1 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
6 (cgraph_node::create_alias): Likewise.
7 (cgraph_node::get_availability): Check ifunc_resolver instead
8 of looking up ifunc attribute.
9 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
10 * varasm.c (do_assemble_alias): Likewise.
11 (assemble_alias): Likewise.
12 (default_binds_local_p_3): Likewise.
13 * cgraph.h (cgraph_node): Add ifunc_resolver.
14 (cgraph_node::only_called_directly_or_aliased_p): Return false
16 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
19 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
20 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
21 instead of looking up ifunc attribute.
23 2018-05-22 Luis Machado <luis.machado@linaro.org>
25 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
27 2018-05-22 Martin Sebor <msebor@redhat.com>
29 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
31 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
32 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34 * config/aarch64/aarch64-ldpstp.md: Replace uses of
35 aarch64_mem_pair_operand with memory_operand and delete operand swapping
37 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
38 Add check for legitimate_address.
39 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
40 (aarch64_swap_ldrstr_operands): New.
41 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
44 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
45 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
47 * config/aarch64/aarch64.md: New patterns to generate stp
49 (store_pair_sw, store_pair_dw): New patterns to generate stp for
50 single words and double words.
51 (load_pair_sw, load_pair_dw): Likewise.
52 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
54 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
56 * config/aarch64/aarch64-ldpstp.md: Modify peephole
57 for different mode ldpstp and add peephole for merged zero stores.
59 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
61 (aarch64_gen_store_pair): Rename calls to match new patterns.
62 (aarch64_gen_load_pair): Rename calls to match new patterns.
63 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
64 (load_pair<DREG:mode><DREG2:mode>): ... This.
65 (store_pair<mode>): Rename to...
66 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
67 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
69 (V_INT_EQUIV): Handle SImode.
70 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
73 2018-05-22 Martin Sebor <msebor@redhat.com>
76 * calls.c (maybe_warn_nonstring_arg): Use string length to set
77 or ajust the presumed bound on an operation to avoid unnecessary
80 2018-05-22 Martin Sebor <msebor@redhat.com>
82 PR tree-optimization/85826
83 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
84 assuming that a DECL necesarily has a constant size.
86 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
89 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
91 2018-05-22 Richard Biener <rguenther@suse.de>
93 PR tree-optimization/85834
94 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
95 non-constant and non-zero memset arguments.
97 2018-05-22 Martin Liska <mliska@suse.cz>
100 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
102 2018-05-22 Richard Biener <rguenther@suse.de>
104 PR tree-optimization/85863
105 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
106 comparisons when vectype is specified.
107 (vectorizable_condition): Do not specify vectype for
108 vect_is_simple_cond when SLP vectorizing.
110 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
113 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
114 define __ibm128 as long double.
115 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
116 as a distinct type when IEEE 128-bit support is enabled.
117 (init_float128_ieee): Fix up conversions between IFmode and IEEE
118 128-bit types to use the correct functions.
119 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
120 convert between 128-bit floating point types that have different
121 modes but the same representation, instead of using gen_lowpart to
123 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
125 (IFKF_reg): New attributes to give the register constraints for
127 (extend<mode>tf2_internal): New insns to mark an explicit
128 conversion between 128-bit floating point types that have a
129 different mode but share the same representation.
131 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
133 PR tree-optimization/85814
134 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
135 a null return from get_strinfo when unsharing the next
136 strinfo in the chain.
138 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
141 * varasm.c (weak_finish): Clean up weak_decls.
143 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
145 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
146 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
147 UNSPEC_UADALP values.
148 * config/aarch64/iterators.md (ABAL): New int iterator.
151 (sur): Add mappings for the above.
152 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
154 (aarch64_<sur>abal<mode>_4): Likewise.
155 (aarch64_<sur>adalp<mode>_3): Likewise.
156 (<sur>sadv16qi): New define_expand.
158 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
160 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
161 (*movdf_internal): Ditto.
162 (*rcpsf2_sse): Ditto.
163 (*rsqrtsf2_sse): Ditto.
164 (*sqrt<mode>2_sse): Ditto.
166 2018-05-21 Tamar Christina <tamar.christina@arm.com>
168 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
170 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
171 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
172 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
173 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
175 * config/aarch64/arm_neon.h: Likewise.
176 * config/aarch64/iterators.md (VQ_I): New.
178 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
180 * config.gcc: Add arc/t-multilib-linux to tmake_file for
182 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
185 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
187 * config/nds32/constraints.md (S): New constraint.
188 * config/nds32/nds32.md (call_internal): Use constraint S.
189 (call_value_internal): Likewise.
190 (sibcall_internal): Likewise.
191 (sibcall_value_internal): Likewise.
193 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
194 Chung-Ju Wu <jasonwucj@gmail.com>
196 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
199 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
200 Chung-Ju Wu <jasonwucj@gmail.com>
202 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
203 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
204 (nds32_rtx_costs_impl): Simplify.
205 (nds32_address_cost_impl): Simplify.
206 (nds32_init_rtx_costs): New function.
207 (nds32_rtx_costs_speed_prefer): Likewise.
208 (nds32_rtx_costs_size_prefer): Likewise.
209 (nds32_address_cost_speed_prefer): Likewise.
210 (nds32_address_cost_speed_fwprop): Likewise.
211 (nds32_address_cost_size_prefer): Likewise.
212 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
213 * config/nds32/nds32.c (nds32_option_override): Use
214 nds32_init_rtx_costs function.
216 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
218 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
219 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
220 (TARGET_PIPELINE_N8): Likewise.
221 (TARGET_PIPELINE_N10): Likewise.
222 (TARGET_PIPELINE_N13): Likewise.
223 (TARGET_PIPELINE_GRAYWOLF): Likewise.
225 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
227 * config/nds32/nds32-fpu.md: Update copyright year.
229 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
231 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
233 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
235 * config/nds32/nds32.c
236 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
237 * config/nds32/nds32.opt (minline-asm-r15): New option.
239 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
241 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
243 * config/nds32/nds32.md (abssi2): New pattern.
245 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
247 * config/i386/i386.md (rex64namesuffix): New mode attribute.
248 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
249 Merge insn pattern from sse_cvtsi2ss<round_name> and
250 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
251 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
252 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
253 using SWI48 mode iterator.
254 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
255 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
256 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
257 pattern from sse_cvttss2si<round_saeonly_name>
258 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
259 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
260 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
261 using SWI48 mode iterator.
262 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
263 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
264 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
265 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
266 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
267 using SWI48 mode iterator.
268 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
269 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
270 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
271 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
272 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
274 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
275 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
276 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
277 pattern from sse_cvttsd2si<round_saeonly_name>
278 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
280 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
282 * config/nds32/nds32-md-auxiliary.c
283 (nds32_valid_smw_lwm_base_p): Refine.
284 (nds32_output_smw_single_word): Refine.
285 (nds32_output_smw_double_word): New.
286 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
288 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
290 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
291 (nds32_output_stack_pop): Refine.
292 (nds32_expand_unaligned_load): Refine.
293 (nds32_expand_unaligned_store): Refine.
295 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
296 Chung-Ju Wu <jasonwucj@gmail.com>
298 * config/nds32/constants.md: Add TP_REGNUM constant.
299 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
300 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
302 * config/nds32/nds32-doubleword.md: Consider flag_pic.
303 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
304 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
305 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
306 and PIC code generation.
307 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
309 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
311 * config/nds32/nds32.md: Support TLS and PIC.
312 * config/nds32/nds32.c: Support TLS and PIC.
313 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
314 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
317 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
319 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
322 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
323 Chung-Ju Wu <jasonwucj@gmail.com>
325 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
326 * config/nds32/nds32-md-auxiliary.c
327 (symbolic_reference_mentioned_p): New.
328 (nds32_legitimize_ict_address): New.
329 (nds32_expand_ict_move): New.
330 (nds32_indirect_call_referenced_p): New.
331 (nds32_symbol_binds_local_p): Delete.
332 (nds32_long_call_p): Modify.
333 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
334 * config/nds32/nds32-protos.h
335 (symbolic_reference_mentioned_p): Declare.
336 (nds32_legitimize_ict_address): Declare.
337 (nds32_expand_ict_move): Declare.
338 (nds32_indirect_call_referenced_p): Declare.
339 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
340 (nds32_relax_group): Use nds32_ict_const_p as condition.
341 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
342 (nds32_asm_file_start): Output ict_model directive in asm code.
343 (nds32_legitimate_address_p): Consider indirect call.
344 (nds32_print_operand): Consider indirect call.
345 (nds32_print_operand_address): Consider indirect call.
346 (nds32_insert_attributes): Handle "indirect_call" attribute.
347 (TARGET_LEGITIMATE_ADDRESS_P): Define.
348 (TARGET_LEGITIMATE_CONSTANT_P): Define.
349 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
350 (TARGET_DELEGITIMIZE_ADDRESS): Define.
351 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
352 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
353 (TARGET_ICT_MODEL_SMALL): Define.
354 (TARGET_ICT_MODEL_LARGE): Define.
355 * config/nds32/nds32.md (movsi): Consider ict model.
356 (call, call_value): Consider ict model.
357 (sibcall, sibcall_value): Consider ict model.
358 * config/nds32/nds32.opt (mict-model): New option.
359 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
362 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
363 Monk Chiang <sh.chiang04@gmail.com>
364 Jim Wilson <jimw@sifive.com>
366 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
367 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
368 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
369 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
370 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
371 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
372 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
373 compute save_libcall_adjustment properly.
374 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
375 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
376 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
377 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
378 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
379 (ABI_SPEC): Handle mabi=ilp32e.
380 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
382 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
383 <-march>: Add rv32e as an example.
385 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
388 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
389 (intra_create_variable_infos): Handle C++ constructors.
391 2018-05-18 Martin Liska <mliska@suse.cz>
393 * passes.def: Remove a redundant pass.
395 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
398 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
400 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
402 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
404 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
406 (armv2, armv2a, armv3, armv3m): Delete architectures.
407 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
408 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
409 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
411 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
412 (*mulsidi3adddi): Likewise.
413 (mulsidi3): Likewise.
414 (*mulsidi3_nov6): Likewise.
415 (umulsidi3): Likewise.
416 (umulsidi3_nov6): Likewise.
417 (umaddsidi4): Likewise.
418 (*umulsidi3adddi): Likewise.
419 (smulsi3_highpart): Likewise.
420 (*smulsi3_highpart_nov6): Likewise.
421 (umulsi3_highpart): Likewise.
422 (*umulsi3_highpart_nov6): Likewise.
423 * config/arm/arm.h (arm_arch3m): Delete.
424 * config/arm/arm.c (arm_arch3m): Delete.
425 (arm_option_override_internal): Update armv3-related comment.
426 (arm_configure_build_target): Delete use of isa_bit_mode32.
427 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
428 (arm_rtx_costs_internal): Delete check of arm_arch3m.
429 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
432 * config/arm/arm-protos.h (arm_arch3m): Delete.
433 * config/arm/arm-tables.opt: Regenerate.
434 * config/arm/arm-tune.md: Likewise.
435 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
436 deleted architectures.
438 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
440 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
441 (armv5t, armv5te): New features.
442 (ARMv5, ARMv5e): Delete fgroups.
443 (ARMv5t, ARMv5te): Adjust for above changes.
445 (armv5, armv5e): Delete arches.
446 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
448 (*call_reg_arm): Likewise.
449 (*call_value_reg_armv5): Likewise.
450 (*call_value_reg_arm): Likewise.
451 (*call_symbol): Likewise.
452 (*call_value_symbol): Likewise.
453 (*sibcall_insn): Likewise.
454 (*sibcall_value_insn): Likewise.
456 (prefetch): Likewise.
457 (define_split and define_peephole2 dependent on arm_arch5):
459 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
461 (TARGET_ARM_QBIT): Likewise.
462 (TARGET_DSP_MULTIPLY): Likewise.
463 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
464 (arm_arch5, arm_arch5e): Delete.
465 (arm_arch5t, arm_arch5te): Declare.
466 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
467 (arm_arch5t): Declare.
468 (arm_option_reconfigure_globals): Update for the above.
469 (arm_options_perform_arch_sanity_checks): Update comment, replace
470 use of arm_arch5 with arm_arch5t.
471 (use_return_insn): Likewise.
472 (arm_emit_call_insn): Likewise.
473 (output_return_instruction): Likewise.
474 (arm_final_prescan_insn): Likewise.
475 (arm_coproc_builtin_available): Likewise.
476 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
477 arm_arch5e with arm_arch5t and arm_arch5te.
478 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
479 (arm_arch5t, arm_arch5te): Declare.
