses in stmts when they need
8197 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8200 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8201 * configure: Regenerate.
8203 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8205 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8207 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8208 Use gen_rtx_REG rather than gen_lowpart.
8210 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8212 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8213 rather than 0 when creating partial subregs.
8215 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8217 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8220 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8223 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8224 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8225 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8226 flags. This restores the settings used before the 2017-07-24.
8227 Turning off pre increment/decrement/modify allows IVOPTS to
8228 optimize DF/SF loops where the index is an int.
8230 2018-01-29 Richard Biener <rguenther@suse.de>
8231 Kelvin Nilsen <kelvin@gcc.gnu.org>
8234 * tree-vect-stmts.c (vectorizable_call): Don't call
8235 targetm.vectorize_builtin_md_vectorized_function if callee is
8238 2018-01-22 Carl Love <cel@us.ibm.com>
8240 * doc/extend.tex: Fix typo in second arg in
8241 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8243 2018-01-29 Richard Biener <rguenther@suse.de>
8245 PR tree-optimization/84086
8246 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8247 (flush_ssaname_freelist): When SSA names were released reset
8248 the SCEV hash table.
8250 2018-01-29 Richard Biener <rguenther@suse.de>
8252 PR tree-optimization/84057
8253 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8254 removed paths when removing edges.
8256 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8258 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8259 -mfunction-return=@var{choice}.
8261 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8264 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8266 (layout::print_source_line): Likewise.
8267 (test_get_line_width_without_trailing_whitespace): Add test cases.
8269 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8272 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8275 2018-01-26 Jim Wilson <jimw@sifive.com>
8277 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8279 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8282 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8284 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8285 and CMP + SUB-immediate -> SUBS.
8287 2018-01-26 Martin Sebor <msebor@redhat.com>
8289 PR tree-optimization/83896
8290 * tree-ssa-strlen.c (get_string_len): Rename...
8291 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8292 Avoid assuming length is constant.
8293 (handle_char_store): Use HOST_WIDE_INT for string length.
8295 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8298 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8299 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8301 2018-01-26 Richard Biener <rguenther@suse.de>
8303 PR rtl-optimization/84003
8304 * dse.c (record_store): Only record redundant stores when
8305 the earlier store aliases at least all accesses the later one does.
8307 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8309 PR rtl-optimization/83985
8310 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8311 REG_CFA_RESTORE insns.
8312 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8313 REG_CFA_RESTORE insns here.
8316 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8317 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8319 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8321 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8322 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8323 (arc_init): Likewise.
8324 (arc_override_options): Likewise.
8325 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8327 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8329 * config/arc/arc.h (TARGET_DBNZ): Define.
8330 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8331 properly set the tune attribute.
8332 (dbnz): Use TARGET_DBNZ guard.
8333 * config/arc/arc.opt (mtune): Add core3 option.
8335 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8337 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8338 recognize new pic like addresses.
8339 (arc_delegitimize_address): Clean up.
8341 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8343 * config/arc/arc-arches.def: Option mrf16 valid for all
8345 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8346 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8347 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8348 * config/arc/arc-tables.opt: Regenerate.
8349 * config/arc/arc.c (arc_conditional_register_usage): Handle
8350 reduced register file case.
8351 (arc_file_start): Set must have build attributes.
8352 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8354 * config/arc/arc.opt (mrf16): Add new option.
8355 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8356 * config/arc/genmultilib.awk: Handle new mrf16 option.
8357 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8358 * config/arc/t-multilib: Regenerate.
8359 * doc/invoke.texi (ARC Options): Document mrf16 option.
8361 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8363 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8364 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8365 (arc_attribute_table): Add 'secure_call' attribute.
8366 (arc_print_operand): Print secure call operand.
8367 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8369 (arc_is_secure_call_p): New function. * config/arc/arc.md
8370 (call_i): Add support for sjli instruction.
8371 (call_value_i): Likewise.
8372 * config/arc/constraints.md (Csc): New constraint.
8374 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8375 John Eric Martin <John.Martin@emmicro-us.com>
8377 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8378 * config/arc/arc.c (_arc_jli_section): New struct.
8379 (arc_jli_section): New type.
8380 (rc_jli_sections): New static variable.
8381 (arc_handle_jli_attribute): New function.
8382 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8383 (arc_file_end): New function.
8384 (TARGET_ASM_FILE_END): Define.
8385 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8386 (arc_add_jli_section): New function.
8387 (jli_call_scan): Likewise.
8388 (arc_reorg): Call jli_call_scan.
8389 (arc_output_addsi): Remove 'S' from printing asm operand.
8390 (arc_is_jli_call_p): New function.
8391 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8393 (movhi_insn): Likewise.
8394 (movsi_insn): Likewise.
8395 (movsi_set_cc_insn): Likewise.
8396 (loadqi_update): Likewise.
8397 (load_zeroextendqisi_update): Likewise.
8398 (load_signextendqisi_update): Likewise.
8399 (loadhi_update): Likewise.
8400 (load_zeroextendhisi_update): Likewise.
8401 (load_signextendhisi_update): Likewise.
8402 (loadsi_update): Likewise.
8403 (loadsf_update): Likewise.
8404 (movsicc_insn): Likewise.
8405 (bset_insn): Likewise.
8406 (bxor_insn): Likewise.
8407 (bclr_insn): Likewise.
8408 (bmsk_insn): Likewise.
8409 (bicsi3_insn): Likewise.
8410 (cmpsi_cc_c_insn): Likewise.
8411 (movsi_ne): Likewise.
8412 (movsi_cond_exec): Likewise.
8413 (clrsbsi2): Likewise.
8424 (call_i): Remove 'S' asm letter, add jli instruction.
8425 (call_value_i): Likewise.
8426 * config/arc/arc.op (mjli-always): New option.
8427 * config/arc/constraints.md (Cji): New constraint.
8428 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8430 (subsf3_fpx): Likewise.
8431 (mulsf3_fpx): Likewise.
8432 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8434 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8436 * doc/invoke.texi (ARC): Document mjli-always option.
8438 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8440 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8441 avoid addition with 0 and use incw and decw where possible.
8443 2018-01-26 Richard Biener <rguenther@suse.de>
8445 PR tree-optimization/81082
8446 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8447 association if it requires casting to unsigned.
8448 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8449 from fold_plusminus_mult_expr to catch important cases late when
8450 range info is available.
8452 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8454 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8455 * configure.ac (hidden_linkonce): New test.
8456 * configure: Regenerate.
8457 * config.in: Regenerate.
8459 2018-01-26 Julia Koval <julia.koval@intel.com>
8461 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8462 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8463 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8464 _mm_mask_bitshuffle_epi64_mask): Fix type.
8465 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8466 USI_FTYPE_V4DI_V4DI_USI): Remove.
8467 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8468 __builtin_ia32_vpshufbitqmb256_mask,
8469 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8470 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8471 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8473 2018-01-26 Alan Modra <amodra@gmail.com>
8476 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8477 UNSPEC_VBPERMQ. Sort other unspecs.
8479 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8481 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8483 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8486 * predict.c (drop_profile): Do not push/pop cfun; update also
8488 (handle_missing_profiles): Fix logic looking for zero profiles.
8490 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8493 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8494 on functions with #pragma omp declare simd or functions with simd
8496 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8497 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8498 Remove trailing \n from warning_at calls.
8500 2018-01-25 Tom de Vries <tom@codesourcery.com>
8503 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8504 for neutered workers.
8506 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8509 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8510 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8512 2018-01-24 Jeff Law <law@redhat.com>
8515 * i386.c (get_probe_interval): Move to earlier point.
8516 (ix86_compute_frame_layout): If -fstack-clash-protection and
8517 the frame is larger than the probe interval, then use pushes
8518 to save registers rather than reg->mem moves.
8519 (ix86_expand_prologue): Remove conditional for int_registers_saved
8522 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8525 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8526 min/max for never referenced object.
8528 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8531 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8533 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8534 attributes from DECL_ATTRIBUTES (decl) without affecting
8535 DECL_ATTRIBUTES (current_function_decl).
8536 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8537 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8539 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8541 PR tree-optimization/83979
8542 * fold-const.c (fold_comparison): Use constant_boolean_node
8543 instead of boolean_{true,false}_node.
8545 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8547 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8550 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8552 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8553 Simplify the clause that sets the length attribute.
8554 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8555 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8556 clause that sets the length attribute.
8557 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8559 2018-01-24 Tom de Vries <tom@codesourcery.com>
8562 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8563 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8564 Add strict parameter.
8565 (prevent_branch_around_nothing): Insert dummy insn between branch to
8566 label and label with no ptx insn inbetween.
8567 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8569 2018-01-24 Tom de Vries <tom@codesourcery.com>
8572 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8573 for neutered threads in warp.
8574 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8576 2018-01-24 Richard Biener <rguenther@suse.de>
8578 PR tree-optimization/83176
8579 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8582 2018-01-24 Richard Biener <rguenther@suse.de>
8584 PR tree-optimization/82819
8585 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8586 code generating pluses that are no-ops in the target precision.
8588 2018-01-24 Richard Biener <rguenther@suse.de>
8591 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8593 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8595 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8596 to merge probabilities.
8597 * predict.c (probably_never_executed): Also mark as cold functions
8598 with global 0 profile and guessed local profile.
8599 * profile-count.c (profile_probability::combine_with_count): New
8601 * profile-count.h (profile_probability::operator*,
8602 profile_probability::operator*=, profile_probability::operator/,
8603 profile_probability::operator/=): Reduce precision to adjusted
8604 and set value to guessed on contradictory divisions.
8605 (profile_probability::combine_with_freq): Remove.
8606 (profile_probability::combine_wiht_count): Declare.
8607 (profile_count::force_nonzero):: Set to adjusted.
8608 (profile_count::probability_in):: Set quality to adjusted.
8609 * tree-ssa-tail-merge.c (replace_block_by): Use
8612 2018-01-23 Andrew Waterman <andrew@sifive.com>
8613 Jim Wilson <jimw@sifive.com>
8615 * config/riscv/riscv.c (riscv_stack_boundary): New.
8616 (riscv_option_override): Set riscv_stack_boundary. Handle
8617 riscv_preferred_stack_boundary_arg.
8618 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8619 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8620 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8621 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8622 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8623 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8625 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8628 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8629 of struct ix86_frame.
8630 (ix86_expand_epilogue): Likewise. Add a local variable for
8631 the reg_save_offset field in struct ix86_frame.
8633 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8635 PR tree-optimization/82604
8636 * tree-loop-distribution.c (enum partition_kind): New enum item
8637 PKIND_PARTIAL_MEMSET.
8638 (partition_builtin_p): Support above new enum item.
8639 (generate_code_for_partition): Ditto.
8640 (compute_access_range): Differentiate cases that equality can be
8641 proven at all loops, the innermost loops or no loops.
8642 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8643 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8644 (finalize_partitions, distribute_loop): Don't fuse partition of
8645 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8646 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8649 2018-01-23 Martin Liska <mliska@suse.cz>
8651 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8652 order to ignore the predictor.
8653 (PRED_POLYMORPHIC_CALL): Likewise.
8654 (PRED_RECURSIVE_CALL): Likewise.
8656 2018-01-23 Martin Liska <mliska@suse.cz>
8658 * tree-profile.c (tree_profiling): Print function header to
8659 aware reader which function we are working on.
8660 * value-prof.c (gimple_find_values_to_profile): Do not print
8661 not interesting value histograms.
8663 2018-01-23 Martin Liska <mliska@suse.cz>
8665 * profile-count.h (enum profile_quality): Add
8666 profile_uninitialized as the first value. Do not number values
8667 as they are zero based.
8668 (profile_count::verify): Update sanity check.
8669 (profile_probability::verify): Likewise.
8671 2018-01-23 Nathan Sidwell <nathan@acm.org>
8673 * doc/invoke.texi (ffor-scope): Deprecate.
8675 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8677 PR tree-optimization/83510
8678 * domwalk.c (set_all_edges_as_executable): New function.
8679 (dom_walker::dom_walker): Convert bool param
8680 "skip_unreachable_blocks" to enum reachability. Move setup of
8681 edge flags to set_all_edges_as_executable and only do it when
8682 reachability is REACHABLE_BLOCKS.
8683 * domwalk.h (enum dom_walker::reachability): New enum.
8684 (dom_walker::dom_walker): Convert bool param
8685 "skip_unreachable_blocks" to enum reachability.
8686 (set_all_edges_as_executable): New decl.
8687 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8688 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8690 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8691 but converting true to REACHABLE_BLOCKS.
8692 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8694 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8695 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8696 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8698 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8699 if check_all_array_refs will be called.
8701 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8703 * tree.c (selftest::test_location_wrappers): Add more test
8706 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8708 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8709 (selftest::test_bit_in_range): Likewise.
8711 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8714 * doc/sourcebuild.texi (vect_float): Say that the selector
8715 only describes the situation when -funsafe-math-optimizations is on.
8716 (vect_float_strict): Document.
8718 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8720 PR tree-optimization/83965
8721 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8722 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8723 instead of checking only for a reduction.
8724 (vect_recog_widen_sum_pattern): Likewise.
8726 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8728 * predict.c (probably_never_executed): Only use precise profile info.
8729 (compute_function_frequency): Skip after inlining hack since we now
8730 have quality checking.
8732 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8734 * profile-count.h (profile_probability::very_unlikely,
8735 profile_probability::unlikely, profile_probability::even): Set
8736 precision to guessed.
8738 2018-01-23 Richard Biener <rguenther@suse.de>
8740 PR tree-optimization/83963
8741 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8742 Properly terminate dominator walk when crossing the exit edge not
8743 when visiting its source block.
8745 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8748 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8749 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8751 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8753 PR tree-optimization/83957
8754 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8755 semicolon after for body surrounded by braces.
8757 PR tree-optimization/83081
8758 * profile-count.h (profile_probability::split): New method.
8759 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8760 Use profile_probability::split.
8761 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8762 when splitting a single conditional jump into 2.
8764 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8766 PR tree-optimization/69452
8767 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8770 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8772 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8773 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8774 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8776 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8778 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8779 * config/rl78/rl78.md (movdi): New define_expand.
8780 * config/rl78/rl78.c (rl78_split_movdi): New function.
8782 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8785 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8787 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8788 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8789 128-bit to produce an UNSPEC move to get the double word with the
8790 signbit and then a shift directly to do signbit.
8791 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8792 implementation with a new version that just does either a direct
8793 move or a regular move. Move memory interface to separate insns.
8794 Move insns so they are next to the expander.
8795 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8796 with signbit move. Split big and little endian case.
8797 (signbit<mode>2_dm_mem_le): Likewise.
8798 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8799 (signbit<mode>2_dm2): Likewise.
8801 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8803 * config/rl78/rl78.md (anddi3): New define_expand.
8805 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8807 * config/rl78/rl78.md (umindi3): New define_expand.
8809 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8811 * config/rl78/rl78.md (smindi3): New define_expand.
8813 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8815 * config/rl78/rl78.md (smaxdi3): New define_expand.
8817 2018-01-22 Carl Love <cel@us.ibm.com>
8819 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8820 LVX_V1TI): Add macro expansion.
8821 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8822 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8823 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8824 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8825 Change check to determine if the instruction is a byte reversing
8826 entry. Fix typo in comment.
8827 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8828 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8829 Add def_builtin calls for new builtins.
8830 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8831 Add define_insn expansion.
8833 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8835 * config/rl78/rl78.md (umaxdi3): New define_expand.
8837 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8839 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8840 for non-QImode registers.
8842 2018-01-22 Richard Biener <rguenther@suse.de>
8844 PR tree-optimization/83963
8845 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8846 including the loop exit block.
8847 (scop_detection::merge_sese): Likewise.
8848 (scop_detection::add_scop): Do it here instead.
8850 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8852 * doc/sourcebuild.texi (arm_softfloat): Document.
8854 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8857 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8858 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8859 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8861 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8862 David Edelsohn <dje.gcc@gmail.com>
8865 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8866 Change "crset eq" to "crset 2".
8867 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8868 (*call_indirect_aix<mode>_nospec): Likewise.
8869 (*call_value_indirect_aix<mode>_nospec): Likewise.
8870 (*call_indirect_elfv2<mode>_nospec): Likewise.
8871 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8872 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8873 change assembly output from . to $.
8874 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8875 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8876 (*tablejump<mode>_internal1_nospec): Likewise.
8878 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8881 * config/sh/sh_optimize_sett_clrt.cc:
8882 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8884 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8886 PR tree-optimization/83940
8887 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8888 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8889 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8890 use it to pass back the definition type.
8891 (vect_check_store_rhs): Likewise rhs_dt_out.
8892 (vect_build_gather_load_calls): Add a mask_dt argument and use
8893 it instead of a call to vect_is_simple_use.
8894 (vectorizable_store): Update calls to vect_check_load_store_mask
8895 and vect_check_store_rhs. Use the dt returned by the latter instead
8896 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8897 instead of calls to vect_is_simple_use. Pass the scalar rather
8898 than the vector operand to vect_is_simple_use when handling
8899 second and subsequent copies of an rhs value.
8900 (vectorizable_load): Update calls to vect_check_load_store_mask
8901 and vect_build_gather_load_calls. Use the cached mask_dt and
8902 gs_info.offset_dt instead of calls to vect_is_simple_use.
8904 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8907 * tree-emutls.c: Include gimplify.h.
8908 (lower_emutls_2): New function.
8909 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8910 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8911 it before further processing.
8914 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8915 UINTVAL (trueop1) instead of INTVAL (op1).
8917 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8921 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8922 INCOMING_FRAME_SP_OFFSET if not defined.
8923 (scan_trace): Add ENTRY argument. If true and
8924 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8925 emit a note to adjust the CFA offset.
8926 (create_cfi_notes): Adjust scan_trace callers.
8927 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8928 INCOMING_FRAME_SP_OFFSET in the CIE.
8929 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8930 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8932 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8933 * doc/tm.texi: Regenerated.
8935 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8937 PR rtl-optimization/83147
8938 * lra-constraints.c (remove_inheritance_pseudos): Use
8939 lra_substitute_pseudo_within_insn.
8941 2018-01-19 Tom de Vries <tom@codesourcery.com>
8942 Cesar Philippidis <cesar@codesourcery.com>
8945 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8947 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8950 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8951 spaces for function labels.
8953 2018-01-19 Martin Liska <mliska@suse.cz>
8955 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8956 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8957 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8958 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8959 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8960 (PRED_CONST_RETURN): Change from 69 to 65.
8961 (PRED_NULL_RETURN): Change from 91 to 71.
8962 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8963 (PRED_LOOP_GUARD): Change from 66 to 73.
8965 2018-01-19 Martin Liska <mliska@suse.cz>
8967 * predict.c (predict_insn_def): Add new assert.
8968 (struct branch_predictor): Change type to signed integer.
8969 (test_prediction_value_range): Amend test to cover
8971 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8972 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8973 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8974 (PRED_LOOP_IV_COMPARE): Likewise.
8975 * predict.h (PROB_UNINITIALIZED): Define new constant.
8977 2018-01-19 Martin Liska <mliska@suse.cz>
8979 * predict.c (dump_prediction): Add new format for
8980 analyze_brprob.py script which is enabled with -details
8982 * profile-count.h (precise_p): New function.
8984 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8986 PR tree-optimization/83922
8987 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8988 there are no statements that need masking.
8989 (vect_active_double_reduction_p): New function.
8990 (vect_analyze_loop_operations): Use it when handling phis that
8991 are not in the loop header.
8993 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8995 PR tree-optimization/83914
8996 * tree-vect-loop.c (vectorizable_induction): Don't convert
8997 init_expr or apply the peeling adjustment for inductions
8998 that are nested within the vectorized loop.
9000 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9002 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9005 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9009 * function.h (gimplify_parameters): Add gimple_seq * argument.
9010 * function.c: Include gimple.h and options.h.
9011 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9012 for the added local temporaries if needed.
9013 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9014 if there are any parameter cleanups, wrap whole body into a
9015 try/finally with the cleanups.
9017 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9020 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9021 Use GET_MODE_CLASS for scalar floating point.
9023 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9027 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9028 Fix call of call_cgraph_insertion_hooks.
9030 2018-01-18 Martin Sebor <msebor@redhat.com>
9032 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9034 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9037 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9040 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9043 * common.opt: (-ffile-prefix-map): New option.
9044 * opts.c (common_handle_option): Defer it.
9045 * opts-global.c (handle_common_deferred_options): Handle it.
9046 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9047 * file-prefix-map.h: New file.
9048 (remap_debug_filename, add_debug_prefix_map): ...here.
9049 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9050 * final.c (debug_prefix_map, add_debug_prefix_map
9051 remap_debug_filename): Move to...
9052 * file-prefix-map.c: New file.
9053 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9054 generalize, get rid of alloca(), use strrchr() instead of strchr().
9055 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9056 Implement in terms of add_prefix_map().
9057 (remap_macro_filename, remap_debug_filename): Implement in term of
9059 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9060 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9061 * dbxout.c: Include file-prefix-map.h.
9062 * varasm.c: Likewise.
9063 * vmsdbgout.c: Likewise.
9064 * xcoffout.c: Likewise.
9065 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9066 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9067 * doc/invoke.texi (-ffile-prefix-map): Document.
9068 (-fdebug-prefix-map): Update description.
9070 2018-01-18 Martin Liska <mliska@suse.cz>
9072 * config/i386/i386.c (indirect_thunk_name): Document that also
9074 (output_indirect_thunk): Document why both instructions
9075 (pause and lfence) are generated.
9077 2018-01-18 Richard Biener <rguenther@suse.de>
9079 PR tree-optimization/83887
9080 * graphite-scop-detection.c
9081 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9082 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9083 (scop_detection::merge_sese): Re-implement with a flood-fill
9084 algorithm that properly finds a SESE region if it exists.
9086 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9089 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9090 pointer_diff optimizations use view_convert instead of convert.
9092 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9094 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9095 Generate different code for -mno-speculate-indirect-jumps.
9096 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9097 (*call_indirect_aix<mode>): Disable for
9098 -mno-speculate-indirect-jumps.
9099 (*call_indirect_aix<mode>_nospec): New define_insn.
9100 (*call_value_indirect_aix<mode>): Disable for
9101 -mno-speculate-indirect-jumps.
9102 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9103 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9104 -mno-speculate-indirect-jumps.
9105 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9107 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9109 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9110 long double type, set the flags for noting the default long double
9111 type, even if we don't pass or return a long double type.
9113 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9116 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9119 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9121 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9122 support for merge[hl].
9123 (fold_mergehl_helper): New helper function.
9124 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9125 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9126 (altivec_vmrglw_direct): Add xxmrglw insn.
9128 2018-01-17 Andrew Waterman <andrew@sifive.com>
9130 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9131 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9133 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9136 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9137 call the lto_location_cache before reading the
9138 DECL_SOURCE_LOCATION of the types.
9140 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9141 Richard Sandiford <richard.sandiford@linaro.org>
9143 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9144 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9145 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9146 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9147 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9149 * config/aarch64/constraints.md (aarch64_movti_operand):
9151 * config/aarch64/predicates.md (Uti): Add new constraint.
9153 2018-01-17 Carl Love <cel@us.ibm.com>
9155 * config/rs6000/vsx.md (define_expand xl_len_r,
9156 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9157 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9159 (define_expand, define_insn): Move the shift left from the
9160 define_insn to the define_expand for lxvl and stxvl instructions.
9161 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9162 and XL_LEN_R definitions to PURE.
9164 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9166 * config/i386/i386.c (indirect_thunk_name): Declare regno
9167 as unsigned int. Compare regno with INVALID_REGNUM.
9168 (output_indirect_thunk): Ditto.
9169 (output_indirect_thunk_function): Ditto.
9170 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9171 in the call to output_indirect_thunk_function.
9173 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9176 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9177 rather than the size of inner_type to determine the stack slot size
9178 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9180 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9183 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9186 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9188 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9189 endian Linux systems to optionally enable multilibs for selecting
9190 the long double type if the user configured an explicit type.
9191 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9192 have no long double multilibs if not defined.
9193 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9194 warn if the user used -mabi={ieee,ibm}longdouble and we built
9195 multilibs for long double.
9196 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9197 appropriate multilib option.
9198 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9200 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9201 for building long double multilibs.
9202 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9204 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9206 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9209 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9211 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9214 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9217 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9220 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9222 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9223 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9225 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9227 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9228 different rtl trees depending on TARGET_64BIT.
9229 (rs6000_gen_lvx): Likewise.
9231 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9233 * config/visium/visium.md (nop): Tweak comment.
9234 (hazard_nop): Likewise.
9236 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9238 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9239 -mspeculate-indirect-jumps.
9240 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9241 for -mno-speculate-indirect-jumps.
9242 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9243 (*call_value_indirect_elfv2<mode>): Disable for
9244 -mno-speculate-indirect-jumps.
9245 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9246 (indirect_jump): Emit different RTL for
9247 -mno-speculate-indirect-jumps.
9248 (*indirect_jump<mode>): Disable for
9249 -mno-speculate-indirect-jumps.
9250 (*indirect_jump<mode>_nospec): New define_insn.
9251 (tablejump): Emit different RTL for
9252 -mno-speculate-indirect-jumps.
9253 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9254 (tablejumpsi_nospec): New define_expand.
9255 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9256 (tablejumpdi_nospec): New define_expand.
9257 (*tablejump<mode>_internal1): Disable for
9258 -mno-speculate-indirect-jumps.
9259 (*tablejump<mode>_internal1_nospec): New define_insn.
9260 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9263 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9265 * caller-save.c (insert_save): Drop unnecessary parameter. All
9268 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9269 Richard Biener <rguenth@suse.de>
9272 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9273 return early, inline manually is_gimple_sizepos. Make sure if we
9274 call gimplify_expr we don't end up with a gimple constant.
9275 * tree.c (variably_modified_type_p): Don't return true for
9276 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9277 * gimplify.h (is_gimple_sizepos): Remove.
9279 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9281 PR tree-optimization/83857
9282 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9283 vectorizable_live_operation for pure SLP statements.
9284 (vectorizable_live_operation): Handle PHIs.
9286 2018-01-16 Richard Biener <rguenther@suse.de>
9288 PR tree-optimization/83867
9289 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9290 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9292 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9295 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9296 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9297 If off is not INTEGER_CST, issue a may not be aligned warning
9298 rather than isn't aligned. Use isn%'t rather than isn't.
9299 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9301 <case MULT_EXPR>: Improve the case when bottom and one of the
9302 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9303 operand, in that case check if the other operand is multiple of
9304 bottom divided by the INTEGER_CST operand.
9306 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9309 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9310 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9311 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9312 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9313 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9314 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9315 (pa_function_arg_size): New function.
9317 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9319 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9320 in a separate statement.
9322 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9324 PR tree-optimization/83847
9325 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9326 group gathers and scatters.
9328 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9330 PR rtl-optimization/86620
9331 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9333 PR rtl-optimization/83213
9334 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9335 to last if both are JUMP_INSNs.
9337 PR tree-optimization/83843
9338 * gimple-ssa-store-merging.c
9339 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9340 store_immediate_info for bswap/nop orig_stores.
9342 2018-01-15 Andrew Waterman <andrew@sifive.com>
9344 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9346 <UDIV>: Increase cost if !TARGET_DIV.
9348 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9350 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9351 (define_attr "cr_logical_3op"): New.
9352 (cceq_ior_compare): Adjust.
9353 (cceq_ior_compare_complement): Adjust.
9354 (*cceq_rev_compare): Adjust.
9355 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9356 (is_cracked_insn): Adjust.
9357 (insn_must_be_first_in_group): Adjust.
9358 * config/rs6000/40x.md: Adjust.
9359 * config/rs6000/440.md: Adjust.
9360 * config/rs6000/476.md: Adjust.
9361 * config/rs6000/601.md: Adjust.
9362 * config/rs6000/603.md: Adjust.
9363 * config/rs6000/6xx.md: Adjust.
9364 * config/rs6000/7450.md: Adjust.
9365 * config/rs6000/7xx.md: Adjust.
9366 * config/rs6000/8540.md: Adjust.
9367 * config/rs6000/cell.md: Adjust.
9368 * config/rs6000/e300c2c3.md: Adjust.
9369 * config/rs6000/e500mc.md: Adjust.
9370 * config/rs6000/e500mc64.md: Adjust.
9371 * config/rs6000/e5500.md: Adjust.
9372 * config/rs6000/e6500.md: Adjust.
9373 * config/rs6000/mpc.md: Adjust.
9374 * config/rs6000/power4.md: Adjust.
9375 * config/rs6000/power5.md: Adjust.
9376 * config/rs6000/power6.md: Adjust.
9377 * config/rs6000/power7.md: Adjust.
9378 * config/rs6000/power8.md: Adjust.
9379 * config/rs6000/power9.md: Adjust.
9380 * config/rs6000/rs64.md: Adjust.
9381 * config/rs6000/titan.md: Adjust.
9383 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9385 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9386 ix86_indirect_branch_register logic.
9388 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9390 * config/i386/constraints.md (Bs): Update
9391 ix86_indirect_branch_register check. Don't check
9392 ix86_indirect_branch_register with GOT_memory_operand.
9394 * config/i386/predicates.md (GOT_memory_operand): Don't check
9395 ix86_indirect_branch_register here.
9396 (GOT32_symbol_operand): Likewise.
9398 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9400 * config/i386/predicates.md (constant_call_address_operand):
9401 Rewrite ix86_indirect_branch_register logic.
9402 (sibcall_insn_operand): Likewise.
9404 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9406 * config/i386/constraints.md (Bs): Replace
9407 ix86_indirect_branch_thunk_register with
9408 ix86_indirect_branch_register.
9410 * config/i386/i386.md (indirect_jump): Likewise.
9411 (tablejump): Likewise.
9412 (*sibcall_memory): Likewise.
9413 (*sibcall_value_memory): Likewise.
9414 Peepholes of indirect call and jump via memory: Likewise.
9415 * config/i386/i386.opt: Likewise.
9416 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9417 (GOT_memory_operand): Likewise.
9418 (call_insn_operand): Likewise.
9419 (sibcall_insn_operand): Likewise.
9420 (GOT32_symbol_operand): Likewise.
9422 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9425 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9426 type rather than type addr's type points to.
9427 (expand_omp_atomic_mutex): Likewise.
9428 (expand_omp_atomic): Likewise.
9430 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9433 * config/i386/i386.c (output_indirect_thunk_function): Use
9434 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9435 for __x86_return_thunk.
9437 2018-01-15 Richard Biener <rguenther@suse.de>
9440 * expmed.c (extract_bit_field_1): Fix typo.
9442 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9445 * config/arm/iterators.md (VF): New mode iterator.
9446 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9447 Remove integer-related logic from pattern.
9448 (neon_vabd<mode>_3): Likewise.
9450 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9453 * common.opt (fstrict-overflow): No longer an alias.
9454 (fwrapv-pointer): New option.
9455 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9456 also for pointer types based on flag_wrapv_pointer.
9457 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9458 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9459 opts->x_flag_wrapv got set.
9460 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9461 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9462 POINTER_TYPE_OVERFLOW_UNDEFINED.
9463 * match.pd: Likewise in address comparison pattern.
9464 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9466 2018-01-15 Richard Biener <rguenther@suse.de>
9469 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9470 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9471 Reset type names to their identifier if their TYPE_DECL doesn't
9472 have linkage (and thus is used for ODR and devirt).
9473 (save_debug_info_for_decl): Remove.
9474 (save_debug_info_for_type): Likewise.
9475 (add_tree_to_fld_list): Adjust.
9476 * tree-pretty-print.c (dump_generic_node): Make dumping of
9477 type names more robust.
9479 2018-01-15 Richard Biener <rguenther@suse.de>
9481 * BASE-VER: Bump to 8.0.1.
9483 2018-01-14 Martin Sebor <msebor@redhat.com>
9486 * builtins.c (check_access): Avoid warning when the no-warning bit
9489 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9491 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9492 * ira-color (allocno_hard_regs_compare): Likewise.
9494 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9497 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9498 Use .pushsection/.popsection.
9500 2018-01-14 Martin Sebor <msebor@redhat.com>
9503 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9505 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9507 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9508 entry from extra_headers.
9509 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9510 extra_headers, make the list bitwise identical to the i?86-*-* one.
9512 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9514 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9515 -mcmodel=large with -mindirect-branch=thunk,
9516 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9517 -mfunction-return=thunk-extern.
9518 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9519 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9520 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9522 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9524 * config/i386/i386.c (print_reg): Print the name of the full
9525 integer register without '%'.
9526 (ix86_print_operand): Handle 'V'.
9527 * doc/extend.texi: Document 'V' modifier.
9529 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9531 * config/i386/constraints.md (Bs): Disallow memory operand for
9532 -mindirect-branch-register.
9534 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9535 (GOT_memory_operand): Likewise.
9536 (call_insn_operand): Likewise.
9537 (sibcall_insn_operand): Likewise.
9538 (GOT32_symbol_operand): Likewise.
9539 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9540 for -mindirect-branch-register.
9541 (tablejump): Likewise.
9542 (*sibcall_memory): Likewise.
9543 (*sibcall_value_memory): Likewise.
9544 Disallow peepholes of indirect call and jump via memory for
9545 -mindirect-branch-register.
9546 (*call_pop): Replace m with Bw.
9547 (*call_value_pop): Likewise.
9548 (*sibcall_pop_memory): Replace m with Bs.
9549 * config/i386/i386.opt (mindirect-branch-register): New option.
9550 * doc/invoke.texi: Document -mindirect-branch-register option.
9552 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9554 * config/i386/i386-protos.h (ix86_output_function_return): New.
9555 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9556 set function_return_type.
9557 (indirect_thunk_name): Add ret_p to indicate thunk for function
9559 (output_indirect_thunk_function): Pass false to
9560 indirect_thunk_name.
9561 (ix86_output_indirect_branch_via_reg): Likewise.
9562 (ix86_output_indirect_branch_via_push): Likewise.
9563 (output_indirect_thunk_function): Create alias for function
9564 return thunk if regno < 0.
9565 (ix86_output_function_return): New function.
9566 (ix86_handle_fndecl_attribute): Handle function_return.
9567 (ix86_attribute_table): Add function_return.
9568 * config/i386/i386.h (machine_function): Add
9569 function_return_type.
9570 * config/i386/i386.md (simple_return_internal): Use
9571 ix86_output_function_return.
9572 (simple_return_internal_long): Likewise.
9573 * config/i386/i386.opt (mfunction-return=): New option.
9574 (indirect_branch): Mention -mfunction-return=.
9575 * doc/extend.texi: Document function_return function attribute.
9576 * doc/invoke.texi: Document -mfunction-return= option.
9578 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9580 * config/i386/i386-opts.h (indirect_branch): New.
9581 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9582 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9583 with local indirect jump when converting indirect call and jump.
9584 (ix86_set_indirect_branch_type): New.
9585 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9586 (indirectlabelno): New.
9587 (indirect_thunk_needed): Likewise.
9588 (indirect_thunk_bnd_needed): Likewise.
9589 (indirect_thunks_used): Likewise.
9590 (indirect_thunks_bnd_used): Likewise.
9591 (INDIRECT_LABEL): Likewise.
9592 (indirect_thunk_name): Likewise.
9593 (output_indirect_thunk): Likewise.
9594 (output_indirect_thunk_function): Likewise.
9595 (ix86_output_indirect_branch_via_reg): Likewise.
9596 (ix86_output_indirect_branch_via_push): Likewise.
9597 (ix86_output_indirect_branch): Likewise.
9598 (ix86_output_indirect_jmp): Likewise.
9599 (ix86_code_end): Call output_indirect_thunk_function if needed.
9600 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9602 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9603 (ix86_attribute_table): Add indirect_branch.
9604 * config/i386/i386.h (machine_function): Add indirect_branch_type
9605 and has_local_indirect_jump.
9606 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9608 (tablejump): Likewise.
9609 (*indirect_jump): Use ix86_output_indirect_jmp.
9610 (*tablejump_1): Likewise.
9611 (simple_return_indirect_internal): Likewise.
9612 * config/i386/i386.opt (mindirect-branch=): New option.
9613 (indirect_branch): New.
9616 (thunk-inline): Likewise.
9617 (thunk-extern): Likewise.
9618 * doc/extend.texi: Document indirect_branch function attribute.
9619 * doc/invoke.texi: Document -mindirect-branch= option.
9621 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9624 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9626 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9628 * ipa-inline.c (want_inline_small_function_p): Return false if
9629 inlining has already failed with CIF_FINAL_ERROR.
9630 (update_caller_keys): Call want_inline_small_function_p before
9632 (update_callee_keys): Likewise.
9634 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9636 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9638 (rs6000_quadword_masked_address_p): Likewise.
9639 (quad_aligned_load_p): Likewise.
9640 (quad_aligned_store_p): Likewise.
9641 (const_load_sequence_p): Add comment to describe the outer-most loop.
9642 (mimic_memory_attributes_and_flags): New function.
9643 (rs6000_gen_stvx): Likewise.
9644 (replace_swapped_aligned_store): Likewise.
9645 (rs6000_gen_lvx): Likewise.
9646 (replace_swapped_aligned_load): Likewise.
9647 (replace_swapped_load_constant): Capitalize argument name in
9648 comment describing this function.
9649 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9650 and stores that access quad-word aligned addresses and replace
9651 with stvx or lvx instructions when appropriate.
9652 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9653 New function prototype.
9654 (rs6000_quadword_masked_address_p): Likewise.
9655 (rs6000_gen_lvx): Likewise.
9656 (rs6000_gen_stvx): Likewise.
9657 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9658 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9659 when memory address is aligned.
9660 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9661 this split to select lvx instruction when memory address is aligned.
9662 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9663 instruction when memory address is aligned.
9664 (*vsx_le_perm_load_v16qi): Likewise.
9665 (four unnamed splitters): Modify to select the stvx instruction
9666 when memory is aligned.
9668 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9670 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9671 which appears in the queue multiple times.
9673 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9674 Alan Hayward <alan.hayward@arm.com>
9675 David Sherwood <david.sherwood@arm.com>
9677 * tree-vectorizer.h (vec_lower_bound): New structure.
9678 (_loop_vec_info): Add check_nonzero and lower_bounds.
9679 (LOOP_VINFO_CHECK_NONZERO): New macro.
9680 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9681 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9682 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9683 fields. Make seg_len the distance travelled, not including the
9685 (dr_direction_indicator): Declare.
9686 (dr_zero_step_indicator): Likewise.
9687 (dr_known_forward_stride_p): Likewise.
9688 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9690 (runtime_alias_check_p): Allow runtime alias checks with
9692 (operator ==): Compare access_size and align.
9693 (prune_runtime_alias_test_list): Rework for new distinction between
9694 the access_size and seg_len.
9695 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9697 (get_segment_min_max): New function.
9698 (create_intersect_range_checks): Use it.
9699 (dr_step_indicator): New function.
9700 (dr_direction_indicator): Likewise.
9701 (dr_zero_step_indicator): Likewise.
9702 (dr_known_forward_stride_p): Likewise.
9703 * tree-loop-distribution.c (data_ref_segment_size): Return
9704 DR_STEP * (niters - 1).
9705 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9707 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9708 (vect_preserves_scalar_order_p): New function, split out from...
9709 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9710 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9711 (vect_vfa_access_size): New function.
9712 (vect_vfa_align): Likewise.
9713 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9714 (dump_lower_bound): New function.
9715 (vect_check_lower_bound): Likewise.
9716 (vect_small_gap_p): Likewise.
9717 (vectorizable_with_step_bound_p): Likewise.
9718 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9719 depencies if the vectorization factor is 1. Convert the checks
9720 for nonzero steps into checks on the bounds of DR_STEP. Try using
9721 a bunds check for variable steps if the minimum required step is
9722 relatively small. Update calls to the dr_with_seg_len
9723 constructor and to vect_compile_time_alias.
9724 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9726 (vect_loop_versioning): Call it.
9727 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9729 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9731 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9732 Alan Hayward <alan.hayward@arm.com>
9733 David Sherwood <david.sherwood@arm.com>
9735 * doc/sourcebuild.texi (vect_scatter_store): Document.
9736 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9738 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9740 * genopinit.c (main): Add supports_vec_scatter_store and
9741 supports_vec_scatter_store_cached to target_optabs.
9742 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9743 IFN_MASK_SCATTER_STORE.
9744 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9746 * internal-fn.h (internal_store_fn_p): Declare.
9747 (internal_fn_stored_value_index): Likewise.
9748 * internal-fn.c (scatter_store_direct): New macro.
9749 (expand_scatter_store_optab_fn): New function.
9750 (direct_scatter_store_optab_supported_p): New macro.
9751 (internal_store_fn_p): New function.
9752 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9753 IFN_MASK_SCATTER_STORE.