480 * config/arm/arm-tables.opt: Regenerate.
481 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
482 * config/arm/t-multilib: Likewise.
483 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
484 instead of arm_arch5.
485 (*call_reg_thumb1): Likewise.
486 (*call_value_reg_thumb1_v5): Likewise.
487 (*call_value_reg_thumb1): Likewise.
488 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
490 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
492 2018-05-18 Martin Liska <mliska@suse.cz>
494 PR gcov-profile/84846
495 * doc/gcov.texi: Document -t option of gcov tool.
497 2018-05-18 Martin Liska <mliska@suse.cz>
499 PR gcov-profile/84846
500 * gcov.c (print_usage): Add new -t option.
501 (process_args): Handle the option.
502 (generate_results): Use stdout as output when requested by
505 2018-05-18 Martin Liska <mliska@suse.cz>
507 PR gcov-profile/84846
508 * coverage.c (coverage_init): Write PWD to .gcno file.
509 * doc/gcov.texi: Document how working directory is printed.
510 * gcov-dump.c (dump_gcov_file): Print PWD.
511 * gcov.c (output_intermediate_file): Likewise.
512 (read_graph_file): Read PWD string.
513 (output_lines): Print PWD.
515 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
518 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
519 for retval and return false if all args to phi are zero.
521 2018-05-18 Richard Biener <rguenther@suse.de>
523 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
525 (evrp_dom_walker::before_dom_children): Call it.
527 2018-05-18 Richard Biener <rguenther@suse.de>
529 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
530 results when processing array refs with variable index.
532 2018-05-18 Toon Moene <toon@moene.org>
534 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
535 directly after that of -floop-interchange. Indicate that both
536 options are enabled by default when specifying -O3.
538 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
540 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
541 iterator. Delete separate integer-mode vec_set<mode> expander.
542 (aarch64_simd_vec_setv2di): Delete.
543 (vec_setv2di): Delete.
544 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
545 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
546 the "w, r" alternative.
548 2018-05-18 Martin Liska <mliska@suse.cz>
550 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
551 * tree-pass.h (make_pass_lower_switch_O0): New function.
552 * tree-switch-conversion.c (node_has_low_bound): Remove.
553 (node_has_high_bound): Likewise.
554 (node_is_bounded): Likewise.
555 (class pass_lower_switch): Make it a template type and create
557 (pass_lower_switch::execute): Add template argument.
558 (make_pass_lower_switch): New function.
559 (make_pass_lower_switch_O0): New function.
560 (do_jump_if_equal): Remove.
561 (emit_case_nodes): Simplify to just handle all 3 cases and leave
562 all the hard work to tree optimization passes.
564 2018-05-18 Martin Liska <mliska@suse.cz>
566 * dbgcnt.c (limit_low): Renamed from limit.
567 (limit_high): New variable.
568 (dbg_cnt_is_enabled): Check for upper limit.
569 (dbg_cnt): Adjust dumping.
570 (dbg_cnt_set_limit_by_index): Add new argument for high
572 (dbg_cnt_set_limit_by_name): Likewise.
573 (dbg_cnt_process_single_pair): Parse new format.
574 (dbg_cnt_process_opt): Use strtok.
575 (dbg_cnt_list_all_counters): Remove 'value' and add
577 * doc/invoke.texi: Document changes.
579 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
581 * doc/sourcebuild.texi (scalar_all_fma): Document.
582 * tree.def (FMA_EXPR): Delete.
583 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
584 * internal-fn.c (ternary_direct): New macro.
585 (expand_ternary_optab_fn): Likewise.
586 (direct_ternary_optab_supported_p): Likewise.
587 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
588 * builtins.c (fold_builtin_fma): Delete.
589 (fold_builtin_3): Don't call it.
590 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
591 * expr.c (expand_expr_real_2): Likewise.
592 * fold-const.c (operand_equal_p): Likewise.
593 (fold_ternary_loc): Likewise.
594 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
595 * gimple.c (DEFTREECODE): Likewise.
596 * gimplify.c (gimplify_expr): Likewise.
597 * optabs-tree.c (optab_for_tree_code): Likewise.
598 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
599 * tree-eh.c (operation_could_trap_p): Likewise.
600 (stmt_could_throw_1_p): Likewise.
601 * tree-inline.c (estimate_operator_cost): Likewise.
602 * tree-pretty-print.c (dump_generic_node): Likewise.
603 (op_code_prio): Likewise.
604 * tree-ssa-loop-im.c (stmt_cost): Likewise.
605 * tree-ssa-operands.c (get_expr_operands): Likewise.
606 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
607 * fold-const-call.h (fold_fma): Delete.
608 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
609 CFN_FNMA and CFN_FNMS.
611 * genmatch.c (combined_fn): New enum.
612 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
613 (commutative_op): New function.
614 (commutate): Use it. Handle more than 2 operands.
615 (dt_operand::gen_gimple_expr): Use commutative_op.
616 (parser::parse_expr): Allow :c to be used with non-binary
617 operators if the commutative operand is known.
618 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
619 CFN_FMS, CFN_FNMA and CFN_FNMS.
620 (backprop::process_assign_use): Remove FMA_EXPR handling.
621 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
622 (gen_hsa_fma): New function.
623 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
624 IFN_FNMA and IFN_FNMS.
625 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
626 * gimple-fold.h (follow_all_ssa_edges): Declare.
627 * gimple-fold.c (follow_all_ssa_edges): New function.
628 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
629 gimple_build interface and use follow_all_ssa_edges to fold the result.
630 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
631 instead of checking for optabs directly.
632 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
633 rather than FMA_EXPRs.
634 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
635 call to IFN_FMA instead of an FMA_EXPR.
637 2018-05-17 Jim Wilson <jimw@sifive.com>
639 * expr.c (do_tablejump): When converting index to Pmode, if we have a
640 sign extended promoted subreg, and the range does not have the sign bit
641 set, then do a sign extend.
643 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
644 test, check for sign extended subreg and/or constant operands, and
645 do a sign extend in that case.
647 2018-05-17 Steve Ellcey <sellcey@cavium.com>
649 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
650 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
652 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
653 Change logics_shift_reg to logics_shift_imm.
654 (thunderx2t99_fp_loadpair_basic): Delete.
655 (thunderx2t99_fp_storepair_basic): Delete.
656 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
657 (thunderx2t99_asimd_polynomial): Delete.
658 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
659 and neon_fp_mul_d_scalar_q.
660 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
661 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
662 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
663 (thunderx2t99_asimd_lut): Add missing tbl types.
664 (thunderx2t99_asimd_ext): Delete.
665 (thunderx2t99_asimd_load1_1_mult): Delete.
666 (thunderx2t99_asimd_load1_2_mult): Delete.
667 (thunderx2t99_asimd_load1_ldp): New.
668 (thunderx2t99_asimd_load1): New.
669 (thunderx2t99_asimd_load2): Add missing *load2* types.
670 (thunderx2t99_asimd_load3): New.
671 (thunderx2t99_asimd_load4): New.
672 (thunderx2t99_asimd_store1_1_mult): Delete.
673 (thunderx2t99_asimd_store1_2_mult): Delete.
674 (thunderx2t99_asimd_store2_mult): Delete.
675 (thunderx2t99_asimd_store2_onelane): Delete.
676 (thunderx2t99_asimd_store_stp): New.
677 (thunderx2t99_asimd_store1): New.
678 (thunderx2t99_asimd_store2): New.
679 (thunderx2t99_asimd_store3): New.
680 (thunderx2t99_asimd_store4): New.
682 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
684 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
685 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
687 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
688 Segher Boessenkool <segher@kernel.crashing.org>
691 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
694 2018-05-17 Richard Biener <rguenther@suse.de>
696 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
697 for pruning loop and prune defs feeding only already visited PHIs.
699 2018-05-17 Richard Biener <rguenther@suse.de>
701 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
703 2018-05-17 Bin Cheng <bin.cheng@arm.com>
704 Richard Biener <rguenther@suse.de>
706 PR tree-optimization/85793
707 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
708 for VMAT_ELEMENTWISE.
710 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
712 * internal-fn.h (lookup_internal_fn): Declare
713 * internal-fn.c (lookup_internal_fn): New function.
714 * gimple.c (gimple_build_call_from_tree): Handle calls to
716 * gimple-pretty-print.c (dump_gimple_call): Print "." before
717 internal function names.
718 * tree-pretty-print.c (dump_generic_node): Likewise.
719 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
721 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
723 * gimple-fold.h (gimple_build): Make the function forms take
724 combined_fn rather than built_in_function.
725 (gimple_simplify): Likewise.
726 * gimple-match-head.c (gimple_simplify): Likewise.
727 * gimple-fold.c (gimple_build): Likewise.
728 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
729 rather than gimple_build_call_internal.
730 (get_initial_defs_for_reduction): Likewise.
731 (vect_create_epilog_for_reduction): Likewise.
732 (vectorizable_live_operation): Likewise.
734 2018-05-17 Martin Liska <mliska@suse.cz>
736 * gimple-ssa-sprintf.c (format_directive): Do not use
737 space in between 'G_' and '('.
739 2018-05-17 Jakub Jelinek <jakub@redhat.com>
742 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
743 even if the mask is not all ones.
746 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
748 (ix86_gimple_fold_builtin): Likewise.
751 * config/i386/i386.c: Include tree-vector-builder.h.
752 (ix86_vector_shift_count): New function.
753 (ix86_fold_builtin): Fold shift builtins by scalar count.
754 (ix86_gimple_fold_builtin): Likewise.
756 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
757 _mm512_setzero): New intrinsics.
759 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
760 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
762 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
763 code generation for cases where splatting a value is not useful.
764 * simplify-rtx.c (simplify_ternary_operation): Simplify
765 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
766 mode to a vec_concat.
768 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
770 * config.gcc: Support "goldmont-plus".
771 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
773 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
774 PROCESSOR_GOLDMONT_PLUS.
775 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
776 (processor_target_table): Add "goldmont-plus".
777 (PTA_GOLDMONT_PLUS): Define.
778 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
779 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
780 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
781 (fold_builtin_cpu): Add "goldmont-plus".
782 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
783 (ix86_option_override_internal): Add "goldmont-plus".
784 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
785 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
786 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
787 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
789 2018-05-17 Richard Biener <rguenther@suse.de>
791 PR tree-optimization/85757
792 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
793 remove defs that only feed that PHI from further processing.
795 2018-05-16 Jim Wilson <jimw@sifive.com>
797 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
799 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
801 2018-05-16 Mark Wielaard <mark@klomp.org>
803 * dwarf2out.c (count_index_strings): New function.
804 (output_indirect_strings): Call count_index_strings and generate
805 header for dwarf_version >= 5.
807 2018-05-16 Mark Wielaard <mark@klomp.org>
809 * dwarf2out.c (dwarf_FORM): New function.
810 (set_indirect_string): Use dwarf_FORM.
811 (reset_indirect_string): Likewise.
812 (size_of_die): Likewise.
813 (value_format): Likewise.
814 (output_die): Likewise.
815 (add_skeleton_AT_string): Likewise.
816 (output_macinfo_op): Likewise.
817 (index_string): Likewise.
818 (output_index_string_offset): Likewise.
819 (output_index_string): Likewise.
820 (count_index_strings): Likewise.
822 2018-05-16 Carl Love <cel@us.ibm.com>
824 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
825 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
827 2018-05-16 Martin Jambor <mjambor@suse.cz>
829 * ipa-prop.c (ipa_free_all_edge_args): Remove.
830 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
832 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
834 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
835 (fnma<mode>4): Likewise.
836 (fms<mode>4): Likewise.
837 (fnms<mode>4): Likewise.
838 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
839 (aarch64_fnma<mode>4): Likewise.
840 (aarch64_fms<mode>4): Likewise.
841 (aarch64_fnms<mode>4): Likewise.
842 (aarch64_fnmadd<mode>4): Likewise.
844 2018-05-16 Jason Merrill <jason@redhat.com>
846 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
848 2018-05-16 Richard Biener <rguenther@suse.de>
850 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
851 (dump_stmt_cost): Declare.
852 (add_stmt_cost): Dump cost we add.
853 (add_stmt_costs): New function.
854 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
856 (vect_analyze_stmt): Adjust prototype.
857 (vectorizable_condition): Likewise.
858 (vectorizable_live_operation): Likewise.
859 (vectorizable_reduction): Likewise.
860 (vectorizable_induction): Likewise.
861 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
862 cost vector to pass to vectorizable_ and record afterwards.
863 (vect_model_reduction_cost): Take cost vector argument and adjust.