9754 (internal_fn_mask_index): Likewise.
9755 (internal_fn_stored_value_index): New function.
9756 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9758 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9759 * optabs-query.c (supports_vec_scatter_store_p): New function.
9760 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9761 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9762 true for scatter stores.
9763 (vect_gather_scatter_fn_p): Handle scatter stores too.
9764 (vect_check_gather_scatter): Consider using scatter stores if
9765 supports_vec_scatter_store_p.
9766 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9768 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9769 internal_fn_stored_value_index.
9770 (check_load_store_masking): Handle scatter stores too.
9771 (vect_get_store_rhs): Make public.
9772 (vectorizable_call): Use internal_store_fn_p.
9773 (vectorizable_store): Handle scatter store internal functions.
9774 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9775 when deciding whether the end of the group has been reached.
9776 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9777 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9778 (mask_scatter_store<mode>): New insns.
9780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9781 Alan Hayward <alan.hayward@arm.com>
9782 David Sherwood <david.sherwood@arm.com>
9784 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9785 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9786 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9788 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9789 Use vect_truncate_gather_scatter_offset if we can't treat the
9790 operation as a normal gather load or scatter store.
9791 (get_group_load_store_type): Take the gather_scatter_info
9792 as argument. Try using a gather load or scatter store for
9793 single-element groups.
9794 (get_load_store_type): Update calls to get_group_load_store_type
9795 and vect_use_strided_gather_scatters_p.
9797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9798 Alan Hayward <alan.hayward@arm.com>
9799 David Sherwood <david.sherwood@arm.com>
9801 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9802 optional tree argument.
9803 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9805 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9806 but continue to use the current value as a fallback.
9807 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9808 to compare the updates.
9809 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9810 (get_load_store_type): Use it when handling a strided access.
9811 (vect_get_strided_load_store_ops): New function.
9812 (vect_get_data_ptr_increment): Likewise.
9813 (vectorizable_load): Handle strided gather loads. Always pass
9814 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9817 Alan Hayward <alan.hayward@arm.com>
9818 David Sherwood <david.sherwood@arm.com>
9820 * doc/md.texi (gather_load@var{m}): Document.
9821 (mask_gather_load@var{m}): Likewise.
9822 * genopinit.c (main): Add supports_vec_gather_load and
9823 supports_vec_gather_load_cached to target_optabs.
9824 * optabs-tree.c (init_tree_optimization_optabs): Use
9825 ggc_cleared_alloc to allocate target_optabs.
9826 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9827 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9829 * internal-fn.h (internal_load_fn_p): Declare.
9830 (internal_gather_scatter_fn_p): Likewise.
9831 (internal_fn_mask_index): Likewise.
9832 (internal_gather_scatter_fn_supported_p): Likewise.
9833 * internal-fn.c (gather_load_direct): New macro.
9834 (expand_gather_load_optab_fn): New function.
9835 (direct_gather_load_optab_supported_p): New macro.
9836 (direct_internal_fn_optab): New function.
9837 (internal_load_fn_p): Likewise.
9838 (internal_gather_scatter_fn_p): Likewise.
9839 (internal_fn_mask_index): Likewise.
9840 (internal_gather_scatter_fn_supported_p): Likewise.
9841 * optabs-query.c (supports_at_least_one_mode_p): New function.
9842 (supports_vec_gather_load_p): Likewise.
9843 * optabs-query.h (supports_vec_gather_load_p): Declare.
9844 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9845 and memory_type field.
9846 (NUM_PATTERNS): Bump to 15.
9847 * tree-vect-data-refs.c: Include internal-fn.h.
9848 (vect_gather_scatter_fn_p): New function.
9849 (vect_describe_gather_scatter_call): Likewise.
9850 (vect_check_gather_scatter): Try using internal functions for
9851 gather loads. Recognize existing calls to a gather load function.
9852 (vect_analyze_data_refs): Consider using gather loads if
9853 supports_vec_gather_load_p.
9854 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9855 (vect_get_gather_scatter_offset_type): Likewise.
9856 (vect_convert_mask_for_vectype): Likewise.
9857 (vect_add_conversion_to_patterm): Likewise.
9858 (vect_try_gather_scatter_pattern): Likewise.
9859 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9860 (vect_vect_recog_func_ptrs): Add it.
9861 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9862 internal_fn_mask_index and internal_gather_scatter_fn_p.
9863 (check_load_store_masking): Take the gather_scatter_info as an
9864 argument and handle gather loads.
9865 (vect_get_gather_scatter_ops): New function.
9866 (vectorizable_call): Check internal_load_fn_p.
9867 (vectorizable_load): Likewise. Handle gather load internal
9869 (vectorizable_store): Update call to check_load_store_masking.
9870 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9871 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9872 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9873 (aarch64_gather_scale_operand_d): New predicates.
9874 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9875 (mask_gather_load<mode>): New insns.
9877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9878 Alan Hayward <alan.hayward@arm.com>
9879 David Sherwood <david.sherwood@arm.com>
9881 * optabs.def (fold_left_plus_optab): New optab.
9882 * doc/md.texi (fold_left_plus_@var{m}): Document.
9883 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9884 * internal-fn.c (fold_left_direct): Define.
9885 (expand_fold_left_optab_fn): Likewise.
9886 (direct_fold_left_optab_supported_p): Likewise.
9887 * fold-const-call.c (fold_const_fold_left): New function.
9888 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9889 * tree-parloops.c (valid_reduction_p): New function.
9890 (gather_scalar_reductions): Use it.
9891 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9892 (vect_finish_replace_stmt): Declare.
9893 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9894 (needs_fold_left_reduction_p): New function, split out from...
9895 (vect_is_simple_reduction): ...here. Accept reductions that
9896 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9897 (vect_force_simple_reduction): Also store the reduction type in
9898 the assignment's STMT_VINFO_REDUC_TYPE.
9899 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9900 (merge_with_identity): New function.
9901 (vect_expand_fold_left): Likewise.
9902 (vectorize_fold_left_reduction): Likewise.
9903 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9904 scalar phi in place for it. Check for target support and reject
9905 cases that would reassociate the operation. Defer the transform
9906 phase to vectorize_fold_left_reduction.
9907 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9908 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9909 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9913 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9914 call to ifc_temp_var.
9916 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9917 Alan Hayward <alan.hayward@arm.com>
9918 David Sherwood <david.sherwood@arm.com>
9920 * target.def (legitimize_address_displacement): Take the original
9921 offset as a poly_int.
9922 * targhooks.h (default_legitimize_address_displacement): Update
9924 * targhooks.c (default_legitimize_address_displacement): Likewise.
9925 * doc/tm.texi: Regenerate.
9926 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9927 as an argument, moving assert of ad->disp == ad->disp_term to...
9928 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9929 Try calling targetm.legitimize_address_displacement before expanding
9930 the address rather than afterwards, and adjust for the new interface.
9931 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9932 Match the new hook interface. Handle SVE addresses.
9933 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9936 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9938 * Makefile.in (OBJS): Add early-remat.o.
9939 * target.def (select_early_remat_modes): New hook.
9940 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9941 * doc/tm.texi: Regenerate.
9942 * targhooks.h (default_select_early_remat_modes): Declare.
9943 * targhooks.c (default_select_early_remat_modes): New function.
9944 * timevar.def (TV_EARLY_REMAT): New timevar.
9945 * passes.def (pass_early_remat): New pass.
9946 * tree-pass.h (make_pass_early_remat): Declare.
9947 * early-remat.c: New file.
9948 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9950 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9953 Alan Hayward <alan.hayward@arm.com>
9954 David Sherwood <david.sherwood@arm.com>
9956 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9957 vfm1 with a bound_epilog parameter.
9958 (vect_do_peeling): Update calls accordingly, and move the prologue
9959 call earlier in the function. Treat the base bound_epilog as 0 for
9960 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9961 this base when peeling for gaps.
9962 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9963 with fully-masked loops.
9964 (vect_estimate_min_profitable_iters): Handle the single peeled
9965 iteration in that case.
9967 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9968 Alan Hayward <alan.hayward@arm.com>
9969 David Sherwood <david.sherwood@arm.com>
9971 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9972 single-element interleaving even if the size is not a power of 2.
9973 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9974 accesses for single-element interleaving if the group size is
9977 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9978 Alan Hayward <alan.hayward@arm.com>
9979 David Sherwood <david.sherwood@arm.com>
9981 * doc/md.texi (fold_extract_last_@var{m}): Document.
9982 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9983 * optabs.def (fold_extract_last_optab): New optab.
9984 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9985 * internal-fn.c (fold_extract_direct): New macro.
9986 (expand_fold_extract_optab_fn): Likewise.
9987 (direct_fold_extract_optab_supported_p): Likewise.
9988 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9989 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9990 EXTRACT_LAST_REDUCTION.
9991 (get_initial_def_for_reduction): Do not create an initial vector
9992 for EXTRACT_LAST_REDUCTION reductions.
9993 (vectorizable_reduction): Leave the scalar phi in place for
9994 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9995 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9996 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9997 transform phase to vectorizable_condition.
9998 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10000 (vect_finish_stmt_generation): ...here.
10001 (vect_finish_replace_stmt): New function.
10002 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10003 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10005 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10008 Alan Hayward <alan.hayward@arm.com>
10009 David Sherwood <david.sherwood@arm.com>
10011 * doc/md.texi (extract_last_@var{m}): Document.
10012 * optabs.def (extract_last_optab): New optab.
10013 * internal-fn.def (EXTRACT_LAST): New internal function.
10014 * internal-fn.c (cond_unary_direct): New macro.
10015 (expand_cond_unary_optab_fn): Likewise.
10016 (direct_cond_unary_optab_supported_p): Likewise.
10017 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10018 loops using EXTRACT_LAST.
10019 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10020 (extract_last_<mode>): ...this optab.
10021 (vec_extract<mode><Vel>): Update accordingly.
10023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10024 Alan Hayward <alan.hayward@arm.com>
10025 David Sherwood <david.sherwood@arm.com>
10027 * target.def (empty_mask_is_expensive): New hook.
10028 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10029 * doc/tm.texi: Regenerate.
10030 * targhooks.h (default_empty_mask_is_expensive): Declare.
10031 * targhooks.c (default_empty_mask_is_expensive): New function.
10032 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10033 if the target says that empty masks are expensive.
10034 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10036 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10039 Alan Hayward <alan.hayward@arm.com>
10040 David Sherwood <david.sherwood@arm.com>
10042 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10043 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10044 (vect_use_loop_mask_for_alignment_p): New function.
10045 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10046 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10047 niters_skip argument. Make sure that the first niters_skip elements
10048 of the first iteration are inactive.
10049 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10050 Update call to vect_set_loop_masks_directly.
10051 (get_misalign_in_elems): New function, split out from...
10052 (vect_gen_prolog_loop_niters): ...here.
10053 (vect_update_init_of_dr): Take a code argument that specifies whether
10054 the adjustment should be added or subtracted.
10055 (vect_update_init_of_drs): Likewise.
10056 (vect_prepare_for_masked_peels): New function.
10057 (vect_do_peeling): Skip prologue peeling if we're using a mask
10058 instead. Update call to vect_update_inits_of_drs.
10059 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10061 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10062 alignment. Do not include the number of peeled iterations in
10063 the minimum threshold in that case.
10064 (vectorizable_induction): Adjust the start value down by
10065 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10066 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10067 Take the number of skipped iterations into account when calculating
10069 * tree-vect-stmts.c (vect_gen_while_not): New function.
10071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10072 Alan Hayward <alan.hayward@arm.com>
10073 David Sherwood <david.sherwood@arm.com>
10075 * doc/sourcebuild.texi (vect_fully_masked): Document.
10076 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10077 default value to 0.
10078 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10080 (vect_analyze_loop_2): ...here. Don't check the vectorization
10081 factor against the number of loop iterations if the loop is
10084 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10085 Alan Hayward <alan.hayward@arm.com>
10086 David Sherwood <david.sherwood@arm.com>
10088 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10089 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10090 (dump_groups): Update accordingly.
10091 (iv_use::mem_type): New member variable.
10092 (address_p): New function.
10093 (record_use): Add a mem_type argument and initialize the new
10095 (record_group_use): Add a mem_type argument. Use address_p.
10096 Remove obsolete null checks of base_object. Update call to record_use.
10097 (find_interesting_uses_op): Update call to record_group_use.
10098 (find_interesting_uses_cond): Likewise.
10099 (find_interesting_uses_address): Likewise.
10100 (get_mem_type_for_internal_fn): New function.
10101 (find_address_like_use): Likewise.
10102 (find_interesting_uses_stmt): Try find_address_like_use before
10103 calling find_interesting_uses_op.
10104 (addr_offset_valid_p): Use the iv mem_type field as the type
10105 of the addressed memory.
10106 (add_autoinc_candidates): Likewise.
10107 (get_address_cost): Likewise.
10108 (split_small_address_groups_p): Use address_p.
10109 (split_address_groups): Likewise.
10110 (add_iv_candidate_for_use): Likewise.
10111 (autoinc_possible_for_pair): Likewise.
10112 (rewrite_groups): Likewise.
10113 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10114 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10115 (get_alias_ptr_type_for_ptr_address): New function.
10116 (rewrite_use_address): Rewrite address uses in calls that were
10117 identified by find_address_like_use.
10119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10120 Alan Hayward <alan.hayward@arm.com>
10121 David Sherwood <david.sherwood@arm.com>
10123 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10125 * gimple-expr.h (is_gimple_addressable: Likewise.
10126 * gimple-expr.c (is_gimple_address): Likewise.
10127 * internal-fn.c (expand_call_mem_ref): New function.
10128 (expand_mask_load_optab_fn): Use it.
10129 (expand_mask_store_optab_fn): Likewise.
10131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10132 Alan Hayward <alan.hayward@arm.com>
10133 David Sherwood <david.sherwood@arm.com>
10135 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10136 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10137 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10138 (cond_umax@var{mode}): Document.
10139 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10140 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10141 (cond_umin_optab, cond_umax_optab): New optabs.
10142 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10143 (COND_IOR, COND_XOR): New internal functions.
10144 * internal-fn.h (get_conditional_internal_fn): Declare.
10145 * internal-fn.c (cond_binary_direct): New macro.
10146 (expand_cond_binary_optab_fn): Likewise.
10147 (direct_cond_binary_optab_supported_p): Likewise.
10148 (get_conditional_internal_fn): New function.
10149 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10150 Cope with reduction statements that are vectorized as calls rather
10152 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10153 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10154 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10155 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10156 (UNSPEC_COND_EOR): New unspecs.
10157 (optab): Add mappings for them.
10158 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10159 (sve_int_op, sve_fp_op): New int attributes.
10161 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10162 Alan Hayward <alan.hayward@arm.com>
10163 David Sherwood <david.sherwood@arm.com>
10165 * optabs.def (while_ult_optab): New optab.
10166 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10167 * internal-fn.def (WHILE_ULT): New internal function.
10168 * internal-fn.h (direct_internal_fn_supported_p): New override
10169 that takes two types as argument.
10170 * internal-fn.c (while_direct): New macro.
10171 (expand_while_optab_fn): New function.
10172 (convert_optab_supported_p): Likewise.
10173 (direct_while_optab_supported_p): New macro.
10174 * wide-int.h (wi::udiv_ceil): New function.
10175 * tree-vectorizer.h (rgroup_masks): New structure.
10176 (vec_loop_masks): New typedef.
10177 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10178 and fully_masked_p.
10179 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10180 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10181 (vect_max_vf): New function.
10182 (slpeel_make_loop_iterate_ntimes): Delete.
10183 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10184 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10185 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10186 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10187 internal-fn.h, stor-layout.h and optabs-query.h.
10188 (vect_set_loop_mask): New function.
10189 (add_preheader_seq): Likewise.
10190 (add_header_seq): Likewise.
10191 (interleave_supported_p): Likewise.
10192 (vect_maybe_permute_loop_masks): Likewise.
10193 (vect_set_loop_masks_directly): Likewise.
10194 (vect_set_loop_condition_masked): Likewise.
10195 (vect_set_loop_condition_unmasked): New function, split out from
10196 slpeel_make_loop_iterate_ntimes.
10197 (slpeel_make_loop_iterate_ntimes): Rename to..
10198 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10199 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10200 (vect_do_peeling): Update call accordingly.
10201 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10203 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10204 mask_compare_type, can_fully_mask_p and fully_masked_p.
10205 (release_vec_loop_masks): New function.
10206 (_loop_vec_info): Use it to free the loop masks.
10207 (can_produce_all_loop_masks_p): New function.
10208 (vect_get_max_nscalars_per_iter): Likewise.
10209 (vect_verify_full_masking): Likewise.
10210 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10211 retries, and free the mask rgroups before retrying. Check loop-wide
10212 reasons for disallowing fully-masked loops. Make the final decision
10213 about whether use a fully-masked loop or not.
10214 (vect_estimate_min_profitable_iters): Do not assume that peeling
10215 for the number of iterations will be needed for fully-masked loops.
10216 (vectorizable_reduction): Disable fully-masked loops.
10217 (vectorizable_live_operation): Likewise.
10218 (vect_halve_mask_nunits): New function.
10219 (vect_double_mask_nunits): Likewise.
10220 (vect_record_loop_mask): Likewise.
10221 (vect_get_loop_mask): Likewise.
10222 (vect_transform_loop): Handle the case in which the final loop
10223 iteration might handle a partial vector. Call vect_set_loop_condition
10224 instead of slpeel_make_loop_iterate_ntimes.
10225 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10226 (check_load_store_masking): New function.
10227 (prepare_load_store_mask): Likewise.
10228 (vectorizable_store): Handle fully-masked loops.
10229 (vectorizable_load): Likewise.
10230 (supportable_widening_operation): Use vect_halve_mask_nunits for
10232 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10233 (vect_gen_while): New function.
10234 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10235 (aarch64_uqdec<mode>): New insn.
10237 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10238 Alan Hayward <alan.hayward@arm.com>
10239 David Sherwood <david.sherwood@arm.com>
10241 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10242 (reduc_xor_scal_optab): New optabs.
10243 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10244 (reduc_xor_scal_@var{m}): Document.
10245 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10246 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10247 internal functions.
10248 * fold-const-call.c (fold_const_call): Handle them.
10249 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10250 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10251 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10252 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10253 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10254 (UNSPEC_XORV): New unspecs.
10255 (optab): Add entries for them.
10256 (BITWISEV): New int iterator.
10257 (bit_reduc_op): New int attributes.
10259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10260 Alan Hayward <alan.hayward@arm.com>
10261 David Sherwood <david.sherwood@arm.com>
10263 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10264 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10265 * optabs.def (vec_shl_insert_optab): New optab.
10266 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10267 (duplicate_and_interleave): Likewise.
10268 * tree-vect-loop.c: Include internal-fn.h.
10269 (neutral_op_for_slp_reduction): New function, split out from
10270 get_initial_defs_for_reduction.
10271 (get_initial_def_for_reduction): Handle option 2 for variable-length
10272 vectors by loading the neutral value into a vector and then shifting
10273 the initial value into element 0.
10274 (get_initial_defs_for_reduction): Replace the code argument with
10275 the neutral value calculated by neutral_op_for_slp_reduction.
10276 Use gimple_build_vector for constant-length vectors.
10277 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10278 but the first group_size elements have a neutral value.
10279 Use duplicate_and_interleave otherwise.
10280 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10281 Update call to get_initial_defs_for_reduction. Handle SLP
10282 reductions for variable-length vectors by creating one vector
10283 result for each scalar result, with the elements associated
10284 with other scalar results stubbed out with the neutral value.
10285 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10286 Require IFN_VEC_SHL_INSERT for double reductions on
10287 variable-length vectors, or SLP reductions that have
10288 a neutral value. Require can_duplicate_and_interleave_p
10289 support for variable-length unchained SLP reductions if there
10290 is no neutral value, such as for MIN/MAX reductions. Also require
10291 the number of vector elements to be a multiple of the number of
10292 SLP statements when doing variable-length unchained SLP reductions.
10293 Update call to vect_create_epilog_for_reduction.
10294 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10295 and remove initial values.
10296 (duplicate_and_interleave): Make public.
10297 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10298 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10300 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10301 Alan Hayward <alan.hayward@arm.com>
10302 David Sherwood <david.sherwood@arm.com>
10304 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10305 (can_duplicate_and_interleave_p): New function.
10306 (vect_get_and_check_slp_defs): Take the vector of statements
10307 rather than just the current one. Remove excess parentheses.
10308 Restriction rejectinon of vect_constant_def and vect_external_def
10309 for variable-length vectors to boolean types, or types for which
10310 can_duplicate_and_interleave_p is false.
10311 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10312 (duplicate_and_interleave): New function.
10313 (vect_get_constant_vectors): Use gimple_build_vector for
10314 constant-length vectors and suitable variable-length constant
10315 vectors. Use duplicate_and_interleave for other variable-length
10316 vectors. Don't defer the update when inserting new statements.
10318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10319 Alan Hayward <alan.hayward@arm.com>
10320 David Sherwood <david.sherwood@arm.com>
10322 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10323 min_profitable_iters doesn't go negative.
10325 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10326 Alan Hayward <alan.hayward@arm.com>
10327 David Sherwood <david.sherwood@arm.com>
10329 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10330 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10331 * optabs.def (vec_mask_load_lanes_optab): New optab.
10332 (vec_mask_store_lanes_optab): Likewise.
10333 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10334 (MASK_STORE_LANES): Likewise.
10335 * internal-fn.c (mask_load_lanes_direct): New macro.
10336 (mask_store_lanes_direct): Likewise.
10337 (expand_mask_load_optab_fn): Handle masked operations.
10338 (expand_mask_load_lanes_optab_fn): New macro.
10339 (expand_mask_store_optab_fn): Handle masked operations.
10340 (expand_mask_store_lanes_optab_fn): New macro.
10341 (direct_mask_load_lanes_optab_supported_p): Likewise.
10342 (direct_mask_store_lanes_optab_supported_p): Likewise.
10343 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10345 (vect_load_lanes_supported): Likewise.
10346 * tree-vect-data-refs.c (strip_conversion): New function.
10347 (can_group_stmts_p): Likewise.
10348 (vect_analyze_data_ref_accesses): Use it instead of checking
10349 for a pair of assignments.
10350 (vect_store_lanes_supported): Take a masked_p parameter.
10351 (vect_load_lanes_supported): Likewise.
10352 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10353 vect_store_lanes_supported and vect_load_lanes_supported.
10354 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10355 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10356 parameter. Don't allow gaps for masked accesses.
10357 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10358 and vect_load_lanes_supported.
10359 (get_load_store_type): Take a masked_p parameter and update
10360 call to get_group_load_store_type.
10361 (vectorizable_store): Update call to get_load_store_type.
10362 Handle IFN_MASK_STORE_LANES.
10363 (vectorizable_load): Update call to get_load_store_type.
10364 Handle IFN_MASK_LOAD_LANES.
10366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10367 Alan Hayward <alan.hayward@arm.com>
10368 David Sherwood <david.sherwood@arm.com>
10370 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10372 * config/aarch64/aarch64-protos.h
10373 (aarch64_sve_struct_memory_operand_p): Declare.
10374 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10375 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10376 (VPRED, vpred): Handle SVE structure modes.
10377 * config/aarch64/constraints.md (Utx): New constraint.
10378 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10379 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10380 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10381 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10382 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10383 structure modes. Split into pieces after RA.
10384 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10385 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10387 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10388 SVE structure modes.
10389 (aarch64_classify_address): Likewise.
10390 (sizetochar): Move earlier in file.
10391 (aarch64_print_operand): Handle SVE register lists.
10392 (aarch64_array_mode): New function.
10393 (aarch64_sve_struct_memory_operand_p): Likewise.
10394 (TARGET_ARRAY_MODE): Redefine.
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 * target.def (array_mode): New target hook.
10401 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10402 * doc/tm.texi: Regenerate.
10403 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10404 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10405 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10406 targetm.array_mode.
10407 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10410 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10411 Alan Hayward <alan.hayward@arm.com>
10412 David Sherwood <david.sherwood@arm.com>
10414 * fold-const.c (fold_binary_loc): Check the argument types
10415 rather than the result type when testing for a vector operation.
10417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10419 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10420 * doc/tm.texi: Regenerate.
10422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10423 Alan Hayward <alan.hayward@arm.com>
10424 David Sherwood <david.sherwood@arm.com>
10426 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10427 (sve): Document new AArch64 extension.
10428 * doc/md.texi (w): Extend the description of the AArch64
10429 constraint to include SVE vectors.
10430 (Upl, Upa): Document new AArch64 predicate constraints.
10431 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10433 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10434 (msve-vector-bits=): New option.
10435 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10436 SVE when these are disabled.
10437 (sve): New extension.
10438 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10439 modes. Adjust their number of units based on aarch64_sve_vg.
10440 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10441 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10442 aarch64_addr_query_type.
10443 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10444 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10445 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10446 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10447 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10448 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10449 (aarch64_simd_imm_zero_p): Delete.
10450 (aarch64_check_zero_based_sve_index_immediate): Declare.
10451 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10452 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10453 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10454 (aarch64_sve_float_mul_immediate_p): Likewise.
10455 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10456 rather than an rtx.
10457 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10458 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10459 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10460 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10461 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10462 (aarch64_regmode_natural_size): Likewise.
10463 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10464 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10466 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10467 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10468 for VG and the SVE predicate registers.
10469 (V_ALIASES): Add a "z"-prefixed alias.
10470 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10471 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10472 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10473 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10474 (REG_CLASS_NAMES): Add entries for them.
10475 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10476 and the predicate registers.
10477 (aarch64_sve_vg): Declare.
10478 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10479 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10480 (REGMODE_NATURAL_SIZE): Define.
10481 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10483 * config/aarch64/aarch64.c: Include cfgrtl.h.
10484 (simd_immediate_info): Add a constructor for series vectors,
10485 and an associated step field.
10486 (aarch64_sve_vg): New variable.
10487 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10488 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10489 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10490 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10491 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10492 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10493 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10494 (aarch64_get_mask_mode): New functions.
10495 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10496 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10497 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10498 predicate modes and predicate registers. Explicitly restrict
10499 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10500 to store a vector mode if it is recognized by
10501 aarch64_classify_vector_mode.
10502 (aarch64_regmode_natural_size): New function.
10503 (aarch64_hard_regno_caller_save_mode): Return the original mode
10505 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10506 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10507 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10508 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10510 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10511 does not overlap dest if the function is frame-related. Handle
10513 (aarch64_split_add_offset): New function.
10514 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10515 them aarch64_add_offset.
10516 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10517 and update call to aarch64_sub_sp.
10518 (aarch64_add_cfa_expression): New function.
10519 (aarch64_expand_prologue): Pass extra temporary registers to the
10520 functions above. Handle the case in which we need to emit new
10521 DW_CFA_expressions for registers that were originally saved
10522 relative to the stack pointer, but now have to be expressed
10523 relative to the frame pointer.
10524 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10526 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10527 IP0 and IP1 values for SVE frames.
10528 (aarch64_expand_vec_series): New function.
10529 (aarch64_expand_sve_widened_duplicate): Likewise.
10530 (aarch64_expand_sve_const_vector): Likewise.
10531 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10532 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10533 into the register, rather than emitting a SET directly.
10534 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10535 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10536 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10537 (offset_9bit_signed_scaled_p): New functions.
10538 (aarch64_replicate_bitmask_imm): New function.
10539 (aarch64_bitmask_imm): Use it.
10540 (aarch64_cannot_force_const_mem): Reject expressions involving
10541 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10542 (aarch64_classify_index): Handle SVE indices, by requiring
10543 a plain register index with a scale that matches the element size.
10544 (aarch64_classify_address): Handle SVE addresses. Assert that
10545 the mode of the address is VOIDmode or an integer mode.
10546 Update call to aarch64_classify_symbol.
10547 (aarch64_classify_symbolic_expression): Update call to
10548 aarch64_classify_symbol.
10549 (aarch64_const_vec_all_in_range_p): New function.
10550 (aarch64_print_vector_float_operand): Likewise.
10551 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10552 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10553 and the FP immediates 1.0 and 0.5.
10554 (aarch64_print_address_internal): Handle SVE addresses.
10555 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10556 (aarch64_regno_regclass): Handle predicate registers.
10557 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10559 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10560 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10561 (aarch64_convert_sve_vector_bits): New function.
10562 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10563 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10564 rather than an rtx.
10565 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10566 Handle SVE vector and predicate modes. Accept VL-based constants
10567 that need only one temporary register, and VL offsets that require
10568 no temporary registers.
10569 (aarch64_conditional_register_usage): Mark the predicate registers
10570 as fixed if SVE isn't available.
10571 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10572 Return true for SVE vector and predicate modes.
10573 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10574 rather than an unsigned int. Handle SVE modes.
10575 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10577 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10579 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10580 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10581 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10582 (aarch64_sve_float_mul_immediate_p): New functions.
10583 (aarch64_sve_valid_immediate): New function.
10584 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10585 Explicitly reject structure modes. Check for INDEX constants.
10586 Handle PTRUE and PFALSE constants.
10587 (aarch64_check_zero_based_sve_index_immediate): New function.
10588 (aarch64_simd_imm_zero_p): Delete.
10589 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10590 vector modes. Accept constants in the range of CNT[BHWD].
10591 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10592 ask for an Advanced SIMD mode.
10593 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10594 (aarch64_simd_vector_alignment): Handle SVE predicates.
10595 (aarch64_vectorize_preferred_vector_alignment): New function.
10596 (aarch64_simd_vector_alignment_reachable): Use it instead of
10598 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10599 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10601 (MAX_VECT_LEN): Delete.
10602 (expand_vec_perm_d): Add a vec_flags field.
10603 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10604 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10605 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10607 (aarch64_evpc_rev): Rename to...
10608 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10609 (aarch64_evpc_rev_global): New function.
10610 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10611 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10613 (aarch64_evpc_sve_tbl): New function.
10614 (aarch64_expand_vec_perm_const_1): Update after rename of
10615 aarch64_evpc_rev. Handle SVE permutes too, trying
10616 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10617 than aarch64_evpc_tbl.
10618 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10619 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10620 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10621 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10622 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10623 (aarch64_expand_sve_vcond): New functions.
10624 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10625 of aarch64_vector_mode_p.
10626 (aarch64_dwarf_poly_indeterminate_value): New function.
10627 (aarch64_compute_pressure_classes): Likewise.
10628 (aarch64_can_change_mode_class): Likewise.
10629 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10630 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10631 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10632 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10633 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10634 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10635 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10636 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10638 (Dn, Dl, Dr): Accept const as well as const_vector.
10639 (Dz): Likewise. Compare against CONST0_RTX.
10640 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10641 of "vector" where appropriate.
10642 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10643 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10644 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10645 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10646 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10647 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10648 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10649 (v_int_equiv): Extend to SVE modes.
10650 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10652 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10653 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10654 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10655 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10656 (SVE_COND_FP_CMP): New int iterators.
10657 (perm_hilo): Handle the new unpack unspecs.
10658 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10660 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10661 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10662 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10663 (aarch64_equality_operator, aarch64_constant_vector_operand)
10664 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10665 (aarch64_sve_nonimmediate_operand): Likewise.
10666 (aarch64_sve_general_operand): Likewise.
10667 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10668 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10669 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10670 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10671 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10672 (aarch64_sve_float_arith_immediate): Likewise.
10673 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10674 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10675 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10676 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10677 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10678 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10679 (aarch64_sve_float_arith_operand): Likewise.
10680 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10681 (aarch64_sve_float_mul_operand): Likewise.
10682 (aarch64_sve_vec_perm_operand): Likewise.
10683 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10684 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10685 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10686 as well as const_vector.
10687 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10688 in file. Use CONST0_RTX and CONSTM1_RTX.
10689 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10690 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10691 Use aarch64_simd_imm_zero.
10692 * config/aarch64/aarch64-sve.md: New file.
10693 * config/aarch64/aarch64.md: Include it.
10694 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10695 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10696 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10697 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10698 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10699 (sve): New attribute.
10700 (enabled): Disable instructions with the sve attribute unless
10702 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10703 aarch64_expand_mov_immediate.
10704 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10705 CNT[BHSD] immediates.
10706 (movti): Split CONST_POLY_INT moves into two halves.
10707 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10708 Split additions that need a temporary here if the destination
10709 is the stack pointer.
10710 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10711 (*add<mode>3_poly_1): New instruction.
10712 (set_clobber_cc): New expander.
10714 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10716 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10717 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10718 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10719 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10720 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10721 Change innermode from fixed_mode_size to machine_mode.
10722 (simplify_subreg): Update call accordingly. Handle a constant-sized
10723 subreg of a variable-length CONST_VECTOR.
10725 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10726 Alan Hayward <alan.hayward@arm.com>
10727 David Sherwood <david.sherwood@arm.com>
10729 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10730 (add_offset_to_base): New function, split out from...
10731 (create_mem_ref): ...here. When handling a scale other than 1,
10732 check first whether the address is valid without the offset.
10733 Add it into the base if so, leaving the index and scale as-is.
10735 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10738 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10739 fold_for_warn before checking if arg2 is INTEGER_CST.
10741 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10743 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10744 (store_multiple_operation): Delete.
10745 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10746 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10747 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10748 guarded by TARGET_STRING.
10749 (rs6000_output_load_multiple): Delete.
10750 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10751 OPTION_MASK_STRING / TARGET_STRING handling.
10752 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10753 (const rs6000_opt_masks) <"string">: Change mask to 0.
10754 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10755 (MASK_STRING): Delete.
10756 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10758 (load_multiple): Delete.
10765 (store_multiple): Delete.
10772 (movmemsi_8reg): Delete.
10773 (corresponding unnamed define_insn): Delete.
10774 (movmemsi_6reg): Delete.
10775 (corresponding unnamed define_insn): Delete.
10776 (movmemsi_4reg): Delete.
10777 (corresponding unnamed define_insn): Delete.
10778 (movmemsi_2reg): Delete.
10779 (corresponding unnamed define_insn): Delete.
10780 (movmemsi_1reg): Delete.
10781 (corresponding unnamed define_insn): Delete.
10782 * config/rs6000/rs6000.opt (mno-string): New.
10783 (mstring): Replace by deprecation warning stub.
10784 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10786 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10788 * regrename.c (regrename_do_replace): If replacing the same
10789 reg multiple times, try to reuse last created gen_raw_REG.
10792 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10793 main to workaround a bug in GDB.
10795 2018-01-12 Tom de Vries <tom@codesourcery.com>
10798 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10800 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10802 PR rtl-optimization/80481
10803 * ira-color.c (get_cap_member): New function.
10804 (allocnos_conflict_by_live_ranges_p): Use it.
10805 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10806 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10808 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10811 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10812 (*saddl_se_1): Ditto.
10813 (*ssubsi_1): Ditto.
10814 (*ssubl_se_1): Ditto.
10816 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10818 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10819 rather than wi::to_widest for DR_INITs.
10820 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10821 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10822 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10824 (vect_analyze_group_access_1): Note that here.
10826 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10828 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10829 polynomial type sizes.
10831 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10833 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10834 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10835 (gimple_add_tmp_var): Likewise.
10837 2018-01-12 Martin Liska <mliska@suse.cz>
10839 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10840 (gimple_alloc_sizes): Likewise.
10841 (dump_gimple_statistics): Use PRIu64 in printf format.
10842 * gimple.h: Change uint64_t to int.
10844 2018-01-12 Martin Liska <mliska@suse.cz>
10846 * tree-core.h: Use uint64_t instead of int.
10847 * tree.c (tree_node_counts): Likewise.
10848 (tree_node_sizes): Likewise.
10849 (dump_tree_statistics): Use PRIu64 in printf format.
10851 2018-01-12 Martin Liska <mliska@suse.cz>
10853 * Makefile.in: As qsort_chk is implemented in vec.c, add
10854 vec.o to linkage of gencfn-macros.
10855 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10856 passing the info to record_node_allocation_statistics.
10857 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10859 * ggc-common.c (struct ggc_usage): Add operator== and use
10860 it in operator< and compare function.
10861 * mem-stats.h (struct mem_usage): Likewise.
10862 * vec.c (struct vec_usage): Remove operator< and compare
10863 function. Can be simply inherited.
10865 2018-01-12 Martin Jambor <mjambor@suse.cz>
10868 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10869 * tree-ssa-math-opts.c: Include domwalk.h.
10870 (convert_mult_to_fma_1): New function.
10871 (fma_transformation_info): New type.
10872 (fma_deferring_state): Likewise.
10873 (cancel_fma_deferring): New function.
10874 (result_of_phi): Likewise.
10875 (last_fma_candidate_feeds_initial_phi): Likewise.
10876 (convert_mult_to_fma): Added deferring logic, split actual
10877 transformation to convert_mult_to_fma_1.
10878 (math_opts_dom_walker): New type.
10879 (math_opts_dom_walker::after_dom_children): New method, body moved
10880 here from pass_optimize_widening_mul::execute, added deferring logic
10882 (pass_optimize_widening_mul::execute): Moved most of code to
10883 math_opts_dom_walker::after_dom_children.
10884 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10885 * config/i386/i386.c (ix86_option_override_internal): Added
10886 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10888 2018-01-12 Richard Biener <rguenther@suse.de>
10891 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10892 inline instance vars.
10894 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10897 * config/rx/rx.c (rx_is_restricted_memory_address):
10898 Handle SUBREG case.
10900 2018-01-12 Richard Biener <rguenther@suse.de>
10902 PR tree-optimization/80846
10903 * target.def (split_reduction): New target hook.
10904 * targhooks.c (default_split_reduction): New function.
10905 * targhooks.h (default_split_reduction): Declare.
10906 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10907 target requests first reduce vectors by combining low and high
10909 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10910 (get_vectype_for_scalar_type_and_size): Export.
10911 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10912 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10913 * doc/tm.texi: Regenerate.
10914 * config/i386/i386.c (ix86_split_reduction): Implement
10915 TARGET_VECTORIZE_SPLIT_REDUCTION.
10917 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10920 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10921 in PIC mode except for TARGET_VXWORKS_RTP.
10922 * config/sparc/sparc.c: Include cfgrtl.h.
10923 (TARGET_INIT_PIC_REG): Define.
10924 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10925 (sparc_pic_register_p): New predicate.
10926 (sparc_legitimate_address_p): Use it.
10927 (sparc_legitimize_pic_address): Likewise.
10928 (sparc_delegitimize_address): Likewise.
10929 (sparc_mode_dependent_address_p): Likewise.
10930 (gen_load_pcrel_sym): Remove 4th parameter.
10931 (load_got_register): Adjust call to above. Remove obsolete stuff.
10932 (sparc_expand_prologue): Do not call load_got_register here.
10933 (sparc_flat_expand_prologue): Likewise.
10934 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10935 (sparc_use_pseudo_pic_reg): New function.
10936 (sparc_init_pic_reg): Likewise.
10937 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10938 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10940 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10942 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10943 Add item for branch_cost.
10945 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10947 PR rtl-optimization/83565
10948 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10949 not extend the result to a larger mode for rotate operations.
10950 (num_sign_bit_copies1): Likewise.
10952 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10955 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10957 Use values-Xc.o for -pedantic.
10958 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10960 2018-01-12 Martin Liska <mliska@suse.cz>
10963 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10965 (possible_polymorphic_call_targets): Use it.
10966 (ipa_devirt): Likewise.
10968 2018-01-12 Martin Liska <mliska@suse.cz>
10970 * profile-count.h (enum profile_quality): Use 0 as invalid
10971 enum value of profile_quality.
10973 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10975 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10976 -mext-string options.
10978 2018-01-12 Richard Biener <rguenther@suse.de>
10980 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10981 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10982 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10984 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10986 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10988 * configure.ac (--with-long-double-format): Add support for the
10989 configuration option to change the default long double format on
10991 * config.gcc (powerpc*-linux*-*): Likewise.
10992 * configure: Regenerate.
10993 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10994 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10995 used without modification.
10997 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10999 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11000 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11001 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11002 MISC_BUILTIN_SPEC_BARRIER.
11003 (rs6000_init_builtins): Likewise.
11004 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11006 (speculation_barrier): New define_insn.
11007 * doc/extend.texi: Document __builtin_speculation_barrier.
11009 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11012 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11013 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11014 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11016 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11017 integral modes instead of "ss" and "sd".
11018 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11019 vectors with 32-bit and 64-bit elements.
11020 (vecdupssescalarmodesuffix): New mode attribute.
11021 (vec_dup<mode>): Use it.
11023 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11026 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11027 frame if argument is passed on stack.
11029 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11032 * ree.c (combine_reaching_defs): Optimize also
11033 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11034 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11036 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11038 PR middle-end/83189
11039 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11041 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11043 PR middle-end/83718
11044 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11045 after they are computed.