864 (vect_model_induction_cost): Likewise.
865 (vectorizable_reduction): Likewise.
866 (vectorizable_induction): Likewise.
867 (vectorizable_live_operation): Likewise.
868 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
869 SLP_TREE_NUMBER_OF_VEC_STMTS.
870 (vect_analyze_slp_cost_1): Remove.
871 (vect_analyze_slp_cost): Likewise.
872 (vect_slp_analyze_node_operations): Take visited args and
873 a target cost vector. Avoid processing already visited stmt sets.
874 (vect_slp_analyze_operations): Use a local cost vector to gather
875 costs and register those of non-discarded instances.
876 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
877 (vect_schedule_slp_instance): Remove copying of
878 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
880 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
881 adding cost. Record cost entry location.
882 (vect_prologue_cost_for_slp_op): Function to compute cost of
883 a constant or invariant generated for SLP vect in the prologue,
884 split out from vect_analyze_slp_cost_1.
885 (vect_model_simple_cost): Make static. Adjust for SLP costing.
886 (vect_model_promotion_demotion_cost): Likewise.
887 (vect_model_store_cost): Likewise, make static.
888 (vect_model_load_cost): Likewise.
889 (vectorizable_bswap): Add cost vector arg and adjust.
890 (vectorizable_call): Likewise.
891 (vectorizable_simd_clone_call): Likewise.
892 (vectorizable_conversion): Likewise.
893 (vectorizable_assignment): Likewise.
894 (vectorizable_shift): Likewise.
895 (vectorizable_operation): Likewise.
896 (vectorizable_store): Likewise.
897 (vectorizable_load): Likewise.
898 (vectorizable_condition): Likewise.
899 (vectorizable_comparison): Likewise.
900 (can_vectorize_live_stmts): Likewise.
901 (vect_analyze_stmt): Likewise.
902 (vect_transform_stmt): Adjust calls to vectorizable_*.
903 * tree-vectorizer.c: Include gimple-pretty-print.h.
904 (dump_stmt_cost): New function.
906 2018-05-16 Richard Biener <rguenther@suse.de>
908 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
909 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
910 * tree-ssa-dse.c: Include tree-ssa-loop.h.
911 (check_name): New callback.
912 (dse_classify_store): Track cycles via a visited bitmap of PHI
913 defs and simplify handling of in-loop and across loop dead stores
914 and properly fail for loop-variant refs. Handle byte-tracking with
915 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
918 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
920 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
921 (vect_get_mask_type_for_stmt): Likewise.
922 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
924 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
925 to determine the statement's vector type and the vector type that
926 should be used for calculating nunits. Deal with cases in which
927 the type has to be deferred.
928 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
929 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
930 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
931 (vect_determine_vf_for_stmt): New functions, split out from...
932 (vect_determine_vectorization_factor): ...here.
933 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
934 (vect_get_mask_type_for_stmt): New functions, split out from
935 vect_determine_vectorization_factor.
937 2018-05-16 Richard Biener <rguenther@suse.de>
939 * tree-cfg.c (verify_gimple_assign_ternary): Properly
940 verify the [VEC_]COND_EXPR embedded comparison.
942 2018-05-15 Martin Sebor <msebor@redhat.com>
944 PR tree-optimization/85753
945 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
946 RECORD_TYPE in addition to ARRAY_TYPE.
948 2018-05-15 Martin Sebor <msebor@redhat.com>
951 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
953 2018-05-15 Richard Biener <rguenther@suse.de>
955 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
956 add by_clobber_p one. Change algorithm to collect all defs
957 representing uses we need to walk and try reducing them to
958 a single one before failing.
959 (dse_dom_walker::dse_optimize_stmt): Adjust.
961 2018-05-13 Mark Wielaard <mark@klomp.org>
963 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
964 (size_of_loc_descr): Likewise.
965 (output_loc_operands): Likewise.
966 (output_loc_operands_raw): Likewise.
967 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
968 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
969 (hash_loc_operands): Likewise.
970 (compare_loc_operands): Likewise.
972 2018-05-14 Mark Wielaard <mark@klomp.org>
974 * dwarf2out.c (count_index_addrs): New function.
975 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
977 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
979 PR tree-optimization/83648
980 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
981 return value as malloc candidate.
983 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
986 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
987 as true in call to suggest_attribute.
989 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
991 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
994 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
996 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
998 (aarch64_crypto_aesd_fused): Likewise.
1000 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1002 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1003 (movsi_aarch64): Likewise.
1004 (load_pairsi): Likewise.
1005 (load_pairdi): Likewise.
1006 (store_pairsi): Likewise.
1007 (store_pairdi): Likewise.
1008 (load_pairsf): Likewise.
1009 (load_pairdf): Likewise.
1010 (store_pairsf): Likewise.
1011 (store_pairdf): Likewise.
1012 (zero_extend): Likewise.
1013 (trunc): Swap alternatives.
1014 (fcvt_target): Add '?' to prefer w over r.
1016 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1019 * config/i386/i386.md: Disallow non-commutative arithmetics in
1020 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1021 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1022 in the peephole2 before it.
1024 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1026 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1027 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1028 (ix86_handle_option): Handle -mcldemote.
1029 * config.gcc: New header.
1030 * config/i386/cldemoteintrin.h: New file.
1031 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1032 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1034 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1035 OPTION_MASK_ISA_CLDEMOTE.
1036 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1037 (ix86_valid_target_attribute_inner_p): Ditto.
1038 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1039 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1040 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1041 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1042 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1044 * config/i386/i386.opt: Add -mcldemote.
1045 * config/i386/x86intrin.h: New header.
1046 * doc/invoke.texi: Add -mcldemote.
1048 2018-05-14 Richard Biener <rguenther@suse.de>
1050 * doc/match-and-simplify.texi: Adjust :s documentation.
1052 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1054 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1055 intended memcpy size.
1056 (REORDER_45): Likewise.
1058 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1060 * sort.cc: New file.
1061 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1062 * vec.c (qsort_chk): Use gcc_qsort.
1063 * Makefile.in (OBJS-libcommon): Add sort.o.
1064 (build/sort.o): New target. Use it...
1065 (BUILD_RTL): ... here, and...
1066 (build/gencfn-macros): ... here, and...
1067 (build/genmatch): ... here.
1069 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1070 Chung-Ju Wu <jasonwucj@gmail.com>
1072 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1073 * config/nds32/nds32-graywolf.md: New file.
1074 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1075 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1077 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1078 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1079 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1080 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1081 * config/nds32/pipelines.md: Include n15 settings.
1083 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1084 Chung-Ju Wu <jasonwucj@gmail.com>
1086 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1087 * config/nds32/nds32-n13.md: New file.
1088 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1089 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1091 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1092 * config/nds32/nds32.md (pipeline_model): Add n13.
1093 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1094 * config/nds32/pipelines.md: Include n13 settings.
1096 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1097 Chung-Ju Wu <jasonwucj@gmail.com>
1099 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1100 * config/nds32/nds32-n10.md: New file.
1101 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1102 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1104 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1105 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1106 * config/nds32/nds32.md (pipeline_model): Add n10.
1107 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1108 * config/nds32/pipelines.md: Include n10 settings.
1110 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1111 Kito Cheng <kito.cheng@gmail.com>
1112 Chung-Ju Wu <jasonwucj@gmail.com>
1114 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1115 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1116 Add enum values for DSP extension instructions.
1117 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1119 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1120 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1122 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1123 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1124 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1125 * config/nds32/nds32-intrinsic.md: Likewise.
1126 * config/nds32/nds32_intrinsic.h: Likewise.
1127 * config/nds32/nds32-md-auxiliary.c: Likewise.
1128 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1129 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1130 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1131 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1132 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1133 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1134 TYPE_DMAC in switch statement.
1135 * config/nds32/nds32.c: New checking and implementation for DSP
1136 extension instructions.
1137 * config/nds32/nds32.h: Likewise.
1138 * config/nds32/nds32.md: Likewise.
1139 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1140 * config/nds32/predicates.md: Implement new predicates for DSP
1143 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1145 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1146 Reformat alternatives and attributes so it is easier to identify
1147 which constraints/attributes go with which instruction.
1148 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1149 (mov<mode>_softfloat32, FMOVE64): Likewise.
1150 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1151 (mov<mode>_softfloat64, FMOVE64): Likewise.
1153 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1155 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1157 (Basic PowerPC Built-in Functions): The new name of the
1158 subsection previously known as "PowerPC Built-in Functions".
1159 (Basic PowerPC Built-in Functions Available on all Configurations):
1161 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1162 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1163 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1164 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1166 2018-05-11 Martin Jambor <mjambor@suse.cz>
1169 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1172 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1175 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1177 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1179 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1180 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1181 (ix86_handle_option): Handle -mwaitpkg.
1182 * config.gcc: New header.
1183 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1184 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1185 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1187 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1188 OPTION_MASK_ISA_WAITPKG.
1189 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1190 (ix86_option_override_internal): Add PTA_WAITPKG.
1191 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1192 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1193 IX86_BUILTIN_TPAUSE.
1194 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1195 __builtin_ia32_umwait and __builtin_ia32_tpause.
1196 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1197 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1198 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1199 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1200 UNSPECV_TPAUSE): New.
1201 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1202 * config/i386/i386.opt: Add -mwaitpkg.
1203 * config/i386/waitpkgintrin.h: New file.
1204 * config/i386/x86intrin.h: New header.
1205 * doc/invoke.texi: Add -mwaitpkg.
1207 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1210 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1212 (cortex-m0): Use armv6s-m isa.
1213 (cortex-m0plus): Likewise.
1214 (cortex-m1): Likewise.
1215 (cortex-m0.small-multiply): Likewise.
1216 (cortex-m0plus.small-multiply): Likewise.
1217 (cortex-m1.small-multiply): Likewise.
1219 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1220 Jakub Jelinek <jakub@redhat.com>
1222 PR tree-optimization/85692
1223 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1224 source permute as well.
1226 2018-05-11 Martin Liska <mliska@suse.cz>
1229 * doc/extend.texi: Document LLVM style format for no_sanitize
1232 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1234 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1235 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1236 (mode_supports_vsx_dform_quad): Likewise.
1237 (mode_supports_vmx_dform): Move these functions to be next to the
1238 other mode_supports functions.
1239 (mode_supports_dq_form): Likewise.
1240 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1241 mode_supports_dq_form.
1242 (reg_offset_addressing_ok_p): Likewise.
1243 (offsettable_ok_by_alignment): Likewise.
1244 (rs6000_legitimate_offset_address_p): Likewise.
1245 (legitimate_lo_sum_address_p): Likewise.
1246 (rs6000_legitimize_address): Likewise.
1247 (rs6000_legitimize_reload_address): Likewise.
1248 (rs6000_secondary_reload_inner): Likewise.
1249 (rs6000_preferred_reload_class): Likewise.
1250 (rs6000_output_move_128bit): Likewise.
1252 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1254 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1255 Generate SImode target register for null target.
1256 <case IX86_BUILTIN_XGETBV>: Ditto.
1257 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1258 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1260 2018-05-10 Carl Love <cel@us.ibm.com>
1262 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1263 dcbtt and dcbtstt if operands[2] is 0.
1265 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1268 * config/i386/sse.md (usadv64qi): New expander.
1270 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1272 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1273 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1274 -maltivec=be support.
1275 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1276 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1277 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1278 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1279 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1280 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1281 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1282 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1283 altivec_vsumsws): Adjust.
1284 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1285 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1286 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1288 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1289 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1290 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1291 (altivec_lve<VI_char>x): Delete expand.
1292 (*altivec_lve<VI_char>x_internal): Rename to...
1293 (altivec_lve<VI_char>x): ... this.
1294 (altivec_lvxl_<mode>): Delete expand.
1295 (*altivec_lvxl_<mode>_internal): Rename to ...
1296 (altivec_lvxl_<mode>): ... this.
1297 (altivec_stvxl_<mode>): Delete expand.
1298 (*altivec_stvxl_<mode>_internal): Rename to ...
1299 (altivec_stvxl_<mode>): ... this.
1300 (altivec_stve<VI_char>x): Delete expand.
1301 (*altivec_stve<VI_char>x_internal): Rename to ...
1302 (altivec_stve<VI_char>x): ... this.
1303 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1304 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1305 reduc_plus_scal_<mode>): Adjust.
1306 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1308 (rs6000_cpu_cpp_builtins): Adjust.
1309 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1310 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1311 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1312 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1313 -maltivec=be support.
1314 (rs6000_split_vec_extract_var): Adjust.
1315 (rs6000_split_v4si_init): Adjust.