11047 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11049 PR tree-optimization/83695
11050 * gimple-loop-linterchange.cc
11051 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11052 reset cached scev information after interchange.
11053 (pass_linterchange::execute): Remove call to scev_reset_htab.
11055 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11057 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11058 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11059 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11060 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11061 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11062 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11063 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11064 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11065 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11066 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11067 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11068 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11069 (V_lane_reg): Likewise.
11070 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11072 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11073 (vfmal_lane_low<mode>_intrinsic,
11074 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11075 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11076 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11077 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11078 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11079 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11081 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11083 * config/arm/arm-cpus.in (fp16fml): New feature.
11084 (ALL_SIMD): Add fp16fml.
11085 (armv8.2-a): Add fp16fml as an option.
11086 (armv8.3-a): Likewise.
11087 (armv8.4-a): Add fp16fml as part of fp16.
11088 * config/arm/arm.h (TARGET_FP16FML): Define.
11089 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11091 * config/arm/arm-modes.def (V2HF): Define.
11092 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11093 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11094 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11095 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11096 vfmsl_low, vfmsl_high): New set of builtins.
11097 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11098 (vfml_op): New code attribute.
11099 (VFMLHALVES): New int iterator.
11100 (VFML, VFMLSEL): New mode attributes.
11101 (V_reg): Define mapping for V2HF.
11102 (V_hi, V_lo): New mode attributes.
11103 (VF_constraint): Likewise.
11104 (vfml_half, vfml_half_selector): New int attributes.
11105 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11107 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11108 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11110 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11111 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11112 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11113 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11115 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11116 Document new effective target and option set.
11118 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11120 * config/arm/arm-cpus.in (armv8_4): New feature.
11121 (ARMv8_4a): New fgroup.
11122 (armv8.4-a): New arch.
11123 * config/arm/arm-tables.opt: Regenerate.
11124 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11125 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11126 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11127 Add matching rules for -march=armv8.4-a and extensions.
11128 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11130 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11133 * config/rx/rx.md (BW): New mode attribute.
11134 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11136 2018-01-11 Richard Biener <rguenther@suse.de>
11138 PR tree-optimization/83435
11139 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11140 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11141 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11143 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11144 Alan Hayward <alan.hayward@arm.com>
11145 David Sherwood <david.sherwood@arm.com>
11147 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11149 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11150 (aarch64_print_address_internal): Use it to check for a zero offset.
11152 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11153 Alan Hayward <alan.hayward@arm.com>
11154 David Sherwood <david.sherwood@arm.com>
11156 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11157 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11158 Return a poly_int64 rather than a HOST_WIDE_INT.
11159 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11160 rather than a HOST_WIDE_INT.
11161 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11162 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11163 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11164 final_offset from HOST_WIDE_INT to poly_int64.
11165 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11166 to_constant when getting the number of units in an Advanced SIMD
11168 (aarch64_builtin_vectorized_function): Check for a constant number
11170 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11172 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11173 attribute instead of GET_MODE_NUNITS.
11174 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11175 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11176 GET_MODE_SIZE for fixed-size registers.
11177 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11178 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11179 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11180 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11181 (aarch64_print_operand, aarch64_print_address_internal)
11182 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11183 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11184 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11185 Handle polynomial GET_MODE_SIZE.
11186 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11187 wider than SImode without modification.
11188 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11189 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11190 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11191 passing and returning SVE modes.
11192 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11193 rather than GEN_INT.
11194 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11195 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11196 (aarch64_allocate_and_probe_stack_space): Likewise.
11197 (aarch64_layout_frame): Cope with polynomial offsets.
11198 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11199 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11200 polynomial offsets.
11201 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11202 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11203 poly_int64 rather than a HOST_WIDE_INT.
11204 (aarch64_get_separate_components, aarch64_process_components)
11205 (aarch64_expand_prologue, aarch64_expand_epilogue)
11206 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11207 (aarch64_anchor_offset): New function, split out from...
11208 (aarch64_legitimize_address): ...here.
11209 (aarch64_builtin_vectorization_cost): Handle polynomial
11210 TYPE_VECTOR_SUBPARTS.
11211 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11213 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11214 number of elements from the PARALLEL rather than the mode.
11215 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11216 rather than GET_MODE_BITSIZE.
11217 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11218 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11219 (aarch64_expand_vec_perm_const_1): Handle polynomial
11220 d->perm.length () and d->perm elements.
11221 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11222 Apply to_constant to d->perm elements.
11223 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11224 polynomial CONST_VECTOR_NUNITS.
11225 (aarch64_move_pointer): Take amount as a poly_int64 rather
11227 (aarch64_progress_pointer): Avoid temporary variable.
11228 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11229 the mode attribute instead of GET_MODE.
11231 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11232 Alan Hayward <alan.hayward@arm.com>
11233 David Sherwood <david.sherwood@arm.com>
11235 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11236 x exists before using it.
11237 (aarch64_add_constant_internal): Rename to...
11238 (aarch64_add_offset_1): ...this. Replace regnum with separate
11239 src and dest rtxes. Handle the case in which they're different,
11240 including when the offset is zero. Replace scratchreg with an rtx.
11241 Use 2 additions if there is no spare register into which we can
11242 move a 16-bit constant.
11243 (aarch64_add_constant): Delete.
11244 (aarch64_add_offset): Replace reg with separate src and dest
11245 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11246 Use aarch64_add_offset_1.
11247 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11248 an rtx rather than an int. Take the delta as a poly_int64
11249 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11250 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11251 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11252 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11253 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11254 and aarch64_add_sp.
11255 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11256 aarch64_add_constant.
11258 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11260 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11261 Use scalar_float_mode.
11263 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11265 * config/aarch64/aarch64-simd.md
11266 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11267 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11268 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11269 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11270 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11271 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11272 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11273 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11274 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11275 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11277 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11280 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11281 targ_options->x_arm_arch_string is non NULL.
11283 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11285 * config/aarch64/aarch64.h
11286 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11288 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11291 * expmed.c (emit_store_flag_force): Swap if const op0
11292 and change VOIDmode to mode of op0.
11294 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11296 PR rtl-optimization/83761
11297 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11298 than bytes to mode_for_size.
11300 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11302 PR middle-end/83189
11303 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11304 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11307 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11309 PR middle-end/83575
11310 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11311 when in layout mode.
11312 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11313 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11316 2018-01-10 Michael Collison <michael.collison@arm.com>
11318 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11319 * config/aarch64/aarch64-option-extension.def: Add
11320 AARCH64_OPT_EXTENSION of 'fp16fml'.
11321 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11322 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11323 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11324 * config/aarch64/constraints.md (Ui7): New constraint.
11325 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11326 (VFMLA_SEL_W): Ditto.
11329 (VFMLA16_LOW): New int iterator.
11330 (VFMLA16_HIGH): Ditto.
11331 (UNSPEC_FMLAL): New unspec.
11332 (UNSPEC_FMLSL): Ditto.
11333 (UNSPEC_FMLAL2): Ditto.
11334 (UNSPEC_FMLSL2): Ditto.
11335 (f16mac): New code attribute.
11336 * config/aarch64/aarch64-simd-builtins.def
11337 (aarch64_fmlal_lowv2sf): Ditto.
11338 (aarch64_fmlsl_lowv2sf): Ditto.
11339 (aarch64_fmlalq_lowv4sf): Ditto.
11340 (aarch64_fmlslq_lowv4sf): Ditto.
11341 (aarch64_fmlal_highv2sf): Ditto.
11342 (aarch64_fmlsl_highv2sf): Ditto.
11343 (aarch64_fmlalq_highv4sf): Ditto.
11344 (aarch64_fmlslq_highv4sf): Ditto.
11345 (aarch64_fmlal_lane_lowv2sf): Ditto.
11346 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11347 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11348 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11349 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11350 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11351 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11352 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11353 (aarch64_fmlal_lane_highv2sf): Ditto.
11354 (aarch64_fmlsl_lane_highv2sf): Ditto.
11355 (aarch64_fmlal_laneq_highv2sf): Ditto.
11356 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11357 (aarch64_fmlalq_lane_highv4sf): Ditto.
11358 (aarch64_fmlsl_lane_highv4sf): Ditto.
11359 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11360 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11361 * config/aarch64/aarch64-simd.md:
11362 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11363 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11364 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11365 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11366 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11367 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11368 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11369 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11370 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11371 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11372 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11373 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11374 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11375 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11376 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11377 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11378 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11379 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11380 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11381 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11382 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11383 (vfmlsl_low_u32): Ditto.
11384 (vfmlalq_low_u32): Ditto.
11385 (vfmlslq_low_u32): Ditto.
11386 (vfmlal_high_u32): Ditto.
11387 (vfmlsl_high_u32): Ditto.
11388 (vfmlalq_high_u32): Ditto.
11389 (vfmlslq_high_u32): Ditto.
11390 (vfmlal_lane_low_u32): Ditto.
11391 (vfmlsl_lane_low_u32): Ditto.
11392 (vfmlal_laneq_low_u32): Ditto.
11393 (vfmlsl_laneq_low_u32): Ditto.
11394 (vfmlalq_lane_low_u32): Ditto.
11395 (vfmlslq_lane_low_u32): Ditto.
11396 (vfmlalq_laneq_low_u32): Ditto.
11397 (vfmlslq_laneq_low_u32): Ditto.
11398 (vfmlal_lane_high_u32): Ditto.
11399 (vfmlsl_lane_high_u32): Ditto.
11400 (vfmlal_laneq_high_u32): Ditto.
11401 (vfmlsl_laneq_high_u32): Ditto.
11402 (vfmlalq_lane_high_u32): Ditto.
11403 (vfmlslq_lane_high_u32): Ditto.
11404 (vfmlalq_laneq_high_u32): Ditto.
11405 (vfmlslq_laneq_high_u32): Ditto.
11406 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11407 (AARCH64_FL_FOR_ARCH8_4): New.
11408 (AARCH64_ISA_F16FML): New ISA flag.
11409 (TARGET_F16FML): New feature flag for fp16fml.
11410 (doc/invoke.texi): Document new fp16fml option.
11412 2018-01-10 Michael Collison <michael.collison@arm.com>
11414 * config/aarch64/aarch64-builtins.c:
11415 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11416 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11417 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11418 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11419 (AARCH64_ISA_SHA3): New ISA flag.
11420 (TARGET_SHA3): New feature flag for sha3.
11421 * config/aarch64/iterators.md (sha512_op): New int attribute.
11422 (CRYPTO_SHA512): New int iterator.
11423 (UNSPEC_SHA512H): New unspec.
11424 (UNSPEC_SHA512H2): Ditto.
11425 (UNSPEC_SHA512SU0): Ditto.
11426 (UNSPEC_SHA512SU1): Ditto.
11427 * config/aarch64/aarch64-simd-builtins.def
11428 (aarch64_crypto_sha512hqv2di): New builtin.
11429 (aarch64_crypto_sha512h2qv2di): Ditto.
11430 (aarch64_crypto_sha512su0qv2di): Ditto.
11431 (aarch64_crypto_sha512su1qv2di): Ditto.
11432 (aarch64_eor3qv8hi): Ditto.
11433 (aarch64_rax1qv2di): Ditto.
11434 (aarch64_xarqv2di): Ditto.
11435 (aarch64_bcaxqv8hi): Ditto.
11436 * config/aarch64/aarch64-simd.md:
11437 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11438 (aarch64_crypto_sha512su0qv2di): Ditto.
11439 (aarch64_crypto_sha512su1qv2di): Ditto.
11440 (aarch64_eor3qv8hi): Ditto.
11441 (aarch64_rax1qv2di): Ditto.
11442 (aarch64_xarqv2di): Ditto.
11443 (aarch64_bcaxqv8hi): Ditto.
11444 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11445 (vsha512h2q_u64): Ditto.
11446 (vsha512su0q_u64): Ditto.
11447 (vsha512su1q_u64): Ditto.
11448 (veor3q_u16): Ditto.
11449 (vrax1q_u64): Ditto.
11450 (vxarq_u64): Ditto.
11451 (vbcaxq_u16): Ditto.
11452 * config/arm/types.md (crypto_sha512): New type attribute.
11453 (crypto_sha3): Ditto.
11454 (doc/invoke.texi): Document new sha3 option.
11456 2018-01-10 Michael Collison <michael.collison@arm.com>
11458 * config/aarch64/aarch64-builtins.c:
11459 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11460 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11461 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11462 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11463 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11464 (AARCH64_ISA_SM4): New ISA flag.
11465 (TARGET_SM4): New feature flag for sm4.
11466 * config/aarch64/aarch64-simd-builtins.def
11467 (aarch64_sm3ss1qv4si): Ditto.
11468 (aarch64_sm3tt1aq4si): Ditto.
11469 (aarch64_sm3tt1bq4si): Ditto.
11470 (aarch64_sm3tt2aq4si): Ditto.
11471 (aarch64_sm3tt2bq4si): Ditto.
11472 (aarch64_sm3partw1qv4si): Ditto.
11473 (aarch64_sm3partw2qv4si): Ditto.
11474 (aarch64_sm4eqv4si): Ditto.
11475 (aarch64_sm4ekeyqv4si): Ditto.
11476 * config/aarch64/aarch64-simd.md:
11477 (aarch64_sm3ss1qv4si): Ditto.
11478 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11479 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11480 (aarch64_sm4eqv4si): Ditto.
11481 (aarch64_sm4ekeyqv4si): Ditto.
11482 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11483 (sm3part_op): Ditto.
11484 (CRYPTO_SM3TT): Ditto.
11485 (CRYPTO_SM3PART): Ditto.
11486 (UNSPEC_SM3SS1): New unspec.
11487 (UNSPEC_SM3TT1A): Ditto.
11488 (UNSPEC_SM3TT1B): Ditto.
11489 (UNSPEC_SM3TT2A): Ditto.
11490 (UNSPEC_SM3TT2B): Ditto.
11491 (UNSPEC_SM3PARTW1): Ditto.
11492 (UNSPEC_SM3PARTW2): Ditto.
11493 (UNSPEC_SM4E): Ditto.
11494 (UNSPEC_SM4EKEY): Ditto.
11495 * config/aarch64/constraints.md (Ui2): New constraint.
11496 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11497 * config/arm/types.md (crypto_sm3): New type attribute.
11498 (crypto_sm4): Ditto.
11499 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11500 (vsm3tt1aq_u32): Ditto.
11501 (vsm3tt1bq_u32): Ditto.
11502 (vsm3tt2aq_u32): Ditto.
11503 (vsm3tt2bq_u32): Ditto.
11504 (vsm3partw1q_u32): Ditto.
11505 (vsm3partw2q_u32): Ditto.
11506 (vsm4eq_u32): Ditto.
11507 (vsm4ekeyq_u32): Ditto.
11508 (doc/invoke.texi): Document new sm4 option.
11510 2018-01-10 Michael Collison <michael.collison@arm.com>
11512 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11513 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11514 (AARCH64_FL_FOR_ARCH8_4): New.
11515 (AARCH64_FL_V8_4): New flag.
11516 (doc/invoke.texi): Document new armv8.4-a option.
11518 2018-01-10 Michael Collison <michael.collison@arm.com>
11520 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11521 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11522 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11523 * config/aarch64/aarch64-option-extension.def: Add
11524 AARCH64_OPT_EXTENSION of 'sha2'.
11525 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11526 (crypto): Disable sha2 and aes if crypto disabled.
11527 (crypto): Enable aes and sha2 if enabled.
11528 (simd): Disable sha2 and aes if simd disabled.
11529 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11531 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11532 (TARGET_SHA2): New feature flag for sha2.
11533 (TARGET_AES): New feature flag for aes.
11534 * config/aarch64/aarch64-simd.md:
11535 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11536 conditional on TARGET_AES.
11537 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11538 (aarch64_crypto_sha1hsi): Make pattern conditional
11540 (aarch64_crypto_sha1hv4si): Ditto.
11541 (aarch64_be_crypto_sha1hv4si): Ditto.
11542 (aarch64_crypto_sha1su1v4si): Ditto.
11543 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11544 (aarch64_crypto_sha1su0v4si): Ditto.
11545 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11546 (aarch64_crypto_sha256su0v4si): Ditto.
11547 (aarch64_crypto_sha256su1v4si): Ditto.
11548 (doc/invoke.texi): Document new aes and sha2 options.
11550 2018-01-10 Martin Sebor <msebor@redhat.com>
11552 PR tree-optimization/83781
11553 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11556 2018-01-11 Martin Sebor <msebor@gmail.com>
11557 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11559 PR tree-optimization/83501
11560 PR tree-optimization/81703
11562 * tree-ssa-strlen.c (get_string_cst): Rename...
11563 (get_string_len): ...to this. Handle global constants.
11564 (handle_char_store): Adjust.
11566 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11567 Jim Wilson <jimw@sifive.com>
11569 * config/riscv/riscv-protos.h (riscv_output_return): New.
11570 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11571 (riscv_attribute_table, riscv_output_return),
11572 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11573 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11574 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11576 (riscv_expand_prologue): Add early return for naked function.
11577 (riscv_expand_epilogue): Likewise.
11578 (riscv_function_ok_for_sibcall): Return false for naked function.
11579 (riscv_set_current_function): New.
11580 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11581 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11582 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11583 * doc/extend.texi (RISC-V Function Attributes): New.
11585 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11587 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11588 check for 128-bit long double before checking TCmode.
11589 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11590 128-bit long doubles before checking TFmode or TCmode.
11591 (FLOAT128_IBM_P): Likewise.
11593 2018-01-10 Martin Sebor <msebor@redhat.com>
11595 PR tree-optimization/83671
11596 * builtins.c (c_strlen): Unconditionally return zero for the empty
11598 Use -Warray-bounds for warnings.
11599 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11600 for non-constant array indices with COMPONENT_REF, arrays of
11601 arrays, and pointers to arrays.
11602 (gimple_fold_builtin_strlen): Determine and set length range for
11603 non-constant character arrays.
11605 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11607 PR middle-end/81897
11608 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11611 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11613 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11615 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11618 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11619 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11620 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11621 indexed_or_indirect_operand predicate.
11622 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11623 (*vsx_le_perm_load_v8hi): Likewise.
11624 (*vsx_le_perm_load_v16qi): Likewise.
11625 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11626 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11627 (*vsx_le_perm_store_v8hi): Likewise.
11628 (*vsx_le_perm_store_v16qi): Likewise.
11629 (eight unnamed splitters): Likewise.
11631 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11633 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11634 * config/rs6000/emmintrin.h: Likewise.
11635 * config/rs6000/mmintrin.h: Likewise.
11636 * config/rs6000/xmmintrin.h: Likewise.
11638 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11641 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11643 * tree.c (tree_nop_conversion): Return true for location wrapper
11645 (maybe_wrap_with_location): New function.
11646 (selftest::check_strip_nops): New function.
11647 (selftest::test_location_wrappers): New function.
11648 (selftest::tree_c_tests): Call it.
11649 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11650 (maybe_wrap_with_location): New decl.
11651 (EXPR_LOCATION_WRAPPER_P): New macro.
11652 (location_wrapper_p): New inline function.
11653 (tree_strip_any_location_wrapper): New inline function.
11655 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11658 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11659 stack_realign_offset for the largest alignment of stack slot
11661 (ix86_find_max_used_stack_alignment): New function.
11662 (ix86_finalize_stack_frame_flags): Use it. Set
11663 max_used_stack_alignment if we don't realign stack.
11664 * config/i386/i386.h (machine_function): Add
11665 max_used_stack_alignment.
11667 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11669 * config/arm/arm.opt (-mbranch-cost): New option.
11670 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11673 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11676 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11677 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11679 2018-01-10 Richard Biener <rguenther@suse.de>
11682 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11683 early out so it also covers the case where we have a non-NULL
11686 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11688 PR tree-optimization/83753
11689 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11690 for non-strided grouped accesses if the number of elements is 1.
11692 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11695 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11696 * i386.h (TARGET_USE_GATHER): Define.
11697 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11699 2018-01-10 Martin Liska <mliska@suse.cz>
11702 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11703 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11705 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11706 CLEANUP_NO_PARTITIONING is not set.
11708 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11710 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11711 for vectors, as a partial revert of r254296.
11712 * rtl.h (const_vec_p): Delete.
11713 (const_vec_duplicate_p): Don't test for vector CONSTs.
11714 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11715 * expmed.c (make_tree): Likewise.
11718 * common.md (E, F): Use CONSTANT_P instead of checking for
11720 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11721 checking for CONST_VECTOR.
11723 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11725 PR middle-end/83575
11726 * predict.c (force_edge_cold): Handle in more sane way edges
11727 with no prediction.
11729 2018-01-09 Carl Love <cel@us.ibm.com>
11731 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11733 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11734 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11735 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11736 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11737 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11738 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11739 * config/rs6000/rs6000-protos.h: Add extern defition for
11740 rs6000_generate_float2_double_code.
11741 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11743 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11744 (float2_v2df): Add define_expand.
11746 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11749 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11750 op_mode in the force_to_mode call.
11752 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11754 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11755 instead of checking each element individually.
11756 (aarch64_evpc_uzp): Likewise.
11757 (aarch64_evpc_zip): Likewise.
11758 (aarch64_evpc_ext): Likewise.
11759 (aarch64_evpc_rev): Likewise.
11760 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11761 instead of checking each element individually. Return true without
11763 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11764 whether all selected elements come from the same input, instead of
11765 checking each element individually. Remove calls to gen_rtx_REG,
11766 start_sequence and end_sequence and instead assert that no rtl is
11769 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11771 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11772 order of HIGH and CONST checks.
11774 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11776 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11777 if the destination isn't an SSA_NAME.
11779 2018-01-09 Richard Biener <rguenther@suse.de>
11781 PR tree-optimization/83668
11782 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11784 (canonicalize_loop_form): ... here, renamed from ...
11785 (canonicalize_loop_closed_ssa_form): ... this and amended to
11786 swap successor edges for loop exit blocks to make us use
11787 the RPO order we need for initial schedule generation.
11789 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11791 PR tree-optimization/64811
11792 * match.pd: When optimizing comparisons with Inf, avoid
11793 introducing or losing exceptions from comparisons with NaN.
11795 2018-01-09 Martin Liska <mliska@suse.cz>
11798 * asan.c (shadow_mem_size): Add gcc_assert.
11800 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11802 Don't save registers in main().
11805 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11806 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11807 * config/avr/avr.c (avr_set_current_function): Don't error if
11808 naked, OS_task or OS_main are specified at the same time.
11809 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11811 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11813 * common/config/avr/avr-common.c (avr_option_optimization_table):
11814 Switch on -mmain-is-OS_task for optimizing compilations.
11816 2018-01-09 Richard Biener <rguenther@suse.de>
11818 PR tree-optimization/83572
11819 * graphite.c: Include cfganal.h.
11820 (graphite_transform_loops): Connect infinite loops to exit
11821 and remove fake edges at the end.
11823 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11825 * ipa-inline.c (edge_badness): Revert accidental checkin.
11827 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11830 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11831 symbols; not inline clones.
11833 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11836 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11837 hard registers. Formatting fixes.
11839 PR preprocessor/83722
11840 * gcc.c (try_generate_repro): Pass
11841 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11842 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11845 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11846 Kito Cheng <kito.cheng@gmail.com>
11848 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11849 (riscv_leaf_function_p): Delete.
11850 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11852 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11854 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11856 (do_ifelse): New function.
11857 (do_isel): New function.
11858 (do_sub3): New function.
11859 (do_add3): New function.
11860 (do_load_mask_compare): New function.
11861 (do_overlap_load_compare): New function.
11862 (expand_compare_loop): New function.
11863 (expand_block_compare): Call expand_compare_loop() when appropriate.
11864 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11865 option description.
11866 (-mblock-compare-inline-loop-limit): New option.
11868 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11871 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11872 Reverse order of second and third operands in first alternative.
11873 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11874 of first and second elements in UNSPEC_VPERMR vector.
11875 (altivec_expand_vec_perm_le): Likewise.
11877 2018-01-08 Jeff Law <law@redhat.com>
11879 PR rtl-optimizatin/81308
11880 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11881 (process_switch): If group_case_labels makes a change, then set
11883 (pass_convert_switch::execute): If a switch is converted, then
11884 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11886 PR rtl-optimization/81308
11887 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11890 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11892 PR target/83663 - Revert r255946
11893 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11894 generation for cases where splatting a value is not useful.
11895 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11896 across a vec_duplicate and a paradoxical subreg forming a vector
11897 mode to a vec_concat.
11899 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11901 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11902 -march=armv8.3-a variants.
11903 * config/arm/t-multilib: Likewise.
11904 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11906 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11908 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11910 (cceq_ior_compare_complement): Give it a name so I can use it, and
11911 change boolean_or_operator predicate to boolean_operator so it can
11912 be used to generate a crand.
11913 (eqne): New code iterator.
11914 (bd/bd_neg): New code_attrs.
11915 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11916 a single define_insn.
11917 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11918 decrement (bdnzt/bdnzf/bdzt/bdzf).
11919 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11920 with the new names of the branch decrement patterns, and added the
11921 names of the branch decrement conditional patterns.
11923 2018-01-08 Richard Biener <rguenther@suse.de>
11925 PR tree-optimization/83563
11926 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11929 2018-01-08 Richard Biener <rguenther@suse.de>
11931 PR middle-end/83713
11932 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11934 2018-01-08 Richard Biener <rguenther@suse.de>
11936 PR tree-optimization/83685
11937 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11938 references to abnormals.
11940 2018-01-08 Richard Biener <rguenther@suse.de>
11943 * dwarf2out.c (output_indirect_strings): Handle empty
11944 skeleton_debug_str_hash.
11945 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11947 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11949 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11950 (emit_store_direct): Likewise.
11951 (arc_trampoline_adjust_address): Likewise.
11952 (arc_asm_trampoline_template): New function.
11953 (arc_initialize_trampoline): Use asm_trampoline_template.
11954 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11955 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11956 * config/arc/arc.md (flush_icache): Delete pattern.
11958 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11960 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11961 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11964 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11967 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11968 by not USED_FOR_TARGET.
11969 (make_pass_resolve_sw_modes): Likewise.
11971 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11973 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11976 2018-01-08 Richard Biener <rguenther@suse.de>
11978 PR middle-end/83580
11979 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11981 2018-01-08 Richard Biener <rguenther@suse.de>
11983 PR middle-end/83517
11984 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11986 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11988 PR middle-end/81897
11989 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11990 basic blocks with a small number of successors.
11991 (convert_control_dep_chain_into_preds): Improve handling of
11993 (dump_predicates): Split apart into...
11994 (dump_pred_chain): ...here...
11995 (dump_pred_info): ...and here.
11996 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11997 (can_chain_union_be_invalidated_p): Improve check for invalidation
11999 (uninit_uses_cannot_happen): Avoid unnecessary if
12000 convert_control_dep_chain_into_preds yielded nothing.
12002 2018-01-06 Martin Sebor <msebor@redhat.com>
12004 PR tree-optimization/83640
12005 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12006 subtracting negative offset from size.
12007 (builtin_access::overlap): Adjust offset bounds of the access to fall
12008 within the size of the object if possible.
12010 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12012 PR rtl-optimization/83699
12013 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12014 extract_bit_field_as_subreg to cases in which the extracted
12015 value is also a vector.
12017 * lra-constraints.c (process_alt_operands): Test for the equivalence
12018 substitutions when detecting a possible reload cycle.
12020 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12023 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12024 by default if flag_selective_schedling{,2}. Formatting fixes.
12026 PR rtl-optimization/83682
12027 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12028 if it has non-VECTOR_MODE element mode.
12029 (vec_duplicate_p): Likewise.
12031 PR middle-end/83694
12032 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12033 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12035 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12038 * config/i386/i386-builtin.def
12039 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12040 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12041 Require also OPTION_MASK_ISA_AVX512F in addition to
12042 OPTION_MASK_ISA_GFNI.
12043 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12044 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12045 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12046 to OPTION_MASK_ISA_GFNI.
12047 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12048 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12049 OPTION_MASK_ISA_AVX512BW.
12050 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12051 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12052 addition to OPTION_MASK_ISA_GFNI.
12053 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12054 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12055 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12056 to OPTION_MASK_ISA_GFNI.
12057 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12058 a requirement for all ISAs rather than any of them with a few
12060 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12062 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12063 bitmasks to be enabled with 3 exceptions, instead of requiring any
12064 enabled ISA with lots of exceptions.
12065 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12066 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12067 Change avx512bw in isa attribute to avx512f.
12068 * config/i386/sgxintrin.h: Add license boilerplate.
12069 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12070 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12071 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12072 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12074 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12075 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12076 temporarily sse2 rather than sse if not enabled already.
12079 * config/i386/sse.md (VI248_VLBW): Rename to ...
12080 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12081 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12082 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12083 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12084 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12085 mode iterator instead of VI248_VLBW.
12087 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12089 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12090 (record_modified): Skip clobbers; add debug output.
12091 (param_change_prob): Use sreal frequencies.
12093 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12095 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12096 punt for user-aligned variables.
12098 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12100 * tree-chrec.c (chrec_contains_symbols): Return true for
12103 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12106 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12107 of (x|y) == x for BICS pattern.
12109 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12111 PR tree-optimization/83605
12112 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12113 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12116 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12118 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12119 * config/epiphany/rtems.h: New file.
12121 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12122 Uros Bizjak <ubizjak@gmail.com>
12125 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12126 QIreg_operand instead of register_operand predicate.
12127 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12128 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12129 comments instead of -fmitigate[-_]rop.
12131 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12134 * cgraphunit.c (symbol_table::compile): Switch to text_section
12135 before calling assembly_start debug hook.
12136 * run-rtl-passes.c (run_rtl_passes): Likewise.
12139 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12141 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12142 range_int_cst_p rather than !symbolic_range_p before calling
12143 extract_range_from_multiplicative_op_1.
12145 2018-01-04 Jeff Law <law@redhat.com>
12147 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12148 redundant test in assertion.
12150 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12152 * doc/rtl.texi: Document machine_mode wrapper classes.
12154 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12156 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12157 using tree_to_uhwi.
12159 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12161 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12162 the VEC_PERM_EXPR fold to fail.
12164 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12167 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12168 to switched_sections.
12170 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12173 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12174 test for d.testing.
12176 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12179 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12180 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12182 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12185 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12186 is BLKmode and bitpos not zero or mode change is needed.
12188 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12191 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12194 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12197 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12198 instead of MULT rtx. Update all corresponding splitters.
12199 (*saddl_se): Ditto.
12200 (*ssub<modesuffix>): Ditto.
12201 (*ssubl_se): Ditto.
12202 (*cmp_sadd_di): Update split patterns.
12203 (*cmp_sadd_si): Ditto.
12204 (*cmp_sadd_sidi): Ditto.
12205 (*cmp_ssub_di): Ditto.
12206 (*cmp_ssub_si): Ditto.
12207 (*cmp_ssub_sidi): Ditto.
12208 * config/alpha/predicates.md (const23_operand): New predicate.
12209 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12210 Look for ASHIFT, not MULT inner operand.
12211 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12213 2018-01-04 Martin Liska <mliska@suse.cz>
12215 PR gcov-profile/83669
12216 * gcov.c (output_intermediate_file): Add version to intermediate
12218 * doc/gcov.texi: Document new field 'version' in intermediate
12219 file format. Fix location of '-k' option of gcov command.
12221 2018-01-04 Martin Liska <mliska@suse.cz>
12224 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12226 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12228 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12230 2018-01-03 Martin Sebor <msebor@redhat.com>
12232 PR tree-optimization/83655
12233 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12234 checking calls with invalid arguments.
12236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12238 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12239 (vectorizable_mask_load_store): Delete.
12240 (vectorizable_call): Return false for masked loads and stores.
12241 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12242 instead of gimple_assign_rhs1.
12243 (vectorizable_load): Handle IFN_MASK_LOAD.
12244 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12248 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12250 (vectorizable_mask_load_store): ...here.
12251 (vectorizable_load): ...and here.
12253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12255 * tree-vect-stmts.c (vect_build_all_ones_mask)
12256 (vect_build_zero_merge_argument): New functions, split out from...
12257 (vectorizable_load): ...here.
12259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12261 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12263 (vectorizable_mask_load_store): ...here.
12264 (vectorizable_store): ...and here.
12266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12268 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12270 (vectorizable_mask_load_store): ...here.
12272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12274 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12275 (vect_model_store_cost): Take a vec_load_store_type instead of a
12277 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12278 (vect_model_store_cost): Take a vec_load_store_type instead of a
12280 (vectorizable_mask_load_store): Update accordingly.
12281 (vectorizable_store): Likewise.
12282 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12286 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12287 IFN_MASK_LOAD calls here rather than...
12288 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12291 Alan Hayward <alan.hayward@arm.com>
12292 David Sherwood <david.sherwood@arm.com>
12294 * expmed.c (extract_bit_field_1): For vector extracts,
12295 fall back to extract_bit_field_as_subreg if vec_extract
12298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12299 Alan Hayward <alan.hayward@arm.com>
12300 David Sherwood <david.sherwood@arm.com>
12302 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12303 they are variable or constant sized.
12304 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12305 slots for constant-sized data.
12307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12308 Alan Hayward <alan.hayward@arm.com>
12309 David Sherwood <david.sherwood@arm.com>
12311 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12312 handling COND_EXPRs with boolean comparisons, try to find a better
12313 basis for the mask type than the boolean itself.
12315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12317 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12318 is calculated and how it can be overridden.
12319 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12320 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12322 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12326 Alan Hayward <alan.hayward@arm.com>
12327 David Sherwood <david.sherwood@arm.com>
12329 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12330 Remove the mode argument.
12331 (aarch64_simd_valid_immediate): Remove the mode and inverse
12333 * config/aarch64/iterators.md (bitsize): New iterator.
12334 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12335 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12336 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12337 aarch64_simd_valid_immediate.
12338 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12339 (aarch64_reg_or_bic_imm): Likewise.
12340 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12341 with an insn_type enum and msl with a modifier_type enum.
12342 Replace element_width with a scalar_mode. Change the shift
12343 to unsigned int. Add constructors for scalar_float_mode and
12344 scalar_int_mode elements.
12345 (aarch64_vect_float_const_representable_p): Delete.
12346 (aarch64_can_const_movi_rtx_p)
12347 (aarch64_simd_scalar_immediate_valid_for_move)
12348 (aarch64_simd_make_constant): Update call to
12349 aarch64_simd_valid_immediate.
12350 (aarch64_advsimd_valid_immediate_hs): New function.
12351 (aarch64_advsimd_valid_immediate): Likewise.
12352 (aarch64_simd_valid_immediate): Remove mode and inverse
12353 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12354 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12355 and aarch64_float_const_representable_p on the result.
12356 (aarch64_output_simd_mov_immediate): Remove mode argument.
12357 Update call to aarch64_simd_valid_immediate and use of
12358 simd_immediate_info.
12359 (aarch64_output_scalar_simd_mov_immediate): Update call
12362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12363 Alan Hayward <alan.hayward@arm.com>
12364 David Sherwood <david.sherwood@arm.com>
12366 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12367 (mode_nunits): Likewise CONST_MODE_NUNITS.
12368 * machmode.def (ADJUST_NUNITS): Document.
12369 * genmodes.c (mode_data::need_nunits_adj): New field.
12370 (blank_mode): Update accordingly.
12371 (adj_nunits): New variable.
12372 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12374 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12375 listed in adj_nunits.
12376 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12377 listed in adj_nunits. Don't emit case statements for such modes.
12378 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12379 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12380 nothing if adj_nunits is nonnull.
12381 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12382 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12383 (emit_mode_fbit): Update use of print_maybe_const_decl.
12384 (emit_move_size): Likewise. Treat the array as non-const
12386 (emit_mode_adjustments): Handle adj_nunits.
12388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12390 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12391 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12392 (VECTOR_MODES): Use it.
12393 (make_vector_modes): Take the prefix as an argument.
12395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12396 Alan Hayward <alan.hayward@arm.com>
12397 David Sherwood <david.sherwood@arm.com>
12399 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12400 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12401 for MODE_VECTOR_BOOL.
12402 * machmode.def (VECTOR_BOOL_MODE): Document.
12403 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12404 (make_vector_bool_mode): New function.
12405 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12407 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12408 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12410 * stor-layout.c (int_mode_for_mode): Likewise.
12411 * tree.c (build_vector_type_for_mode): Likewise.
12412 * varasm.c (output_constant_pool_2): Likewise.
12413 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12414 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12415 for MODE_VECTOR_BOOL.
12416 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12417 of mode class checks.
12418 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12419 instead of a list of mode class checks.
12420 (expand_vector_scalar_condition): Likewise.
12421 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12424 Alan Hayward <alan.hayward@arm.com>
12425 David Sherwood <david.sherwood@arm.com>
12427 * machmode.h (mode_size): Change from unsigned short to
12429 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12430 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12431 or if measurement_type is not polynomial.
12432 (fixed_size_mode::includes_p): Check for constant-sized modes.
12433 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12434 return a poly_uint16 rather than an unsigned short.
12435 (emit_mode_size): Change the type of mode_size from unsigned short
12436 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12437 (emit_mode_adjustments): Cope with polynomial vector sizes.
12438 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12440 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12442 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12443 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12444 * caller-save.c (setup_save_areas): Likewise.
12445 (replace_reg_with_saved_mem): Likewise.
12446 * calls.c (emit_library_call_value_1): Likewise.
12447 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12448 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12449 (gen_lowpart_for_combine): Likewise.
12450 * convert.c (convert_to_integer_1): Likewise.
12451 * cse.c (equiv_constant, cse_insn): Likewise.
12452 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12453 (cselib_subst_to_values): Likewise.
12454 * dce.c (word_dce_process_block): Likewise.
12455 * df-problems.c (df_word_lr_mark_ref): Likewise.
12456 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12457 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12458 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12459 (rtl_for_decl_location): Likewise.
12460 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12461 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12462 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12463 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12464 (expand_expr_real_1): Likewise.
12465 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12466 (pad_below): Likewise.
12467 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12468 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12469 * ira.c (get_subreg_tracking_sizes): Likewise.
12470 * ira-build.c (ira_create_allocno_objects): Likewise.
12471 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12472 (ira_sort_regnos_for_alter_reg): Likewise.
12473 * ira-costs.c (record_operand_costs): Likewise.
12474 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12475 (resolve_simple_move): Likewise.
12476 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12477 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12478 (lra_constraints): Likewise.
12479 (CONST_POOL_OK_P): Reject variable-sized modes.
12480 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12481 (add_pseudo_to_slot, lra_spill): Likewise.
12482 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12483 * optabs-query.c (get_best_extraction_insn): Likewise.
12484 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12485 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12486 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12487 * recog.c (offsettable_address_addr_space_p): Likewise.
12488 * regcprop.c (maybe_mode_change): Likewise.
12489 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12490 * regrename.c (build_def_use): Likewise.
12491 * regstat.c (dump_reg_info): Likewise.
12492 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12493 (find_reloads, find_reloads_subreg_address): Likewise.
12494 * reload1.c (eliminate_regs_1): Likewise.
12495 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12496 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12497 (simplify_binary_operation_1, simplify_subreg): Likewise.
12498 * targhooks.c (default_function_arg_padding): Likewise.
12499 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12500 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12501 (verify_gimple_assign_ternary): Likewise.
12502 * tree-inline.c (estimate_move_cost): Likewise.
12503 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12504 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12505 (get_address_cost_ainc): Likewise.
12506 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12507 (vect_supportable_dr_alignment): Likewise.
12508 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12509 (vectorizable_reduction): Likewise.
12510 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12511 (vectorizable_operation, vectorizable_load): Likewise.
12512 * tree.c (build_same_sized_truth_vector_type): Likewise.
12513 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12514 * var-tracking.c (emit_note_insn_var_location): Likewise.
12515 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12516 (ADDR_VEC_ALIGN): Likewise.
12518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12519 Alan Hayward <alan.hayward@arm.com>
12520 David Sherwood <david.sherwood@arm.com>
12522 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12524 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12525 or if measurement_type is polynomial.
12526 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12527 * combine.c (make_extraction): Likewise.
12528 * dse.c (find_shift_sequence): Likewise.
12529 * dwarf2out.c (mem_loc_descriptor): Likewise.
12530 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12531 (extract_bit_field, extract_low_bits): Likewise.
12532 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12533 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12534 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12535 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12536 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12537 * reload.c (find_reloads): Likewise.
12538 * reload1.c (alter_reg): Likewise.
12539 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12540 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12541 * tree-if-conv.c (predicate_mem_writes): Likewise.
12542 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12543 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12544 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12545 * valtrack.c (dead_debug_insert_temp): Likewise.
12546 * varasm.c (mergeable_constant_section): Likewise.