1316 (swap_selector_for_mode): Delete.
1317 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1318 altivec_expand_stvex_be): Delete.
1319 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1320 -maltivec=be support.
1321 (rs6000_gimple_fold_builtin): Ditto.
1322 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1324 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1325 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1326 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1327 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1328 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1329 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1330 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1331 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1332 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1333 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1334 anonymous split): Adjust.
1335 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1336 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1338 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1340 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1341 when --with-gxx-include-dir is also specified.
1342 * configure: Regenerate.
1344 2018-05-09 Jim Wilson <jimw@sifive.com>
1347 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1348 * config/riscv/t-withmultilib: New.
1349 * config/riscv/withmultilib.h: New.
1350 * doc/install.texi: Document RISC-V --with-multilib-list support.
1352 2018-05-09 Richard Biener <rguenther@suse.de>
1354 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1356 (vect_bb_vectorization_profitable_p): Adjust. Compute
1357 actual scalar cost using the cost vector and the add_stmt_cost
1360 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1362 PR rtl-optimization/85645
1363 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1364 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1366 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1368 PR rtl-optimization/85645
1369 * shrink-wrap.c (spread_components): Return a boolean saying if
1370 anything was changed.
1371 (try_shrink_wrapping_separate): Iterate spread_components until
1372 nothing changes anymore.
1374 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1376 PR rtl-optimization/85645
1377 * regrename.c (build_def_use): Also kill the chains that include the
1378 destination of a REG_CFA_REGISTER note.
1380 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1382 PR rtl-optimization/85645
1383 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1384 insn that has a REG_CFA_REGISTER note.
1386 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1388 * cfgexpand.c (expand_clobber): New function.
1389 (expand_gimple_stmt_1): Use it.
1390 * tree-vect-stmts.c (vect_clobber_variable): New function,
1392 (vectorizable_simd_clone_call): ...here.
1393 (vectorizable_store): Emit a clobber either side of an
1394 IFN_STORE_LANES sequence.
1395 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1397 2018-05-09 Tom de Vries <tom@codesourcery.com>
1400 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1401 (define_insn "trap_if_false"): Add exit after trap.
1403 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1405 PR rtl-optimization/85638
1406 * bb-reorder.c: Include common/common-target.h.
1407 (create_forwarder_block): New function extracted from...
1408 (fix_up_crossing_landing_pad): ...here. Rename into...
1409 (dw2_fix_up_crossing_landing_pad): ...this.
1410 (sjlj_fix_up_crossing_landing_pad): New function.
1411 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1412 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1413 from both partitions and exit the loop after one iteration.
1415 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1418 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1420 (Basic PowerPC Built-in Functions): The new name of the
1421 subsection previously known as "PowerPC Built-in Functions".
1422 (Basic PowerPC Built-in Functions Available on all Configurations):
1424 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1426 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1428 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1430 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1433 2018-05-08 Jim Wilson <jimw@sifive.com>
1435 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1436 (LD_EMUL_SUFFIX): New.
1437 (LINK_SPEC): Use it.
1439 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1441 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1443 (Basic PowerPC Built-in Functions): The new name of the
1444 subsection previously known as "PowerPC Built-in Functions".
1445 (Basic PowerPC Built-in Functions Available on all Configurations):
1447 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1449 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1451 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1453 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1456 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1459 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1460 after cmpelim optimization.
1462 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1464 * config.gcc: Support "goldmont".
1465 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1466 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1468 * config/i386/i386.c (m_GOLDMONT): Define.
1469 (processor_target_table): Add "goldmont".
1470 (PTA_GOLDMONT): Define.
1471 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1472 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1473 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1474 (fold_builtin_cpu): Add "goldmont".
1475 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1476 (ix86_option_override_internal): Add "goldmont".
1477 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1478 (processor_type): Add PROCESSOR_GOLDMONT.
1479 * config/i386/i386.md: Add CPU "glm".
1480 * config/i386/glm.md: New file.
1481 * config/i386/x86-tune.def: Add m_GOLDMONT.
1482 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1484 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1487 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1489 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1490 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1491 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1494 * config/i386/i386.c (ix86_fold_builtin): Handle
1495 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1498 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1499 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1501 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1504 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1505 (check_arch): Likewise.
1506 (check_fpu): Return the result rather than printing it.
1507 (end arch): Fix operator precedence.
1508 (end cpu): Likewise.
1509 (END): Print the result from check_fpu.
1511 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1512 Alan Hayward <alan.hayward@arm.com>
1513 David Sherwood <david.sherwood@arm.com>
1515 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1516 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1517 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1518 (*fcmuo<mode>_and): New patterns.
1520 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1522 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1523 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1524 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1525 (cmp_op, sve_imm_con): New code attributes.
1526 (SVE_COND_INT_CMP, imm_con): Delete.
1527 (cmp_op): Remove above unspecs from int attribute.
1528 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1530 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1531 comparison-specific unspecs.
1532 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1533 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1534 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1535 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1536 (*vec_fcm<cmp_op><mode>): Rename to...
1537 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1538 (*vec_fcmuo<mode>): Rename to...
1539 (*fcmuo<mode>): ...this and adjust likewise.
1540 (*pred_fcm<cmp_op><mode>): New pattern.
1541 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1542 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1544 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1546 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1547 (aarch64_emit_sve_predicated_cond): New function.
1548 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1549 (aarch64_emit_unspec_cond_or): Replace with...
1550 (aarch64_emit_sve_or_conds): ...this new function. Use
1551 aarch64_emit_sve_ptrue_op for the individual comparisons and
1552 aarch64_emit_binop to OR them together.
1553 (aarch64_emit_inverted_unspec_cond): Replace with...
1554 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1555 aarch64_emit_sve_ptrue_op for the comparison and
1556 aarch64_emit_unop to invert the result.
1557 (aarch64_expand_sve_vec_cmp_float): Update after the above
1558 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1560 2018-05-07 Nathan Sidwell <nathan@acm.org>
1562 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1563 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1564 (Backwards Compatibility): Likewise.
1566 2018-05-07 Luis Machado <luis.machado@linaro.org>
1570 2018-05-07 Luis Machado <luis.machado@linaro.org>
1572 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1573 <prefetch_dynamic_strides>: New const bool field.
1574 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1575 prefetch_dynamic_strides.
1576 (exynosm1_prefetch_tune): Likewise.
1577 (thunderxt88_prefetch_tune): Likewise.
1578 (thunderx_prefetch_tune): Likewise.
1579 (thunderx2t99_prefetch_tune): Likewise.
1580 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1581 (aarch64_override_options_internal): Update to set
1582 PARAM_PREFETCH_DYNAMIC_STRIDES.
1583 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1584 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1585 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1586 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1587 prefetch-dynamic-strides setting.
1589 2018-05-07 Luis Machado <luis.machado@linaro.org>
1591 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1592 <minimum_stride>: New const int field.
1593 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1594 minimum_stride field.
1595 (exynosm1_prefetch_tune): Likewise.
1596 (thunderxt88_prefetch_tune): Likewise.
1597 (thunderx_prefetch_tune): Likewise.
1598 (thunderx2t99_prefetch_tune): Likewise.
1599 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1600 (aarch64_override_options_internal): Update to set
1601 PARAM_PREFETCH_MINIMUM_STRIDE.
1602 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1603 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1604 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1605 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1606 stride is constant and is below the minimum stride threshold.
1608 2018-05-07 Luis Machado <luis.machado@linaro.org>
1610 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1613 2018-05-07 Luis Machado <luis.machado@linaro.org>
1615 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1616 <prefetch_dynamic_strides>: New const bool field.
1617 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1618 prefetch_dynamic_strides.
1619 (exynosm1_prefetch_tune): Likewise.
1620 (thunderxt88_prefetch_tune): Likewise.
1621 (thunderx_prefetch_tune): Likewise.
1622 (thunderx2t99_prefetch_tune): Likewise.
1623 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1624 (aarch64_override_options_internal): Update to set
1625 PARAM_PREFETCH_DYNAMIC_STRIDES.
1626 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1627 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1628 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1629 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1630 prefetch-dynamic-strides setting.
1632 2018-05-07 Luis Machado <luis.machado@linaro.org>
1634 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1635 <minimum_stride>: New const int field.
1636 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1637 minimum_stride field.
1638 (exynosm1_prefetch_tune): Likewise.
1639 (thunderxt88_prefetch_tune): Likewise.
1640 (thunderx_prefetch_tune): Likewise.
1641 (thunderx2t99_prefetch_tune): Likewise.
1642 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1643 (aarch64_override_options_internal): Update to set
1644 PARAM_PREFETCH_MINIMUM_STRIDE.
1645 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1646 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1647 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1648 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1649 stride is constant and is below the minimum stride threshold.
1651 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1654 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1655 the type is addressable. Don't force op into register if it has
1658 2018-05-05 Roland McGrath <mcgrathr@google.com>
1661 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1662 any section for which we don't know a specific type it should have,
1663 regardless of name. Previously this was done only for the exact
1664 names ".init_array", ".fini_array", and ".preinit_array".
1665 (default_elf_asm_named_section): Add comment about
1666 relationship with default_section_type_flags and SECTION_NOTYPE.
1667 (get_section): Don't consider it a type conflict if one side has
1668 SECTION_NOTYPE and the other doesn't, as long as neither has the
1669 SECTION_BSS et al used in the default_section_type_flags logic.
1671 2018-05-05 Tom de Vries <tom@codesourcery.com>
1674 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1675 (workaround_barsyncs): New function.
1676 (nvptx_reorg): Use workaround_barsyncs.
1677 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1678 (define_expand "nvptx_membar_cta"): New define_expand.
1679 (define_insn "*nvptx_membar_cta"): New insn.
1681 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1683 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1684 To improve optimization opportunities.
1685 * builtin-types.def: The new needed builtin types for the above.
1687 2018-05-04 Richard Biener <rguenther@suse.de>
1689 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1690 * gimple-ssa-store-merging.c
1691 (imm_store_chain_info::output_merged_store): Remove redundant create,
1692 release split_store vector contents on failure.
1693 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1694 scalar stmt vector on cache hit.
1696 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1698 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1700 * config.gcc (powerpc-xilinx-eabi*): Remove.
1701 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1703 (fusion_addis_mem_combo_load): Ditto.
1704 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1706 (rs6000_cpu_cpp_builtins): Ditto.
1707 * config/rs6000/rs6000-linux.c
1708 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1709 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1710 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1712 (rs6000_setup_reg_addr_masks): Ditto.
1713 (rs6000_init_hard_regno_mode_ok): Ditto.
1714 (rs6000_option_override_internal): Ditto.
1715 (legitimate_lo_sum_address_p): Ditto.
1716 (rs6000_legitimize_address): Ditto.
1717 (rs6000_legitimize_reload_address): Ditto.
1718 (rs6000_legitimate_address_p): Ditto.
1719 (abi_v4_pass_in_fpr): Ditto.
1720 (setup_incoming_varargs): Ditto.
1721 (rs6000_gimplify_va_arg): Ditto.
1722 (rs6000_split_multireg_move): Ditto.
1723 (rs6000_savres_strategy): Ditto.
1724 (rs6000_emit_prologue_components): Ditto.
1725 (rs6000_emit_epilogue_components): Ditto.
1726 (rs6000_emit_prologue): Ditto.
1727 (rs6000_emit_epilogue): Ditto.
1728 (rs6000_elf_file_end): Ditto.
1729 (rs6000_function_value): Ditto.
1730 (rs6000_libcall_value): Ditto.
1731 * config/rs6000/rs6000.h: Ditto.
1732 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1733 (TARGET_MINMAX): ... this. New.
1734 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1735 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1736 (*movsi_internal1_single): Delete.
1737 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1738 mfpu=, mxilinx-fpu): Delete.
1739 * config/rs6000/singlefp.h: Delete.
1740 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1741 * config/rs6000/t-rs6000: Ditto.
1742 * config/rs6000/t-xilinx: Delete.
1743 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1744 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1745 (VStype_simple): Delete.
1746 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1747 * config/rs6000/xfpu.h: Delete.
1748 * config/rs6000/xfpu.md: Delete.
1749 * config/rs6000/xilinx.h: Delete.
1750 * config/rs6000/xilinx.opt: Delete.
1751 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1752 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1754 2018-05-04 Tom de Vries <tom@codesourcery.com>
1757 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1760 2018-05-04 Richard Biener <rguenther@suse.de>
1763 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1764 (expand_complex_div_wide): Likewise.
1765 (expand_complex_operations_1): Likewise.
1766 (expand_complex_libcall): Preserve EH info of the original stmt.
1767 (tree_lower_complex): Handle removed blocks.
1768 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1769 on complex multiplication and division libcall builtins.
1771 2018-05-04 Richard Biener <rguenther@suse.de>
1774 * fold-const.c (negate_expr_p): Restrict negation of operand
1775 zero of a division to when we know that can happen without
1777 (fold_negate_expr_1): Likewise.
1779 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1782 * real.h (real_nextafter): Declare.
1783 * real.c (real_nextafter): New function.
1784 * fold-const-call.c (fold_const_nextafter): New function.
1785 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1786 CASE_CFN_NEXTTOWARD.
1787 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1788 even when arg1_mode is different from arg0_mode.
1790 2018-05-03 Nathan Sidwell <nathan@acm.org>
1792 * doc/extend.texi (Deprecated Features): Remove
1794 (Backwards Compatibility): Likewise.
1795 * doc/invoke.texi (C++ Language Options): Likewise.
1796 (C++ Dialect Options): Likewise.
1798 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1801 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1802 _mm512_mask_mullox_epi64): New intrinsics.
1804 2018-05-03 Tom de Vries <tom@codesourcery.com>
1807 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1808 dump files): Add offload-tree.
1810 2018-05-03 Richard Biener <rguenther@suse.de>
1812 PR tree-optimization/85615
1813 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1814 to loops not nested in BBs loop father to avoid creating multi-entry
1817 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1819 PR tree-optimization/70291
1820 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1821 arguments. Change return type to tree. Emit libcall as a new
1822 statement rather than replacing existing one when inplace_p is true.
1823 (expand_complex_multiplication_components): New function.
1824 (expand_complex_multiplication): Expand floating-point complex
1825 multiplication using the above.
1826 (expand_complex_division): Rename inner_type parameter to type.
1827 Update expand_complex_libcall call-site.
1828 (expand_complex_operations_1): Update expand_complex_multiplication
1829 and expand_complex_division call-sites.
1831 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1834 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1835 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1836 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1837 the highest significant bit of the shift count mask is clear. In
1838 check whether and[sq]i3 is needed verify that all significant bits
1839 of the shift count other than the highest are set.
1841 2018-05-02 Tom de Vries <tom@codesourcery.com>
1844 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1845 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1846 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1847 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1848 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1849 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1850 __builtin_goacc_parlevel_size.
1852 2018-05-02 Richard Biener <rguenther@suse.de>
1854 PR tree-optimization/85597
1855 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1856 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1859 2018-05-02 Tom de Vries <tom@codesourcery.com>
1862 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1863 dump files): Add ltrans-tree.
1865 2018-05-02 Tom de Vries <tom@codesourcery.com>
1868 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1869 dump files): Add wpa-ipa.
1871 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1873 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1874 powerpc*-*-linux*paired* target.
1875 * config/rs6000/750cl.h: Delete.
1876 * config/rs6000/paired.h: Delete.
1877 * config/rs6000/paired.md: Delete.
1878 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1880 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1881 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1882 comment. Remove paired float support.
1883 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1884 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1886 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1887 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1889 * config/rs6000/rs6000.c: Remove paired float support.
1890 (paired_expand_vector_init, paired_expand_vector_move,
1891 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1892 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1893 paired_expand_builtin, paired_expand_predicate_builtin,
1894 paired_init_builtins): Delete.
1895 * config/rs6000/rs6000.h: Remove paired float support.
1896 * config/rs6000/rs6000.md: Remove paired float support.
1897 (move_from_CR_ov_bit): Delete.
1898 * config/rs6000/rs6000.opt (mpaired): Delete.
1899 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1900 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1902 2018-05-02 Richard Biener <rguenther@suse.de>
1905 * gimplify.c (gimplify_save_expr): When in SSA form allow
1906 SAVE_EXPRs to compute to SSA vars.
1908 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1911 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1912 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1913 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1914 clobber operands[2], instead use a new pseudo. Formatting fixes.
1916 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1918 PR tree-optimization/85586
1919 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1920 exit early for statements in the same group if the accesses are
1923 2018-05-02 Tom de Vries <tom@codesourcery.com>
1926 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1929 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1931 PR tree-optimization/85143
1932 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1934 2018-05-01 Tom de Vries <tom@codesourcery.com>
1937 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1938 not found" error message.
1940 2018-05-01 Tom de Vries <tom@codesourcery.com>
1943 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1944 * vec.c (test_ordered_remove_if): New function.
1945 (vec_c_tests): Call test_ordered_remove_if.
1946 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1947 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1948 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1949 VEC_ORDERED_REMOVE_IF.
1951 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1953 PR tree-optimization/82665
1954 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1955 pointer subtraction where arguments come from a memchr call.
1957 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1959 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1960 --push-state --as-needed and --pop-state instead of --as-needed and
1961 --no-as-needed if ld supports it.
1962 * configure: Regenerated.
1965 * doc/install.texi2html: Replace _002d with - and _002a with * in
1966 generated html files using sed.
1968 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1971 * gcc-rich-location.c (blank_line_before_p): New function.
1972 (use_new_line): New function.
1973 (gcc_rich_location::add_fixit_insert_formatted): New function.
1974 * gcc-rich-location.h
1975 (gcc_rich_location::add_fixit_insert_formatted): New function.
1977 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1979 * selftest.c (assert_streq): Rename "expected" and "actual" to
1980 "val1" and "val2". Extend NULL-handling to cover both inputs
1981 symmetrically, while still requiring both to be non-NULL for a pass.
1982 * selftest.h (assert_streq): Rename "expected" and "actual" to
1984 (ASSERT_EQ): Likewise.
1985 (ASSERT_EQ_AT): Likewise.
1986 (ASSERT_KNOWN_EQ): Likewise.
1987 (ASSERT_KNOWN_EQ_AT): Likewise.
1988 (ASSERT_NE): Likewise.
1989 (ASSERT_MAYBE_NE): Likewise.
1990 (ASSERT_MAYBE_NE_AT): Likewise.
1991 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1992 the assertion to pass.
1993 (ASSERT_STREQ_AT): Likewise.
1995 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1997 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1998 interaction with -pie.
2000 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2002 * selftest.h: Fix alphabetization of per-source-file selftest
2005 2018-04-30 Jason Merrill <jason@redhat.com>
2007 PR c++/61982 - dead stores to destroyed objects.
2008 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2011 2018-04-30 Jason Merrill <jason@redhat.com>
2013 * tree.c (build_clobber): New.
2014 * tree.h: Declare it.
2015 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2017 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2019 * diagnostic-show-locus.c (layout::layout): Update for
2020 location_get_source_line returning a char_span.
2021 (struct char_span): Move to input.h.
2022 (struct correction): Update for fields in char_span becoming
2024 (struct source_line): Update for location_get_source_line
2025 returning a char_span.
2026 (layout::print_line): Likewise.
2027 * edit-context.c (edited_file::print_content): Likewise.
2028 (edited_file::print_diff_hunk): Likewise.
2029 (edited_file::print_run_of_changed_lines): Likewise.
2030 (edited_file::get_num_lines): Likewise.
2031 (edited_line::edited_line): Likewise.
2032 * final.c (asm_show_source): Likewise.
2033 * input.c (location_get_source_line): Convert return type
2034 from const char * to char_span, losing the final "line_len"
2036 (dump_location_info): Update for the above.
2037 (get_substring_ranges_for_loc): Likewise. Use a char_span
2038 when handling the literal within the line.
2039 (test_reading_source_line): Update for location_get_source_line
2040 returning a char_span.
2041 * input.h (class char_span): Move here from
2042 diagnostic-show-locus.c, converting from a struct to a class.
2043 Make data members private.
2044 (char_span::operator bool): New.
2045 (char_span::length): New.
2046 (char_span::get_buffer): New.
2047 (char_span::operator[]): New.
2048 (char_span::subspan): Make const.
2049 (char_span::xstrdup): New.
2050 (location_get_source_line): Convert return type from const char *
2051 to char_span, losing the final "line_size" param.
2053 2018-04-30 Jan Hubicka <jh@suse.cz>
2055 * lto-wrapper.c (ltrans_priorities): New static var.
2056 (cmp_priority): New.
2057 (run_gcc): Read priorities and if doing parallel build order
2058 the Makefile by them.
2060 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2062 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2064 2018-04-30 Richard Biener <rguenther@suse.de>
2066 * tree-cfg.c (verify_address): Remove base argument, add
2067 flag whether to check TREE_ADDRESSABLE and do that.
2068 (verify_expr): Remove.
2069 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2070 (verify_gimple_assign_single): Likewise.
2071 (verify_gimple_switch): Likewise.
2072 (verify_expr_location_1): Dereference tp once. Add (disabled)
2073 piece from verify_expr.
2074 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2076 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2078 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2080 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2082 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2083 (small_data_pattern): Likewise.
2084 (arc_rewrite_small_data): Likewise.
2085 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2086 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2087 (get_symbol_alignment): New function.
2088 (legitimate_small_data_address_p): Likewise.
2089 (legitimate_scaled_address): Update, call
2090 legitimate_small_data_address_p.
2091 (output_sdata): New static variable.
2092 (arc_print_operand): Update how we handle small data operands.
2093 (arc_print_operand_address): Likewise.
2094 (arc_legitimate_address_p): Update, use
2095 legitimate_small_data_address_p.
2096 (arc_rewrite_small_data_p): Remove.
2097 (arc_rewrite_small_data_1): Likewise.
2098 (arc_rewrite_small_data): Likewise.
2099 (small_data_pattern): Likewise.
2100 (compact_sda_memory_operand): Update to use
2101 legitimate_small_data_address_p and get_symbol_alignment.
2102 (prepare_move_operands): Don't rewite sdata pattern.
2103 (prepare_extend_operands): Remove.
2104 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2106 (zero_extendqisi2): Likewise.
2107 (zero_extendhisi2): Likewise.
2108 (extendqihi2): Likewise.
2109 (extendqisi2): Likewise.
2110 (extendhisi2): Likewise.
2114 * config/arc/constraints.md (Usd): Change it to memory constraint.
2116 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2118 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2119 as source of std instructions.
2120 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2121 allow 6-bit constants as source for store instructions.
2122 (movdi_insn): Update instruction pattern to allow 6-bit constants
2123 as source for store instructions.
2125 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2127 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2129 2018-04-30 Nathan Sidwell <nathan@acm.org>
2130 Sandra Loosemore <sandra@codesourcery.com>
2132 * dumpfile.c (dump_open): Allow '-' for stdout.
2133 * doc/invoke.texi (Developer Options): Document dump filename
2134 determination early. Document stdin/stdout selection.
2136 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2138 Microblaze Target: PIC data text relative
2140 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2141 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2143 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2144 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2145 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2146 New addressing mode for data-text relative position indepenedent code.
2147 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2148 'ADDRESS_SYMBOLIC_TXT_REL'.
2149 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2150 (microblaze_legitimate_pic_operand): Exclude function calls from
2151 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2152 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2154 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2155 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2156 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2157 for 'address + offset'.
2158 (microblaze_expand_prologue): Add new function prologue call for
2160 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2161 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2162 table in case of TARGET_PIC_DATA_TEXT_REL.
2163 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2164 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2165 Add new macros 'UNSPEC_TEXT',
2166 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2167 + exclude function calls from 'UNSPEC_PLT' in case of data text
2169 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2170 new target hook for generating address diff vector tables in case of
2172 * doc/tm.texi : Regenerate.
2173 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2174 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2175 of addr diff vector generation.
2176 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2177 target hook definition.
2178 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2179 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2180 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2181 option for data text relative.
2183 2018-04-30 Richard Biener <rguenther@suse.de>
2185 * tree-chrec.h (evolution_function_is_constant_p): Remove
2187 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2189 2018-04-30 Richard Biener <rguenther@suse.de>
2192 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2194 2018-04-30 Richard Biener <rguenther@suse.de>
2196 PR tree-optimization/28364
2197 PR tree-optimization/85275
2198 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2199 copying first exit test.
2201 2018-04-28 Mark Wielaard <mark@klomp.org>
2203 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2205 (dwarf_AT): Handle DW_AT_addr_base.
2206 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2208 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2211 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2212 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2213 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2214 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2216 2018-04-28 Richard Biener <rguenther@suse.de>
2218 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2219 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2220 to reflect use. Only add interesting stmts.
2222 2018-04-27 Martin Jambor <mjambor@suse.cz>
2225 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2226 the jump function allows for passing through aggregate values.
2228 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2230 * input.h (in_system_header_at): Convert from macro to inline
2232 (from_macro_expansion_at): Likewise.
2233 (from_macro_definition_at): Likewise.