12547 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12550 Alan Hayward <alan.hayward@arm.com>
12551 David Sherwood <david.sherwood@arm.com>
12553 * expr.c (expand_assignment): Cope with polynomial mode sizes
12554 when assigning to a CONCAT.
12556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12557 Alan Hayward <alan.hayward@arm.com>
12558 David Sherwood <david.sherwood@arm.com>
12560 * machmode.h (mode_precision): Change from unsigned short to
12562 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12564 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12565 or if measurement_type is not polynomial.
12566 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12567 in which the mode is already known to be a scalar_int_mode.
12568 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12569 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12571 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12572 for GET_MODE_PRECISION.
12573 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12574 for GET_MODE_PRECISION.
12575 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12577 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12578 (expand_field_assignment, make_extraction): Likewise.
12579 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12580 (get_last_value): Likewise.
12581 * convert.c (convert_to_integer_1): Likewise.
12582 * cse.c (cse_insn): Likewise.
12583 * expr.c (expand_expr_real_1): Likewise.
12584 * lra-constraints.c (simplify_operand_subreg): Likewise.
12585 * optabs-query.c (can_atomic_load_p): Likewise.
12586 * optabs.c (expand_atomic_load): Likewise.
12587 (expand_atomic_store): Likewise.
12588 * ree.c (combine_reaching_defs): Likewise.
12589 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12590 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12591 * tree.h (type_has_mode_precision_p): Likewise.
12592 * ubsan.c (instrument_si_overflow): Likewise.
12594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12595 Alan Hayward <alan.hayward@arm.com>
12596 David Sherwood <david.sherwood@arm.com>
12598 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12599 polynomial numbers of units.
12600 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12601 (valid_vector_subparts_p): New function.
12602 (build_vector_type): Remove temporary shim and take the number
12603 of units as a poly_uint64 rather than an int.
12604 (build_opaque_vector_type): Take the number of units as a
12605 poly_uint64 rather than an int.
12606 * tree.c (build_vector_from_ctor): Handle polynomial
12607 TYPE_VECTOR_SUBPARTS.
12608 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12609 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12610 (build_vector_from_val): If the number of units is variable,
12611 use build_vec_duplicate_cst for constant operands and
12612 VEC_DUPLICATE_EXPR otherwise.
12613 (make_vector_type): Remove temporary is_constant ().
12614 (build_vector_type, build_opaque_vector_type): Take the number of
12615 units as a poly_uint64 rather than an int.
12616 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12618 * cfgexpand.c (expand_debug_expr): Likewise.
12619 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12620 (store_constructor, expand_expr_real_1): Likewise.
12621 (const_scalar_mask_from_tree): Likewise.
12622 * fold-const-call.c (fold_const_reduction): Likewise.
12623 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12624 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12625 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12626 (fold_relational_const): Likewise.
12627 (native_interpret_vector): Likewise. Change the size from an
12628 int to an unsigned int.
12629 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12630 TYPE_VECTOR_SUBPARTS.
12631 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12632 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12633 duplicating a non-constant operand into a variable-length vector.
12634 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12635 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12636 * ipa-icf.c (sem_variable::equals): Likewise.
12637 * match.pd: Likewise.
12638 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12639 * print-tree.c (print_node): Likewise.
12640 * stor-layout.c (layout_type): Likewise.
12641 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12642 * tree-cfg.c (verify_gimple_comparison): Likewise.
12643 (verify_gimple_assign_binary): Likewise.
12644 (verify_gimple_assign_ternary): Likewise.
12645 (verify_gimple_assign_single): Likewise.
12646 * tree-pretty-print.c (dump_generic_node): Likewise.
12647 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12648 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12649 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12650 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12651 (vect_shift_permute_load_chain): Likewise.
12652 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12653 (expand_vector_condition, optimize_vector_constructor): Likewise.
12654 (lower_vec_perm, get_compute_type): Likewise.
12655 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12656 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12657 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12658 (vect_recog_mask_conversion_pattern): Likewise.
12659 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12660 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12661 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12662 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12663 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12664 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12665 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12666 (supportable_widening_operation): Likewise.
12667 (supportable_narrowing_operation): Likewise.
12668 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12670 * varasm.c (output_constant): 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-vect-data-refs.c (vect_permute_store_chain): Reorganize
12677 so that both the length == 3 and length != 3 cases set up their
12678 own permute vectors. Add comments explaining why we know the
12679 number of elements is constant.
12680 (vect_permute_load_chain): Likewise.
12682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12683 Alan Hayward <alan.hayward@arm.com>
12684 David Sherwood <david.sherwood@arm.com>
12686 * machmode.h (mode_nunits): Change from unsigned char to
12688 (ONLY_FIXED_SIZE_MODES): New macro.
12689 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12690 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12691 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12693 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12694 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12695 or if measurement_type is not polynomial.
12696 * genmodes.c (ZERO_COEFFS): New macro.
12697 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12699 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12700 Use ZERO_COEFFS when emitting initializers.
12701 * data-streamer.h (bp_pack_poly_value): New function.
12702 (bp_unpack_poly_value): Likewise.
12703 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12704 for GET_MODE_NUNITS.
12705 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12706 for GET_MODE_NUNITS.
12707 * tree.c (make_vector_type): Remove temporary shim and make
12708 the real function take the number of units as a poly_uint64
12709 rather than an int.
12710 (build_vector_type_for_mode): Handle polynomial nunits.
12711 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12712 * emit-rtl.c (const_vec_series_p_1): Likewise.
12713 (gen_rtx_CONST_VECTOR): Likewise.
12714 * fold-const.c (test_vec_duplicate_folding): Likewise.
12715 * genrecog.c (validate_pattern): Likewise.
12716 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12717 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12718 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12719 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12720 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12721 * rtlanal.c (subreg_get_info): Likewise.
12722 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12723 (vect_grouped_load_supported): Likewise.
12724 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12725 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12726 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12727 (simplify_const_unary_operation, simplify_binary_operation_1)
12728 (simplify_const_binary_operation, simplify_ternary_operation)
12729 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12730 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12731 instead of CONST_VECTOR_NUNITS.
12732 * varasm.c (output_constant_pool_2): Likewise.
12733 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12734 explicit-encoded elements in the XVEC for variable-length vectors.
12736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12738 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12741 Alan Hayward <alan.hayward@arm.com>
12742 David Sherwood <david.sherwood@arm.com>
12744 * coretypes.h (fixed_size_mode): Declare.
12745 (fixed_size_mode_pod): New typedef.
12746 * builtins.h (target_builtins::x_apply_args_mode)
12747 (target_builtins::x_apply_result_mode): Change type to
12748 fixed_size_mode_pod.
12749 * builtins.c (apply_args_size, apply_result_size, result_vector)
12750 (expand_builtin_apply_args_1, expand_builtin_apply)
12751 (expand_builtin_return): Update accordingly.
12753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12755 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12756 * cselib.c (cselib_hash_rtx): Likewise.
12757 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12758 CONST_VECTOR encoding.
12760 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12761 Jeff Law <law@redhat.com>
12764 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12765 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12766 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12767 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12770 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12771 explicitly probe *sp in a noreturn function if there were any callee
12772 register saves or frame pointer is needed.
12774 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12777 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12778 BLKmode for ternary, binary or unary expressions.
12781 * var-tracking.c (delete_vta_debug_insn): New inline function.
12782 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12783 insns from get_insns () to NULL instead of each bb separately.
12784 Use delete_vta_debug_insn. No longer static.
12785 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12786 delete_vta_debug_insns callers.
12787 * rtl.h (delete_vta_debug_insns): Declare.
12788 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12789 instead of variable_tracking_main.
12791 2018-01-03 Martin Sebor <msebor@redhat.com>
12793 PR tree-optimization/83603
12794 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12795 arguments past the endof the argument list in functions declared
12796 without a prototype.
12797 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12798 Avoid checking when arguments are null.
12800 2018-01-03 Martin Sebor <msebor@redhat.com>
12803 * doc/extend.texi (attribute const): Fix a typo.
12804 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12805 issuing -Wsuggest-attribute for void functions.
12807 2018-01-03 Martin Sebor <msebor@redhat.com>
12809 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12810 offset_int::from instead of wide_int::to_shwi.
12811 (maybe_diag_overlap): Remove assertion.
12812 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12813 * gimple-ssa-sprintf.c (format_directive): Same.
12814 (parse_directive): Same.
12815 (sprintf_dom_walker::compute_format_length): Same.
12816 (try_substitute_return_value): Same.
12818 2018-01-03 Jeff Law <law@redhat.com>
12820 PR middle-end/83654
12821 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12822 non-constant residual for zero at runtime and avoid probing in
12823 that case. Reorganize code for trailing problem to mirror handling
12826 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12828 PR tree-optimization/83501
12829 * tree-ssa-strlen.c (get_string_cst): New.
12830 (handle_char_store): Call get_string_cst.
12832 2018-01-03 Martin Liska <mliska@suse.cz>
12834 PR tree-optimization/83593
12835 * tree-ssa-strlen.c: Include tree-cfg.h.
12836 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12837 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12838 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12840 (strlen_dom_walker::before_dom_children): Call
12841 gimple_purge_dead_eh_edges. Dump tranformation with details
12843 (strlen_dom_walker::before_dom_children): Update call by adding
12844 new argument cleanup_eh.
12845 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12847 2018-01-03 Martin Liska <mliska@suse.cz>
12850 * cif-code.def (VARIADIC_THUNK): New enum value.
12851 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12854 2018-01-03 Jan Beulich <jbeulich@suse.com>
12856 * sse.md (mov<mode>_internal): Tighten condition for when to use
12857 vmovdqu<ssescalarsize> for TI and OI modes.
12859 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12861 Update copyright years.
12863 2018-01-03 Martin Liska <mliska@suse.cz>
12866 * ipa-visibility.c (function_and_variable_visibility): Skip
12867 functions with noipa attribure.
12869 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12871 * gcc.c (process_command): Update copyright notice dates.
12872 * gcov-dump.c (print_version): Ditto.
12873 * gcov.c (print_version): Ditto.
12874 * gcov-tool.c (print_version): Ditto.
12875 * gengtype.c (create_file): Ditto.
12876 * doc/cpp.texi: Bump @copying's copyright year.
12877 * doc/cppinternals.texi: Ditto.
12878 * doc/gcc.texi: Ditto.
12879 * doc/gccint.texi: Ditto.
12880 * doc/gcov.texi: Ditto.
12881 * doc/install.texi: Ditto.
12882 * doc/invoke.texi: Ditto.
12884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12886 * vector-builder.h (vector_builder::m_full_nelts): Change from
12887 unsigned int to poly_uint64.
12888 (vector_builder::full_nelts): Update prototype accordingly.
12889 (vector_builder::new_vector): Likewise.
12890 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12891 (vector_builder::operator ==): Likewise.
12892 (vector_builder::finalize): Likewise.
12893 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12894 Take the number of elements as a poly_uint64 rather than an
12896 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12897 from unsigned int to poly_uint64.
12898 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12899 (vec_perm_indices::new_vector): Likewise.
12900 (vec_perm_indices::length): Likewise.
12901 (vec_perm_indices::nelts_per_input): Likewise.
12902 (vec_perm_indices::input_nelts): Likewise.
12903 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12904 number of elements per input as a poly_uint64 rather than an
12905 unsigned int. Use the original encoding for variable-length
12906 vectors, rather than clamping each individual element.
12907 For the second and subsequent elements in each pattern,
12908 clamp the step and base before clamping their sum.
12909 (vec_perm_indices::series_p): Handle polynomial element counts.
12910 (vec_perm_indices::all_in_range_p): Likewise.
12911 (vec_perm_indices_to_tree): Likewise.
12912 (vec_perm_indices_to_rtx): Likewise.
12913 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12914 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12915 (tree_vector_builder::new_binary_operation): Handle polynomial
12916 element counts. Return false if we need to know the number
12917 of elements at compile time.
12918 * fold-const.c (fold_vec_perm): Punt if the number of elements
12919 isn't known at compile time.
12921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12923 * vec-perm-indices.h (vec_perm_builder): Change element type
12924 from HOST_WIDE_INT to poly_int64.
12925 (vec_perm_indices::element_type): Update accordingly.
12926 (vec_perm_indices::clamp): Handle polynomial element_types.
12927 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12928 (vec_perm_indices::all_in_range_p): Likewise.
12929 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12931 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12932 polynomial vec_perm_indices element types.
12933 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12934 * fold-const.c (fold_vec_perm): Likewise.
12935 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12936 * tree-vect-generic.c (lower_vec_perm): Likewise.
12937 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12938 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12939 element type to HOST_WIDE_INT.
12941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12942 Alan Hayward <alan.hayward@arm.com>
12943 David Sherwood <david.sherwood@arm.com>
12945 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12946 rather than an int. Use plus_constant.
12947 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12948 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12951 Alan Hayward <alan.hayward@arm.com>
12952 David Sherwood <david.sherwood@arm.com>
12954 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12955 a HOST_WIDE_INT to a poly_int64.
12957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12958 Alan Hayward <alan.hayward@arm.com>
12959 David Sherwood <david.sherwood@arm.com>
12961 * calls.c (load_register_parameters): Cope with polynomial
12962 mode sizes. Require a constant size for BLKmode parameters
12963 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12964 forces a parameter to be padded at the lsb end in order to
12965 fill a complete number of words, require the parameter size
12966 to be ordered wrt UNITS_PER_WORD.
12968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12969 Alan Hayward <alan.hayward@arm.com>
12970 David Sherwood <david.sherwood@arm.com>
12972 * reload1.c (spill_stack_slot_width): Change element type
12973 from unsigned int to poly_uint64_pod.
12974 (alter_reg): Treat mode sizes as polynomial.
12976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12977 Alan Hayward <alan.hayward@arm.com>
12978 David Sherwood <david.sherwood@arm.com>
12980 * reload.c (complex_word_subreg_p): New function.
12981 (reload_inner_reg_of_subreg, push_reload): Use it.
12983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12984 Alan Hayward <alan.hayward@arm.com>
12985 David Sherwood <david.sherwood@arm.com>
12987 * lra-constraints.c (process_alt_operands): Reject matched
12988 operands whose sizes aren't ordered.
12989 (match_reload): Refer to this check here.
12991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12992 Alan Hayward <alan.hayward@arm.com>
12993 David Sherwood <david.sherwood@arm.com>
12995 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12996 that the mode size is in the set {1, 2, 4, 8, 16}.
12998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12999 Alan Hayward <alan.hayward@arm.com>
13000 David Sherwood <david.sherwood@arm.com>
13002 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13003 Use plus_constant instead of gen_rtx_PLUS.
13005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13006 Alan Hayward <alan.hayward@arm.com>
13007 David Sherwood <david.sherwood@arm.com>
13009 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13010 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13011 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13012 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13013 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13014 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13015 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13016 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13017 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13018 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13020 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13021 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13022 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13023 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13024 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13025 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13026 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13027 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13028 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13029 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13031 * expr.c (emit_move_resolve_push): Treat the input and result
13032 of PUSH_ROUNDING as a poly_int64.
13033 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13034 (emit_push_insn): Likewise.
13035 * lra-eliminations.c (mark_not_eliminable): Likewise.
13036 * recog.c (push_operand): Likewise.
13037 * reload1.c (elimination_effects): Likewise.
13038 * rtlanal.c (nonzero_bits1): Likewise.
13039 * calls.c (store_one_arg): Likewise. Require the padding to be
13040 known at compile time.
13042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13043 Alan Hayward <alan.hayward@arm.com>
13044 David Sherwood <david.sherwood@arm.com>
13046 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13047 Use plus_constant instead of gen_rtx_PLUS.
13049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13050 Alan Hayward <alan.hayward@arm.com>
13051 David Sherwood <david.sherwood@arm.com>
13053 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13054 rather than an int.
13056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13057 Alan Hayward <alan.hayward@arm.com>
13058 David Sherwood <david.sherwood@arm.com>
13060 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13061 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13062 via stack temporaries. Treat the mode size as polynomial too.
13064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13065 Alan Hayward <alan.hayward@arm.com>
13066 David Sherwood <david.sherwood@arm.com>
13068 * expr.c (expand_expr_real_2): When handling conversions involving
13069 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13070 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13071 as a poly_uint64 too.
13073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13074 Alan Hayward <alan.hayward@arm.com>
13075 David Sherwood <david.sherwood@arm.com>
13077 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13080 Alan Hayward <alan.hayward@arm.com>
13081 David Sherwood <david.sherwood@arm.com>
13083 * combine.c (can_change_dest_mode): Handle polynomial
13084 REGMODE_NATURAL_SIZE.
13085 * expmed.c (store_bit_field_1): Likewise.
13086 * expr.c (store_constructor): Likewise.
13087 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13088 and polynomial REGMODE_NATURAL_SIZE.
13089 (gen_lowpart_common): Likewise.
13090 * reginfo.c (record_subregs_of_mode): Likewise.
13091 * rtlanal.c (read_modify_subreg_p): Likewise.
13093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13094 Alan Hayward <alan.hayward@arm.com>
13095 David Sherwood <david.sherwood@arm.com>
13097 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13098 numbers of elements.
13100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13101 Alan Hayward <alan.hayward@arm.com>
13102 David Sherwood <david.sherwood@arm.com>
13104 * match.pd: Cope with polynomial numbers of vector elements.
13106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13107 Alan Hayward <alan.hayward@arm.com>
13108 David Sherwood <david.sherwood@arm.com>
13110 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13111 in a POINTER_PLUS_EXPR.
13113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13114 Alan Hayward <alan.hayward@arm.com>
13115 David Sherwood <david.sherwood@arm.com>
13117 * omp-simd-clone.c (simd_clone_subparts): New function.
13118 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13119 (ipa_simd_modify_function_body): Likewise.
13121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13122 Alan Hayward <alan.hayward@arm.com>
13123 David Sherwood <david.sherwood@arm.com>
13125 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13126 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13127 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13128 (expand_vector_condition, vector_element): Likewise.
13129 (subparts_gt): New function.
13130 (get_compute_type): Use subparts_gt.
13131 (count_type_subparts): Delete.
13132 (expand_vector_operations_1): Use subparts_gt instead of
13133 count_type_subparts.
13135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13136 Alan Hayward <alan.hayward@arm.com>
13137 David Sherwood <david.sherwood@arm.com>
13139 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13140 (vect_compile_time_alias): ...this new function. Do the calculation
13141 on poly_ints rather than trees.
13142 (vect_prune_runtime_alias_test_list): Update call accordingly.
13144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13145 Alan Hayward <alan.hayward@arm.com>
13146 David Sherwood <david.sherwood@arm.com>
13148 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13150 (vect_schedule_slp_instance): Likewise.
13152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13153 Alan Hayward <alan.hayward@arm.com>
13154 David Sherwood <david.sherwood@arm.com>
13156 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13157 constant and extern definitions for variable-length vectors.
13158 (vect_get_constant_vectors): Note that the number of units
13159 is known to be constant.
13161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13162 Alan Hayward <alan.hayward@arm.com>
13163 David Sherwood <david.sherwood@arm.com>
13165 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13166 of units as polynomial. Choose between WIDE and NARROW based
13169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13170 Alan Hayward <alan.hayward@arm.com>
13171 David Sherwood <david.sherwood@arm.com>
13173 * tree-vect-stmts.c (simd_clone_subparts): New function.
13174 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13177 Alan Hayward <alan.hayward@arm.com>
13178 David Sherwood <david.sherwood@arm.com>
13180 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13181 vectors as polynomial. Use build_index_vector for
13182 IFN_GOMP_SIMD_LANE.
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 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13189 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13190 for variable-length vectors.
13191 (vectorizable_mask_load_store): Treat the number of units as
13192 polynomial, asserting that it is constant if the condition has
13193 already been enforced.
13194 (vectorizable_store, vectorizable_load): Likewise.
13196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13197 Alan Hayward <alan.hayward@arm.com>
13198 David Sherwood <david.sherwood@arm.com>
13200 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13201 of units as polynomial. Punt if we can't tell at compile time
13202 which vector contains the final result.
13204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13205 Alan Hayward <alan.hayward@arm.com>
13206 David Sherwood <david.sherwood@arm.com>
13208 * tree-vect-loop.c (vectorizable_induction): Treat the number
13209 of units as polynomial. Punt on SLP inductions. Use an integer
13210 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13211 cast of such a series for variable-length floating-point
13214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13215 Alan Hayward <alan.hayward@arm.com>
13216 David Sherwood <david.sherwood@arm.com>
13218 * tree.h (build_index_vector): Declare.