2235 2018-04-27 Jeff Law <law@redhat.com>
2237 * config.gcc: Mark tile* targets as deprecated/obsolete.
2239 2018-04-27 Richard Biener <rguenther@suse.de>
2241 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2244 2018-04-27 Richard Biener <rguenther@suse.de>
2246 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2248 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2250 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2251 with Yd constraint. Set "preferred_for_speed" attribute from
2252 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2254 (*movdi_internal): Ditto.
2255 (movti_interunit splitters): Remove
2256 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2257 (movdi_interunit splitters): Ditto.
2258 * config/i386/constraints.md (Ye): Remove.
2259 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2261 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2264 * config/aarch64/constraints.md (Usg): Limit to 31.
2267 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2269 PR tree-optimization/85529
2270 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2271 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2272 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2273 zero extension or masking of the MSB bit.
2274 (optimize_range_tests): Add FIRST_BB argument, pass it through
2275 to optimize_range_tests_var_bound.
2276 (maybe_optimize_range_tests, reassociate_bb): Adjust
2277 optimize_range_tests callers.
2279 2018-04-26 Richard Biener <rguenther@suse.de>
2280 Jakub Jelinek <jakub@redhat.com>
2282 * cgraph.h (symbol_table): Just declare debug method here.
2283 * symtab.c (symbol_table::debug): Define.
2285 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2287 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2289 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2291 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2292 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2293 (*movdi_internal): Substitute Yi and Yj constraint with x
2294 and Ym and Yn constraint with y constraint. Update "isa"
2295 attribute and set "preferred_for_speed" attribute from
2296 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2297 (*movsi_internal): Ditto.
2298 (*movdf_internal): Ditto.
2299 (*movsf_internal): Ditto.
2300 (*zero_extendsidi2): Ditto.
2301 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2302 (sse2_loadld): Ditto.
2303 (*vec_extract<ssevecmodelower>_0): Ditto.
2304 (*vec_extractv4si_0_zext_sse4): Ditto.
2305 (vec_concatv2di): Ditto.
2306 (*vec_dup<mode>): Ditto.
2307 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2308 * config/i386/constraints.md (Yi): Remove.
2313 2018-04-26 Nathan Sidwell <nathan@acm.org>
2315 * dumpfile.c (dump_open): New.
2316 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2317 (dump_finish): Detect stdio/stderr by value not name.
2319 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2321 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2323 2018-04-26 Tom de Vries <tom@codesourcery.com>
2326 * config/nvptx/nvptx.c (verify_neutering_jumps)
2327 (verify_neutering_labels): New function
2328 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2330 2018-04-26 Tom de Vries <tom@codesourcery.com>
2333 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2334 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2337 2018-04-26 Richard Biener <rguenther@suse.de>
2338 Tom de Vries <tom@codesourcery.com>
2341 * lto-streamer-out.c (output_function): Fixup loops if required to match
2342 discovery done in the reader.
2344 2018-04-26 Richard Biener <rguenther@suse.de>
2346 PR tree-optimization/85116
2347 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2348 have a loop exit from the single latch predecessor. Remove
2349 case of header with just condition.
2350 (ch_base::copy_headers): Exclude infinite loops from any
2352 (pass_ch::execute): Record exits.
2354 2018-04-26 Richard Biener <rguenther@suse.de>
2356 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2357 prologue cost vector and pass it to vect_get_load_cost.
2358 (vect_get_peeling_costs_all_drs): Likewise.
2359 (vect_peeling_hash_get_lowest_cost): Likewise.
2360 (vect_enhance_data_refs_alignment): Likewise.
2362 2018-04-26 Richard Biener <rguenther@suse.de>
2365 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2366 checking of integer<->pointer conversions.
2367 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2368 sign-/zero-extending pointer types.
2369 (expand_omp_for_static_chunk): Likewise.
2371 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2372 Jean Lee <xiaoyur347@gmail.com>
2374 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2375 (TARGET_ASAN_SHADOW_OFFSET): Define.
2376 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2377 true for -fsanitize=address.
2379 2018-04-25 Mark Wielaard <mark@klomp.org>
2381 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2384 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2386 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2387 than "alu", remove explicit "memory" and "imm_disp" attributes.
2388 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2391 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2392 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2393 gen_lowpart_no_emit.
2395 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2398 * config/i386/i386.c (ix86_expand_builtin): Change memory
2399 operand to XI, extend p0 to Pmode.
2400 * config/i386/i386.md: Change unspec volatile and operand
2401 1 mode to XI, change operand 0 mode to P.
2403 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2405 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2406 GET_MODE_MASK before any checking.
2407 (nds32_can_use_bset_p): Likewise.
2408 (nds32_can_use_btgl_p): Likewise.
2410 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2412 * config/nds32/nds32-doubleword.md: New define_split pattern for
2413 illegal register number.
2415 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2417 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2419 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2421 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2423 2018-04-25 Richard Biener <rguenther@suse.de>
2425 * lto-streamer.h (LTO_major_version): Bump to 8.
2427 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2429 * BASE-VER: Set to 9.0.0.
2431 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2433 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2434 in __abskf2 and __powikf2.
2436 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2439 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2440 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2441 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2442 Use the above on operand 2. Reindent.
2443 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2445 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2448 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2450 * config/i386/i386.opt (mcet): Removed.
2451 * doc/install.texi: Remove -mcet documentation.
2452 * doc/invoke.texi: Likewise.
2454 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2457 * doc/install.texi: Remove -mcet from bootstrap-cet.
2459 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2462 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2463 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2467 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2468 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2469 containing a CONST_VECTOR.
2471 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2473 * doc/install.texi: Update newlib dependency for nvptx.
2475 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2478 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2479 instead of INTVAL when shifting x left.
2481 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2483 PR tree-optimization/85478
2484 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2485 vect_grouped_store_supported for single element vectors.
2487 2018-04-24 Richard Biener <rguenther@suse.de>
2490 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2491 load cost increase to the case of non-constant step.
2493 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2496 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2497 destination if any_malformed_asm.
2499 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2502 * expr.c (store_field): In the bitfield case, if the value comes from
2503 a function call and is returned in registers by means of a PARALLEL,
2504 do not change the mode of the temporary unless BLKmode and VOIDmode.
2506 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2508 PR rtl-optimization/85423
2509 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2510 dependencies to debug insns when the previous insn is non-debug.
2512 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2514 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2515 enums into a single definition.
2516 (fls): Fix predicates and printing.
2519 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2521 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2522 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2523 and short u6 immediate.
2524 (check_if_valid_sleep_operand): Remove.
2525 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2527 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2529 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2530 flag_always_save_lp condition.
2531 * config/nds32/nds32.opt (malways-save-lp): New option.
2533 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2535 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2536 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2537 * config/nds32/nds32.h
2538 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2539 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2541 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2543 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2544 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2546 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2547 Chung-Ju Wu <jasonwucj@gmail.com>
2549 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2551 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2552 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2554 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2556 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2558 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2560 * config/nds32/nds32-protos.h (nds32_data_alignment,
2561 nds32_local_alignment): Declare.
2562 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2563 nds32_local_alignment): New functions.
2564 (TARGET_CONSTANT_ALIGNMENT): Define.
2565 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2567 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2569 * config/nds32/nds32.c
2570 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2571 (TARGET_MODES_TIEABLE_P): Likewise.
2573 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2575 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2578 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2579 Chung-Ju Wu <jasonwucj@gmail.com>
2581 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2582 for unaligned access.
2583 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2585 * config/nds32/nds32-intrinsic.md: Likewise.
2586 * config/nds32/nds32_intrinsic.h: Likewise.
2587 * config/nds32/nds32.h (nds32_builtins): Likewise.
2588 * config/nds32/nds32.opt (munaligned-access): New option.
2589 * config/nds32/nds32.c (nds32_asm_file_start): Display
2590 flag_unaligned_access status.
2592 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2594 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2595 -mno-relax is present.
2596 * config/riscv/linux.h (LINK_SPEC): Ditto.
2598 2018-04-20 Martin Sebor <msebor@redhat.com>
2601 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2603 (gimple_fold_builtin_stxcpy_chk): Same.
2604 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2606 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2609 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2610 __powikf2 when long double is IEEE 128-bit.
2612 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2614 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2615 step to make sure stack always aligned.
2617 2018-04-20 Carl Love <cel@us.ibm.com>
2620 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2621 size check for arg0.
2623 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2624 Tom de Vries <tom@codesourcery.com>
2627 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2628 Emit insns for calls too.
2629 (nvptx_find_par): Always look for worker-level predecessor insn.
2630 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2632 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2633 (nvptx_process_pars): Propagate frames for calls.
2635 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2638 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2640 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2641 (ix86_handle_option): Don't handle OPT_mibt.
2642 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2644 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2646 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2647 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2648 (ix86_target_macros): Define __CET__ with flag_cf_protection
2649 for -fcf-protection.
2650 * config/i386/i386.c (isa2_opts): Remove -mibt.
2651 * config/i386/i386.h (TARGET_IBT): Removed.
2652 (TARGET_IBT_P): Likewise.
2653 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2654 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2655 * config/i386/i386.opt (mcet): Update help message.
2658 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2659 -mcet as an alias for -mshstk.
2661 2018-04-20 Richard Biener <rguenther@suse.de>
2664 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2665 complexity by forcing a single use of the multiply operand.
2667 2018-04-20 Martin Jambor <mjambor@suse.cz>
2670 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2671 recursion dependency to only apply to non-clones.
2673 2018-04-20 Martin Jambor <mjambor@suse.cz>
2676 * ipa-cp.c (create_specialized_node): Check that clones of
2677 self-recursive edges exist during IPA-CP.
2679 2018-04-19 Toon Moene <toon@moene.org>
2681 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2684 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2686 PR tree-optimization/85467
2687 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2688 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2689 VECTOR_CST element to type.
2691 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2694 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2695 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2696 (builtin_longjmp): Likewise.
2697 (save_stack_nonlocal): New pattern.
2698 (restore_stack_nonlocal): Likewise.
2700 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2703 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2704 Replace ASM_OUTPUT_LABEL with fprintf.
2706 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2709 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2710 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2711 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2712 define __IBT__ and __SHSTK__ for -fcf-protection.
2713 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2715 (ix86_trampoline_init): Likewise.
2716 (x86_output_mi_thunk): Likewise.
2717 (ix86_notrack_prefixed_insn_p): Likewise.
2718 (ix86_option_override_internal): Don't disallow -fcf-protection.
2719 * config/i386/i386.md (rdssp<mode>): Also enable for
2721 (incssp<mode>): Likewise.
2722 (nop_endbr): Likewise.
2723 * config/i386/i386.opt (mcet): Change help message to built-in
2727 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2728 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2729 enable CET built-in functions.
2731 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2733 * common/config/i386/i386-common.c
2734 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2735 OPTION_MASK_ISA_MOVDIRI_UNSET,
2736 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2737 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2738 * config.gcc (movdirintrin.h): New header.
2739 * config/i386/cpuid.h (bit_MOVDIRI,
2740 bit_MOVDIR64B): New bits.
2741 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2743 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2744 (VOID, PVOID, PCVOID)): New function types.
2745 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2746 __builtin_ia32_directstoreu_u64,
2747 __builtin_ia32_movdir64b): New builtins.
2748 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2749 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2751 (ix86_valid_target_attribute_inner_p): Ditto.
2752 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2753 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2754 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2755 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2756 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2757 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2758 (movdiri<mode>, movdir64b_<mode>): New.
2759 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2760 * config/i386/immintrin.h: Include movdirintrin.h.
2761 * config/i386/movdirintrin.h: New file.
2762 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2764 2018-04-19 Richard Biener <rguenther@suse.de>
2767 * cfg.c (clear_bb_flags): When loop state says we have
2768 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2770 2018-04-19 Richard Biener <rguenther@suse.de>
2772 PR tree-optimization/84737
2773 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2774 copying restrict info.
2775 (vect_setup_realignment): Use it.
2776 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2777 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2778 the first DR to all generated stores.
2779 (vectorizable_load): Likewise for loads.
2781 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2783 PR tree-optimization/85446
2784 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2785 the integral and pointer types to have the same precision.
2787 * doc/install.texi: Document --disable-cet being the default and
2790 2018-04-18 Martin Liska <mliska@suse.cz>
2792 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2795 2018-04-18 Martin Liska <mliska@suse.cz>
2798 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2801 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2802 arguments if they are comparable.
2804 2018-04-18 Martin Liska <mliska@suse.cz>
2807 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2810 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2813 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2816 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2817 ENDBR after calling __morestack.