13219 * tree.c (build_index_vector): New function.
13220 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13221 of units as polynomial, forcibly converting it to a constant if
13222 vectorizable_reduction has already enforced the condition.
13223 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13224 to create a {1,2,3,...} vector.
13225 (vectorizable_reduction): Treat the number of units as polynomial.
13226 Choose vectype_in based on the largest scalar element size rather
13227 than the smallest number of units. Enforce the restrictions
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-data-refs.c (vector_alignment_reachable_p): Treat the
13235 number of units as polynomial.
13237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13238 Alan Hayward <alan.hayward@arm.com>
13239 David Sherwood <david.sherwood@arm.com>
13241 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13242 * target.def (autovectorize_vector_sizes): Return the vector sizes
13243 by pointer, using vector_sizes rather than a bitmask.
13244 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13245 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13246 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13248 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13249 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13250 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13251 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13252 * omp-general.c (omp_max_vf): Likewise.
13253 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13254 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13255 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13256 * tree-vect-slp.c (vect_slp_bb): Likewise.
13257 * doc/tm.texi: Regenerate.
13258 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13260 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13261 the vector size as a poly_uint64 rather than an unsigned int.
13262 (current_vector_size): Change from an unsigned int to a poly_uint64.
13263 (get_vectype_for_scalar_type): Update accordingly.
13264 * tree.h (build_truth_vector_type): Take the size and number of
13265 units as a poly_uint64 rather than an unsigned int.
13266 (build_vector_type): Add a temporary overload that takes
13267 the number of units as a poly_uint64 rather than an unsigned int.
13268 * tree.c (make_vector_type): Likewise.
13269 (build_truth_vector_type): Take the number of units as a poly_uint64
13270 rather than an unsigned int.
13272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13273 Alan Hayward <alan.hayward@arm.com>
13274 David Sherwood <david.sherwood@arm.com>
13276 * target.def (get_mask_mode): Take the number of units and length
13277 as poly_uint64s rather than unsigned ints.
13278 * targhooks.h (default_get_mask_mode): Update accordingly.
13279 * targhooks.c (default_get_mask_mode): Likewise.
13280 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13281 * doc/tm.texi: Regenerate.
13283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13284 Alan Hayward <alan.hayward@arm.com>
13285 David Sherwood <david.sherwood@arm.com>
13287 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13288 * omp-general.c (omp_max_vf): Likewise.
13289 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13290 (expand_omp_simd): Handle polynomial safelen.
13291 * omp-low.c (omplow_simd_context): Add a default constructor.
13292 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13293 (lower_rec_simd_input_clauses): Update accordingly.
13294 (lower_rec_input_clauses): Likewise.
13296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13297 Alan Hayward <alan.hayward@arm.com>
13298 David Sherwood <david.sherwood@arm.com>
13300 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13301 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13302 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13303 (vect_analyze_slp_cost): Likewise.
13304 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13305 (vect_model_load_cost): Likewise.
13307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13308 Alan Hayward <alan.hayward@arm.com>
13309 David Sherwood <david.sherwood@arm.com>
13311 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13312 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13313 from an unsigned int * to a poly_uint64_pod *.
13314 (calculate_unrolling_factor): New function.
13315 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13318 Alan Hayward <alan.hayward@arm.com>
13319 David Sherwood <david.sherwood@arm.com>
13321 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13322 from an unsigned int to a poly_uint64.
13323 (_loop_vec_info::slp_unrolling_factor): Likewise.
13324 (_loop_vec_info::vectorization_factor): Change from an int
13326 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13327 (vect_get_num_vectors): New function.
13328 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13329 (vect_get_num_copies): Use vect_get_num_vectors.
13330 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13331 to an unsigned int *.
13332 (vect_analyze_data_refs): Change min_vf from an int * to a
13334 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13335 than an unsigned HOST_WIDE_INT.
13336 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13337 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13338 to an unsigned int *.
13339 (vect_analyze_data_ref_dependences): Likewise.
13340 (vect_compute_data_ref_alignment): Handle polynomial vf.
13341 (vect_enhance_data_refs_alignment): Likewise.
13342 (vect_prune_runtime_alias_test_list): Likewise.
13343 (vect_shift_permute_load_chain): Likewise.
13344 (vect_supportable_dr_alignment): Likewise.
13345 (dependence_distance_ge_vf): Take the vectorization factor as a
13346 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13347 (vect_analyze_data_refs): Change min_vf from an int * to a
13349 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13350 vfm1 as a poly_uint64 rather than an int. Make the same change
13351 for the returned bound_scalar.
13352 (vect_gen_vector_loop_niters): Handle polynomial vf.
13353 (vect_do_peeling): Likewise. Update call to
13354 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13355 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13357 * tree-vect-loop.c (vect_determine_vectorization_factor)
13358 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13359 (vect_get_known_peeling_cost): Likewise.
13360 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13361 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13362 (vect_transform_loop): Likewise. Use the lowest possible VF when
13363 updating the upper bounds of the loop.
13364 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13365 rather than an int.
13366 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13367 polynomial unroll factors.
13368 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13369 (vect_make_slp_decision): Likewise.
13370 (vect_supported_load_permutation_p): Likewise, and polynomial
13372 (vect_analyze_slp_cost): Handle polynomial vf.
13373 (vect_slp_analyze_node_operations): Likewise.
13374 (vect_slp_analyze_bb_1): Likewise.
13375 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13376 than an unsigned HOST_WIDE_INT.
13377 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13378 (vectorizable_load): Handle polynomial vf.
13379 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13381 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13384 Alan Hayward <alan.hayward@arm.com>
13385 David Sherwood <david.sherwood@arm.com>
13387 * match.pd: Handle bit operations involving three constants
13388 and try to fold one pair.
13390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13392 * tree-vect-loop-manip.c: Include gimple-fold.h.
13393 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13394 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13395 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13396 Add a path that uses a step of VF instead of 1, but disable it
13398 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13399 and niters_no_overflow parameters. Update calls to
13400 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13401 Create a new SSA name if the latter choses to use a ste other
13402 than zero, and return it via niters_vector_mult_vf_var.
13403 * tree-vect-loop.c (vect_transform_loop): Update calls to
13404 vect_do_peeling, vect_gen_vector_loop_niters and
13405 slpeel_make_loop_iterate_ntimes.
13406 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13407 (vect_gen_vector_loop_niters): Update declarations after above changes.
13409 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13411 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13412 128-bit round to integer instructions.
13413 (ceil<mode>2): Likewise.
13414 (btrunc<mode>2): Likewise.
13415 (round<mode>2): Likewise.
13417 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13419 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13420 unaligned VSX load/store on P8/P9.
13421 (expand_block_clear): Allow the use of unaligned VSX
13422 load/store on P8/P9.
13424 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13426 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13428 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13429 swap associated with both a load and a store.
13431 2018-01-02 Andrew Waterman <andrew@sifive.com>
13433 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13434 * config/riscv/riscv.md (clear_cache): Use it.
13436 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13438 * web.c: Remove out-of-date comment.
13440 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13442 * expr.c (fixup_args_size_notes): Check that any existing
13443 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13444 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13445 (emit_single_push_insn): ...here.
13447 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13449 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13450 (const_vector_encoded_nelts): New function.
13451 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13452 (const_vector_int_elt, const_vector_elt): Declare.
13453 * emit-rtl.c (const_vector_int_elt_1): New function.
13454 (const_vector_elt): Likewise.
13455 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13456 of CONST_VECTOR_ELT.
13458 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13460 * expr.c: Include rtx-vector-builder.h.
13461 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13462 directly on the tree encoding.
13463 (const_vector_from_tree): Likewise.
13464 * optabs.c: Include rtx-vector-builder.h.
13465 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13466 sequence of "u" values.
13467 * vec-perm-indices.c: Include rtx-vector-builder.h.
13468 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13469 directly on the vec_perm_indices encoding.
13471 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13473 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13474 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13475 * rtx-vector-builder.h: New file.
13476 * rtx-vector-builder.c: Likewise.
13477 * rtl.h (rtx_def::u2): Add a const_vector field.
13478 (CONST_VECTOR_NPATTERNS): New macro.
13479 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13480 (CONST_VECTOR_DUPLICATE_P): Likewise.
13481 (CONST_VECTOR_STEPPED_P): Likewise.
13482 (CONST_VECTOR_ENCODED_ELT): Likewise.
13483 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13484 (unwrap_const_vec_duplicate): Likewise.
13485 (const_vec_series_p): Check for a non-duplicated vector encoding.
13486 Say that the function only returns true for integer vectors.
13487 * emit-rtl.c: Include rtx-vector-builder.h.
13488 (gen_const_vec_duplicate_1): Delete.
13489 (gen_const_vector): Call gen_const_vec_duplicate instead of
13490 gen_const_vec_duplicate_1.
13491 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13492 (gen_const_vec_duplicate): Use rtx_vector_builder.
13493 (gen_const_vec_series): Likewise.
13494 (gen_rtx_CONST_VECTOR): Likewise.
13495 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13496 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13497 Build a new vector rather than modifying a CONST_VECTOR in-place.
13498 (handle_special_swappables): Update call accordingly.
13499 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13500 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13501 Build a new vector rather than modifying a CONST_VECTOR in-place.
13502 (handle_special_swappables): Update call accordingly.
13504 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13506 * simplify-rtx.c (simplify_const_binary_operation): Use
13507 CONST_VECTOR_ELT instead of XVECEXP.
13509 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13511 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13512 the selector elements to be different from the data elements
13513 if the selector is a VECTOR_CST.
13514 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13515 ssizetype for the selector.
13517 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13519 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13520 before testing each element individually.
13521 * tree-vect-generic.c (lower_vec_perm): Likewise.
13523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13525 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13526 * selftest-run-tests.c (selftest::run_tests): Call it.
13527 * vector-builder.h (vector_builder::operator ==): New function.
13528 (vector_builder::operator !=): Likewise.
13529 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13530 (vec_perm_indices::all_from_input_p): New function.
13531 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13532 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13533 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13534 instead of reading the VECTOR_CST directly. Detect whether both
13535 vector inputs are the same before constructing the vec_perm_indices,
13536 and update the number of inputs argument accordingly. Use the
13537 utility functions added above. Only construct sel2 if we need to.
13539 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13541 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13542 the broadcast of the low byte.
13543 (expand_mult_highpart): Use an explicit encoding for the permutes.
13544 * optabs-query.c (can_mult_highpart_p): Likewise.
13545 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13546 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13547 (vectorizable_bswap): Likewise.
13548 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13549 explicit encoding for the power-of-2 permutes.
13550 (vect_permute_store_chain): Likewise.
13551 (vect_grouped_load_supported): Likewise.
13552 (vect_permute_load_chain): Likewise.
13554 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13556 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13557 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13558 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13559 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13560 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13561 (vect_gen_perm_mask_any): Likewise.
13563 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13565 * int-vector-builder.h: New file.
13566 * vec-perm-indices.h: Include int-vector-builder.h.
13567 (vec_perm_indices): Redefine as an int_vector_builder.
13568 (auto_vec_perm_indices): Delete.
13569 (vec_perm_builder): Redefine as a stand-alone class.
13570 (vec_perm_indices::vec_perm_indices): New function.
13571 (vec_perm_indices::clamp): Likewise.
13572 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13573 (vec_perm_indices::new_vector): New function.
13574 (vec_perm_indices::new_expanded_vector): Update for new
13575 vec_perm_indices class.
13576 (vec_perm_indices::rotate_inputs): New function.
13577 (vec_perm_indices::all_in_range_p): Operate directly on the
13578 encoded form, without computing elided elements.
13579 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13580 encoding. Update for new vec_perm_indices class.
13581 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13582 the given vec_perm_builder.
13583 (expand_vec_perm_var): Update vec_perm_builder constructor.
13584 (expand_mult_highpart): Use vec_perm_builder instead of
13585 auto_vec_perm_indices.
13586 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13587 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13588 or double series encoding as appropriate.
13589 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13590 vec_perm_indices instead of auto_vec_perm_indices.
13591 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13592 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13593 (vect_permute_store_chain): Likewise.
13594 (vect_grouped_load_supported): Likewise.
13595 (vect_permute_load_chain): Likewise.
13596 (vect_shift_permute_load_chain): Likewise.
13597 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13598 (vect_transform_slp_perm_load): Likewise.
13599 (vect_schedule_slp_instance): Likewise.
13600 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13601 (vectorizable_mask_load_store): Likewise.
13602 (vectorizable_bswap): Likewise.
13603 (vectorizable_store): Likewise.
13604 (vectorizable_load): Likewise.
13605 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13606 vec_perm_indices instead of auto_vec_perm_indices. Use
13607 tree_to_vec_perm_builder to read the vector from a tree.
13608 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13609 vec_perm_builder instead of a vec_perm_indices.
13610 (have_whole_vector_shift): Use vec_perm_builder and
13611 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13612 truncation to calc_vec_perm_mask_for_shift.
13613 (vect_create_epilog_for_reduction): Likewise.
13614 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13615 from auto_vec_perm_indices to vec_perm_indices.
13616 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13617 instead of changing individual elements.
13618 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13619 the vector in d.perm.
13620 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13621 from auto_vec_perm_indices to vec_perm_indices.
13622 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13623 instead of changing individual elements.
13624 (arm_vectorize_vec_perm_const): Use new_vector to install
13625 the vector in d.perm.
13626 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13627 Update vec_perm_builder constructor.
13628 (rs6000_expand_interleave): Likewise.
13629 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13630 (rs6000_expand_interleave): Likewise.
13632 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13634 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13635 to qimode could truncate the indices.
13636 * optabs.c (expand_vec_perm_var): Likewise.
13638 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13640 * Makefile.in (OBJS): Add vec-perm-indices.o.
13641 * vec-perm-indices.h: New file.
13642 * vec-perm-indices.c: Likewise.
13643 * target.h (vec_perm_indices): Replace with a forward class
13645 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13646 * optabs.h: Include vec-perm-indices.h.
13647 (expand_vec_perm): Delete.
13648 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13649 (expand_vec_perm_const): Declare.
13650 * target.def (vec_perm_const_ok): Replace with...
13651 (vec_perm_const): ...this new hook.
13652 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13653 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13654 * doc/tm.texi: Regenerate.
13655 * optabs.def (vec_perm_const): Delete.
13656 * doc/md.texi (vec_perm_const): Likewise.
13657 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13658 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13659 expand_vec_perm for constant permutation vectors. Assert that
13660 the mode of variable permutation vectors is the integer equivalent
13661 of the mode that is being permuted.
13662 * optabs-query.h (selector_fits_mode_p): Declare.
13663 * optabs-query.c: Include vec-perm-indices.h.
13664 (selector_fits_mode_p): New function.
13665 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13666 is defined, instead of checking whether the vec_perm_const_optab
13667 exists. Use targetm.vectorize.vec_perm_const instead of
13668 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13669 fit in the vector mode before using a variable permute.
13670 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13671 vec_perm_indices instead of an rtx.
13672 (expand_vec_perm): Replace with...
13673 (expand_vec_perm_const): ...this new function. Take the selector
13674 as a vec_perm_indices rather than an rtx. Also take the mode of
13675 the selector. Update call to shift_amt_for_vec_perm_mask.
13676 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13677 Use vec_perm_indices::new_expanded_vector to expand the original
13678 selector into bytes. Check whether the indices fit in the vector
13679 mode before using a variable permute.
13680 (expand_vec_perm_var): Make global.
13681 (expand_mult_highpart): Use expand_vec_perm_const.
13682 * fold-const.c: Includes vec-perm-indices.h.
13683 * tree-ssa-forwprop.c: Likewise.
13684 * tree-vect-data-refs.c: Likewise.
13685 * tree-vect-generic.c: Likewise.
13686 * tree-vect-loop.c: Likewise.
13687 * tree-vect-slp.c: Likewise.
13688 * tree-vect-stmts.c: Likewise.
13689 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13691 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13692 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13693 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13694 (aarch64_vectorize_vec_perm_const): ...this new function.
13695 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13696 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13697 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13698 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13699 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13700 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13701 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13703 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13704 check for NEON modes.
13705 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13706 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13707 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13708 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13710 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13711 the old VEC_PERM_CONST conditions.
13712 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13713 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13714 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13715 (ia64_vectorize_vec_perm_const_ok): Merge into...
13716 (ia64_vectorize_vec_perm_const): ...this new function.
13717 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13718 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13719 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13720 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13721 * config/mips/mips.c (mips_expand_vec_perm_const)
13722 (mips_vectorize_vec_perm_const_ok): Merge into...
13723 (mips_vectorize_vec_perm_const): ...this new function.
13724 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13725 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13726 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13727 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13728 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13729 (rs6000_expand_vec_perm_const): Delete.
13730 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13732 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13733 (altivec_expand_vec_perm_const_le): Take each operand individually.
13734 Operate on constant selectors rather than rtxes.
13735 (altivec_expand_vec_perm_const): Likewise. Update call to
13736 altivec_expand_vec_perm_const_le.
13737 (rs6000_expand_vec_perm_const): Delete.
13738 (rs6000_vectorize_vec_perm_const_ok): Delete.
13739 (rs6000_vectorize_vec_perm_const): New function.
13740 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13741 an element count and rtx array.
13742 (rs6000_expand_extract_even): Update call accordingly.
13743 (rs6000_expand_interleave): Likewise.
13744 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13745 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13746 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13747 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13748 (rs6000_expand_vec_perm_const): Delete.
13749 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13750 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13751 (altivec_expand_vec_perm_const_le): Take each operand individually.
13752 Operate on constant selectors rather than rtxes.
13753 (altivec_expand_vec_perm_const): Likewise. Update call to
13754 altivec_expand_vec_perm_const_le.
13755 (rs6000_expand_vec_perm_const): Delete.
13756 (rs6000_vectorize_vec_perm_const_ok): Delete.
13757 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13758 reference to the SPE evmerge intructions.
13759 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13760 an element count and rtx array.
13761 (rs6000_expand_extract_even): Update call accordingly.
13762 (rs6000_expand_interleave): Likewise.
13763 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13764 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13766 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13768 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13770 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13771 vector mode and that that mode matches the mode of the data
13773 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13774 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13775 directly using expand_vec_perm_1 when forcing selectors into
13777 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13779 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13781 * optabs-query.h (can_vec_perm_p): Delete.
13782 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13783 * optabs-query.c (can_vec_perm_p): Split into...
13784 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13785 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13786 particular selector is valid.
13787 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13788 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13789 (vect_grouped_load_supported): Likewise.
13790 (vect_shift_permute_load_chain): Likewise.
13791 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13792 (vect_transform_slp_perm_load): Likewise.
13793 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13794 (vectorizable_bswap): Likewise.
13795 (vect_gen_perm_mask_checked): Likewise.
13796 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13797 implementations of variable permutation vectors into account
13798 when deciding which selector to use.
13799 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13800 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13801 with a false third argument.
13802 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13803 to test whether the constant selector is valid and can_vec_perm_var_p
13804 to test whether a variable selector is valid.
13806 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13808 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13809 * optabs-query.c (can_vec_perm_p): Likewise.
13810 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13811 instead of vec_perm_indices.
13812 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13813 (vect_gen_perm_mask_checked): Likewise,
13814 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13815 (vect_gen_perm_mask_checked): Likewise,
13817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13819 * optabs-query.h (qimode_for_vec_perm): Declare.
13820 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13821 (qimode_for_vec_perm): ...this new function.
13822 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13824 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13826 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13827 does not have a conditional at the top.
13829 2018-01-02 Richard Biener <rguenther@suse.de>
13831 * ipa-inline.c (big_speedup_p): Fix expression.
13833 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13836 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13839 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13843 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13844 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13845 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13846 cond_taken_branch_cost 3->4.
13848 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13850 PR tree-optimization/83581
13851 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13852 TODO_cleanup_cfg if any changes have been made.
13854 PR middle-end/83608
13855 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13856 convert_modes if target mode has the right side, but different mode
13859 PR middle-end/83609
13860 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13861 last argument when extracting from CONCAT. If either from_real or
13862 from_imag is NULL, use expansion through memory. If result is not
13863 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13864 the parts directly to inner mode, if even that fails, use expansion
13867 PR middle-end/83623
13868 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13869 check for bswap in mode rather than HImode and use that in expand_unop
13872 Copyright (C) 2018 Free Software Foundation, Inc.
13874 Copying and distribution of this file, with or without modification,
13875 are permitted in any medium without royalty provided the copyright
13876 notice and this notice are preserved.