2819 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2822 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2823 by using gcc_base_ver to generate a gcc_driver_version, and use
2824 it when generating GCC_DRIVER_NAME.
2825 * configure: Regenerate.
2827 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2830 * config.gcc: Obsolete powerpc*-*-*spe*.
2832 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2835 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2836 (stabstr_D): Change type of unum from unsigned int to
2837 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2840 2018-04-17 Jim Wilson <jimw@sifive.com>
2843 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2844 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2845 Set arg_pointer_offset after using pretend_args_size.
2847 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2849 PR rtl-optimization/85431
2850 * dse.c (record_store): Ignore zero width stores.
2853 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2854 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2855 __builtin_stack_restore rather than after it.
2856 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2857 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2858 argument instead of virtual_dynamic_stack_rtx.
2860 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2862 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2864 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2865 Add note to error message to explain internal mapping of overloaded
2866 built-in function name to non-overloaded built-in function name.
2867 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2870 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2873 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2874 where the inputs overlap with the output.
2876 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2878 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2879 (=v, v) alternative and explicit "memory" attribute.
2880 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2881 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2883 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2884 "sselog1" type instead of "sselog".
2885 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2886 "sselog". Remove explicit "memory" attribute.
2887 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2888 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2890 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2891 "sselog1" type instead of "sselog". Remove explicit "memory"
2893 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2894 use "sselog1" type instead of "sselog". Remove explicit "memory"
2896 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2897 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2899 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2900 "sselog1" type instead of "sselog". Remove explicit "memory"
2902 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2903 use "sselog1" type instead of "sselog". Remove explicit "memory"
2907 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2910 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2913 2018-04-17 Martin Jambor <mjambor@suse.cz>
2916 * ipa-cp.c (create_specialized_node): Call
2917 expand_all_artificial_thunks if necessary.
2919 2018-04-17 Martin Liska <mliska@suse.cz>
2922 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2923 in message, remote space in between '_G' and '('.
2925 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2928 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2929 avx512f_vmcmp<mode>3<round_saeonly_name>,
2930 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2931 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2932 avx512f_rndscale<mode><round_saeonly_name>,
2933 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2934 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2935 Use %<iptr>2 instead of %2 for -masm=intel.
2936 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2937 avx512f_vcvttss2usi<round_saeonly_name>,
2938 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2940 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2941 avx512f_vcvttsd2usi<round_saeonly_name>,
2942 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2943 Use %q1 instead of %1 for -masm=intel.
2944 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2945 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2946 of %3 for -masm=intel.
2947 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2949 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2951 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2953 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2954 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2956 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2958 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2959 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2960 %g1 and one with %0 and %1.
2961 (avx512er_vmrcp28<mode><round_saeonly_name>,
2962 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2964 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2965 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2966 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2967 of %0 and %{%4%} for -masm=intel.
2968 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2969 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2970 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2971 order of %0 and %{%5%}%{z%} for -masm=intel.
2973 2018-04-17 Jan Hubicka <jh@suse.cz>
2976 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2978 2018-04-17 Martin Liska <mliska@suse.cz>
2981 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2982 for target_clone error message. Make default implementation
2983 clone to be a local declaration.
2984 (separate_attrs): Add new argument and check for an empty
2986 (expand_target_clones): Handle it.
2987 (ipa_target_clone): Make redirection just for target_clones
2990 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2991 Tom de Vries <tom@codesourcery.com>
2994 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2995 tiled basic blocks without omp continue statements.
2997 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3000 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3001 vec_extract expression as having side effects to make sure it gets
3004 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3007 * config/i386/i386.c (get_builtin_code_for_version): Check
3010 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3013 * gcc/config.gcc: Support "skylake".
3014 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3016 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3017 (processor_target_table): Add "skylake".
3018 (ix86_option_override_internal): Add "skylake".
3019 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3020 PROCESSOR_CANNONLAKE.
3021 (get_builtin_code_for_version): Fix priority for
3022 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3023 PROCESSOR_SKYLAKE-AVX512.
3024 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3025 (processor_type): Add PROCESSOR_SKYLAKE.
3027 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3028 Jason Merrill <jason@redhat.com>
3031 * convert.c (convert_to_integer_1): Use direct recursion for
3032 enumeral types and types with a precision less than the number
3033 of bits in their mode.
3035 2018-04-16 Julia Koval <julia.koval@intel.com>
3038 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3039 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3041 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3044 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3045 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3046 and -mno-direct-move.
3048 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3051 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3052 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3053 Ensure negative shifts result in {0}.
3055 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3057 PR rtl-optimization/79916
3058 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3059 regs (if any) to define how to gnerate SD moves when LRA is in
3062 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3064 PR rtl-optimization/85393
3065 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3066 * except.c (expand_dw2_landing_pad_for_region): Make static.
3067 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3068 a label and unconditional jump to old_bb, rather than
3069 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3072 PR rtl-optimization/85376
3073 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3074 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3075 instead of a specific value.
3077 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3078 Bin Cheng <bin.cheng@arm.com>
3080 PR tree-optimization/82965
3081 PR tree-optimization/83991
3082 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3083 by_profile_only parameter.
3084 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3085 information if the loop was predicted to iterate too many times.
3086 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3088 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3091 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3094 2018-04-13 Martin Liska <mliska@suse.cz>
3095 Jakub Jelinek <jakub@redhat.com>
3098 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3099 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3100 * builtins.c (expand_builtin_memory_copy_args): Use
3101 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3102 handle dest_addr == pc_rtx.
3104 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3107 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3108 asked to not generate direct moves.
3109 (fix_trunc<mode>si2_stfiwx): Similar.
3110 (fix_trunc<mode>si2_internal): Similar.
3112 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3115 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3116 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3117 lookup if dest in some wider mode is known to be const0_rtx and
3118 if so, record permanent equivalence for it to be ZERO_EXTEND of
3119 the narrower mode destination.
3121 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3123 * lto-streamer-out.c (output_function): Revert 259346.
3124 * omp-expand.c (expand_oacc_for): Likewise.
3126 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3128 PR rtl-optimization/85354
3129 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3130 * sel-sched.c (sel_global_init): ... here.
3132 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3135 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3136 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3137 mode for PE-COFF targets.
3138 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3139 (i386_pe_asm_lto_end): Likewise.
3140 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3141 (TARGET_ASM_LTO_END): Likewise.
3142 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3143 (i386_pe_asm_lto_start): New function.
3144 (i386_pe_asm_lto_end): Likewise.
3146 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3147 Richard Biener <rguenther@suse.de>
3150 * lto-streamer-out.c (output_function): Fix CFG loop state before
3152 * omp-expand.c (expand_oacc_for): Handle calls to internal
3153 functions like regular functions.
3155 2018-04-12 Richard Biener <rguenther@suse.de>
3158 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3159 for the early LTO debug to properly generate references to it
3160 during DIE emission. Do not re-use that for the skeleton for
3162 (dwarf2out_early_finish): Likewise.
3164 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3167 * config/i386/sse.md
3168 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3169 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3170 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3171 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3172 and output is a reg, avoid creating invalid lowpart subreg, but
3173 instead split into a 512-bit move. Don't split if not AVX512VL,
3174 input is xmm16+ reg and output is a mem.
3175 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3176 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3177 xmm16+ reg and output is a mem.
3179 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3181 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3182 also for flag_dwarf2_cfi_asm.
3184 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3186 PR rtl-optimization/85342
3187 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3188 a bool scalar var inside of the loop instead. Don't try to update
3189 recog_data.operand after failed apply_change_group.
3191 2018-04-12 Tom de Vries <tom@codesourcery.com>
3194 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3195 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3196 array with flexible array member as array without given dimension.
3197 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3198 argument for undefined param to true.
3200 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3203 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3204 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3205 from PowerPC section.
3206 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3207 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3208 help text that is too long.
3209 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3210 help text that is too long.
3211 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3212 help text that is too long.
3214 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3216 * config/alpha/alpha.md (stack_probe_internal): Rename
3217 from "probe_stack". Update all callers.
3219 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3221 PR rtl-optimization/84566
3222 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3223 sched_macro_fuse_insns.
3225 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3228 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3229 (compute_block_dependences): ... from here.
3231 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3233 PR tree-optimization/85331
3234 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3235 from int to HOST_WIDE_INT.
3237 2018-04-11 Martin Jambor <mjambor@suse.cz>
3240 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3241 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3242 not the same as the source val.
3243 (cgraph_edge_brings_value_p): New parameter.
3244 (gather_edges_for_value): Pass destination value to
3245 cgraph_edge_brings_value_p.
3246 (perhaps_add_new_callers): Likewise.
3247 (get_info_about_necessary_edges): Likewise and exclude values brought
3248 only by self-recursive edges.
3249 (create_specialized_node): Redirect only clones of self-calling edges.
3250 (+self_recursive_pass_through_p): New function.
3251 (find_more_scalar_values_for_callers_subset): Use it.
3252 (find_aggregate_values_for_callers_subset): Likewise.
3253 (known_aggs_to_agg_replacement_list): Removed.
3254 (decide_whether_version_node): Re-calculate known constants for all
3255 remaining context clones.
3257 2018-04-11 Richard Biener <rguenther@suse.de>
3260 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3261 from early DWARF output.
3262 (dwarf2out_early_finish): Output line info unconditionally into
3263 early DWARF and add reference to it.
3265 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3268 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3269 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3270 other than V2DFmode using iptr mode attribute.
3271 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3273 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3275 PR rtl-optimization/84659
3276 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3278 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3281 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3283 (output_loc_list): Pass address of a dummy size variable even in the
3284 locview handling loop.
3285 (index_location_lists): Add comment on why skip_loc_list_entry can't
3288 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3291 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3294 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3297 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3298 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3299 and -mstring-compare-inline-limit.
3301 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3304 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3305 for stack clash protection in a register whenever we need it to be in
3308 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3310 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3311 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3313 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3316 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3318 (mlong-double-): Ditto.
3319 * config/rs6000/sysv4.opt (msdata=): Ditto.
3320 (mtls-size=): Ditto.
3322 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3324 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3325 erroneous entries for
3326 "vector int vec_ldl (int, long int *)", and
3327 "vector unsigned int vec_ldl (int, unsigned long int *)".
3328 Add comments and entries for
3329 "vector bool char vec_ldl (int, bool char *)",
3330 "vector bool short vec_ldl (int, bool short *)",
3331 "vector bool int vec_ldl (int, bool int *)",
3332 "vector bool long long vec_ldl (int, bool long long *)",
3333 "vector pixel vec_ldl (int, pixel *)",
3334 "vector long long vec_ldl (int, long long *)",
3335 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3336 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3337 type tree bool_long_long_type_node and correct definition of
3338 bool_V2DI_type_node to make reference to this new type tree.
3339 (rs6000_mangle_type): Replace erroneous reference to
3340 bool_long_type_node with bool_long_long_type_node.
3341 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3342 comments to emphasize sign distinctions for char and int types and
3343 replace RS6000_BTI_bool_long constant with
3344 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3345 use of RS6000_BTI_pixel.
3346 (bool_long_type_node): Remove this macro definition.
3347 (bool_long_long_type_node): New macro definition
3349 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3351 PR rtl-optimization/85300
3352 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3353 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3354 simplify_unary_operation fails.
3356 2018-04-10 Martin Liska <mliska@suse.cz>
3358 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3359 cgraph_edge and ipa_ref.
3361 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3365 * config/i386/sse.md
3366 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3367 computation of the VEC_MERGE selector from mask.
3368 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3369 Fix decoding of the VEC_MERGE selector into mask.
3371 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3373 PR tree-optimization/85286
3374 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3376 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3378 * final.c (final_1): Set insn_last_address as well as
3379 insn_current_address.
3381 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3384 * explow.c (emit_stack_probe): Call validize_mem on memory location
3385 before passing it to gen_probe_stack. Create address operand and
3386 legitimize it for the probe_stack_address case.
3388 2018-04-09 Jan Hubicka <jh@suse.cz>
3391 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3392 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3393 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3395 (free_lang_data): Rebuild type inheritance graph.
3397 2018-04-09 Martin Sebor <msebor@redhat.com>
3399 * invoke.texi (-finline-small-functions): Mention other optimization
3401 (-findirect-inlining, -fpartial-inlining): Same.
3402 (-finline-functions-called-once): Same.
3403 (-freorder-blocks-and-partition): Same.
3405 2018-04-09 Jan Hubicka <jh@suse.cz>
3408 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3409 jumps; choose last target that matches the criteria (i.e.
3410 no partition changes for non-crossing jumps).
3411 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3412 support for redirecting crossing jumps to non-crossing.
3414 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3416 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3417 also for naked functions.
3419 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3421 * config/arc/arc.md (add_shift): New pattern.
3422 (add_shift2): Likewise.
3423 (sub_shift): Likewise.
3424 (sub_shift_cmp0_noout): Likewise.
3425 (compare_si_ashiftsi): Likewise.
3426 (xbfu_cmp0_noout): New combine pattern.
3427 (xbfu_cmp0"): Likewise.
3428 (movsi_set_cc_insn): Place the predicable variant first.
3429 (commutative_binary_cmp0_noout): Remove clobber.
3430 (commutative_binary_cmp0): New pattern.
3431 (noncommutative_binary_cmp0): Likewise.
3432 (noncommutative_binary_cmp0_noout): Likewise.
3433 (noncommutative_binary_comparison_result_used): Removed.
3434 (rsub_cmp0): New pattern.
3435 (rsub_cmp0_noout): Likewise.
3436 (extzvsi): Changed, keep only meaningful variants.
3437 (SQH, SEZ): New iterators.
3438 (SQH_postfix): New mode attribute.
3439 (SEZ_prefix): New code attribute.
3440 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3441 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3442 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3444 (noncommutative_operator): Check the availability of barrel
3447 2018-04-09 Richard Biener <rguenther@suse.de>
3449 PR tree-optimization/85284
3450 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3451 Only use the niter constraining form of simple_iv when the exit
3454 2018-04-09 Tom de Vries <tom@codesourcery.com>
3457 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3458 (define_expand "*memory_barrier"): New define_expand.
3459 (define_insn "memory_barrier"): New insn.
3461 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3463 PR rtl-optimization/80463
3464 PR rtl-optimization/83972
3465 PR rtl-optimization/83480
3467 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3468 correct producer for the insn.
3469 (tidy_control_flow): Fixup seqnos in case of debug insns.
3471 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3473 PR rtl-optimization/83913
3475 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3476 different sched-times when merging exprs.
3478 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3480 PR rtl-optimization/83962
3482 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3483 tidy_fallthru_edge and tidy_control_flow.
3485 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3487 PR rtl-optimization/83530
3489 * sel-sched.c (force_next_insn): New global variable.
3490 (remove_insn_for_debug): When force_next_insn is true, also leave only
3491 next insn in the ready list.
3492 (sel_sched_region): When the region wasn't scheduled, make another pass
3493 over it with force_next_insn set to 1.
3495 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3497 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3499 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3500 for interrupt control.
3501 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3502 functions for interrupt control.
3503 * config/nds32/nds32-intrinsic.md: Likewise.
3504 * config/nds32/nds32_intrinsic.h: Likewise.
3505 * config/nds32/nds32.h (nds32_builtins): Likewise.
3507 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3509 * config/nds32/nds32.c (nds32_init_machine_status,
3510 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3511 strict_aligned_p field.
3512 (nds32_expand_to_rtl_hook): New function.
3513 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3514 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3516 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3517 Chung-Ju Wu <jasonwucj@gmail.com>
3519 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3520 * config/nds32/nds32-n7.md: New file.
3521 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3522 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3524 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3525 * config/nds32/nds32.md (pipeline_model): Add n7.
3526 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3527 * config/nds32/pipelines.md: Include n7 settings.
3529 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3530 Chung-Ju Wu <jasonwucj@gmail.com>
3532 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3533 * config/nds32/nds32-e8.md: New file.
3534 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3535 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3537 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3538 * config/nds32/nds32.md (pipeline_model): Add e8.
3539 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3540 * config/nds32/pipelines.md: Include e8 settings.
3542 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3543 Chung-Ju Wu <jasonwucj@gmail.com>
3545 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3546 * config/nds32/nds32-n8.md: New file.
3547 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3548 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3550 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3551 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3552 * config/nds32/nds32.md (pipeline_model): Add n8.
3553 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3554 * config/nds32/pipelines.md: Include n8 settings.
3556 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3557 Chung-Ju Wu <jasonwucj@gmail.com>
3559 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3560 * config/nds32/nds32-n9-2r1w.md: New file.
3561 * config/nds32/nds32-n9-3r2w.md: New file.
3562 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3563 nds32_register_ports): New or modify for cpu n9.
3564 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3566 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3567 * config/nds32/nds32-utils.c: New file.
3568 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3569 TARGET_MUL_SLOW): Define.
3570 * config/nds32/nds32.md (pipeline_model): New attribute.
3571 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3572 New options that support cpu n9.
3573 * config/nds32/pipelines.md: Include n9 settings.
3574 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3576 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3578 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3579 information if necessary.
3580 (output_cond_branch_compare_zero): Likewise.
3581 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3582 (nds32_target_alignment): Refine for alignment.
3583 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3584 (FUNCTION_BOUNDARY): Modify.
3585 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3587 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3589 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3591 * config/nds32/constants.md (unspec_volatile_element): Add values for
3592 TLB operation and data prefetch.
3593 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3594 functions for TLB operation and data prefetch.
3595 * config/nds32/nds32-intrinsic.md: Likewise.
3596 * config/nds32/nds32_intrinsic.h: Likewise.
3597 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3598 (nds32_print_operand): Likewise.
3599 * config/nds32/nds32.h (nds32_builtins): Likewise.
3601 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3602 Andrew Pinski <pinsika@gcc.gnu.org>
3605 * match.pd: Use constant_boolean_node of correct type instead of
3606 boolean_true_node or boolean_false_node for simplifying
3607 pointer comparisons to zero.
3609 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3611 PR tree-optimization/80021
3612 * tree.c (verify_type_variant): Make error call in verify_variant_match
3613 translatable and remove final full stop.
3615 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3617 * config/nds32/constants.md (unspec_volatile_element): Add
3618 UNSPEC_VOLATILE_EH_RETURN.
3619 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3620 nds32_output_stack_pop): Support dwarf exception handling process.
3621 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3622 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3623 exception handling process.
3624 (nds32_compute_stack_frame): Likewise.
3625 (nds32_return_addr_rtx): Likewise.
3626 (nds32_initial_elimination_offset): Likewise.
3627 (nds32_expand_prologue): Likewise.
3628 (nds32_expand_epilogue): Likewise.
3629 (nds32_dynamic_chain_address): New function.
3630 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3632 (DYNAMIC_CHAIN_ADDRESS): Define.
3633 (EH_RETURN_DATA_REGNO): Define.
3634 (EH_RETURN_STACKADJ_RTX): Define.
3635 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3636 patterns for dwarf exception handling.
3638 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3640 * config/nds32/nds32.h: Clean up obsolete macros.
3642 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3644 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3645 Add enum values for particular instructions.
3646 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3647 particular intrinsic functions.
3648 * config/nds32/nds32-intrinsic.md: Likewise.
3649 * config/nds32/nds32_intrinsic.h: Likewise.
3650 * config/nds32/nds32.h (nds32_builtins): Likewise.
3651 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3652 (btst, ave): New patterns for particular instructions.
3654 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3656 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3657 Add enum values for atomic load/store and memory sync.
3658 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3660 * config/nds32/nds32-intrinsic.md: Likewise.
3661 * config/nds32/nds32_intrinsic.h: Likewise.
3662 * config/nds32/nds32.h (nds32_builtins): Likewise.
3664 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3666 PR tree-optimization/85257
3667 * fold-const.c (native_encode_vector): If not all elts could fit
3668 and off is -1, return 0 rather than offset.
3669 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3670 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3671 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3672 adjust buffer in native_interpret_expr call.
3674 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3676 * config/nds32/constants.md (unspec_volatile_element): Add cache
3677 control enum values.
3678 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3679 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3680 * config/nds32/nds32.c (nds32_cctl_names): New.
3681 (nds32_print_operand): Handle cache control register names.
3682 * config/nds32/nds32.h (nds32_builtins): New enum values.
3683 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3685 * config/nds32/nds32.md (type): Add mmu.
3686 * config/nds32/pipelines.md (simple_insn): Add mmu.
3688 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3690 * config/nds32/nds32.md (type): Remove call.
3691 * config/nds32/pipelines.md (simple_insn): Likewise.
3693 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3695 * config/nds32/constants.md (unspec_volatile_element): Add
3696 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3697 UNSPEC_VOLATILE_FMFCFG.
3698 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3699 description for fmfcfg and fmfcsr.
3700 (bdesc_1arg): Add fmtcsr.
3701 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3702 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3703 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3704 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3705 unspec_fmfcfg): New patterns.
3706 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3707 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3708 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3709 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3710 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3711 __nds32__fmfcfg): Define.
3713 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3715 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3716 intrinsic register names.
3717 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3718 intrinsic register enum values and macros.
3720 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3722 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3723 for load/store addressing form.
3724 (nds32_print_operand_address): Likewise.
3726 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3729 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3730 based on LABEL_REF. Remove useless assertion.
3731 (pic_address_needs_scratch): Fix formatting.
3732 (sparc_legitimize_pic_address): Minor tweaks.
3733 (sparc_delegitimize_address): Adjust assertion accordingly.
3734 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3735 into symbolic_operand.
3736 (movsi_high_pic_label_ref): Likewise.
3737 (movsi_lo_sum_pic_label_ref): Likewise.
3738 (movdi_pic_label_ref): Likewise.
3739 (movdi_high_pic_label_ref): Likewise.
3740 (movdi_lo_sum_pic_label_ref): Likewise.
3742 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3744 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3745 custom LIB_SPEC setup.
3747 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3748 Kito Cheng <kito.cheng@gmail.com>
3750 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3751 * config/riscv/freebsd.h: New.
3753 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3755 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3756 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3759 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3760 Kito Cheng <kito.cheng@gmail.com>
3762 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3763 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3764 * config/nds32/nds32-protos.h (nds32_output_call,
3765 nds32_output_return): Declare.
3766 * config/nds32/nds32.md: Refine all the call and return patterns.
3768 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3771 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3772 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3774 PR rtl-optimization/84872
3775 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3776 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3779 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3781 * expr.c (copy_blkmode_to_reg): Revert 254862.
3782 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3784 2018-04-06 Richard Biener <rguenther@suse.de>
3787 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3788 after seeing a component reference with an adjacent field. Treat
3789 refs to arrays at struct end of external decls similar to
3790 refs to unconstrained commons.
3792 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3795 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3796 look through SAVE_EXPRs with non-side-effects argument. Adjust
3798 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3801 2018-04-06 Richard Biener <rguenther@suse.de>
3804 * alias.c (find_base_term): New wrapper around find_base_term
3805 unwinding CSELIB_VAL_PTR changes.
3806 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3809 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3811 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3813 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3814 constant definitions.
3815 ("nop"): lr 0,0 -> nopr r0
3816 ("nop_lr0", "nop_lr1"): New insn definitions.
3818 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3820 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3821 NDS32_V3PUSH_AVAILABLE_P macro.
3823 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3824 Chung-Ju Wu <jasonwucj@gmail.com>
3826 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3827 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3828 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3829 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3830 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3831 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3832 * config/nds32/constraints.md: New constraints and checking for hard
3833 float configuration.
3834 * config/nds32/iterators.md: New mode iterator and attribute for hard
3835 float configuration.
3836 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3838 * config/nds32/nds32-fpu.md: New file.
3839 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3840 deal with hard float code generation.
3841 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3843 (abi_type, float_reg_number): New enum type.
3844 * config/nds32/nds32-predicates.c: New predicates for hard float.
3845 * config/nds32/nds32-protos.h: Declare functions for hard float.
3846 * config/nds32/nds32.c: Implementation for hard float configuration.
3847 * config/nds32/nds32.h: Definitions for hard float configuration.
3848 * config/nds32/nds32.md: Include hard float machine description and
3849 modify patterns for hard float configuration.
3850 * config/nds32/nds32.opt: New options for hard float configuration.
3851 * config/nds32/predicates.md: New predicates for hard float
3854 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3856 * common/config/nds32/nds32-common.c
3857 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3859 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3862 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3863 CONSTRUCTOR_ELT (ctor, ...)->value.
3865 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3868 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3870 2018-04-05 Tom de Vries <tom@codesourcery.com>
3873 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3876 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3877 Kito Cheng <kito.cheng@gmail.com>
3879 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3880 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3881 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3882 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3884 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3885 Kito Cheng <kito.cheng@gmail.com>
3887 * config/nds32/constraints.md (Ufe): New memory constraint.
3888 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3889 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3890 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3892 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3893 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3895 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3897 * config/nds32/nds32.md: Use optimize_size in the condition for
3898 alu-shift instructions.
3900 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3902 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3904 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3906 * config/nds32/nds32.md (negsi2): Refine pattern.
3908 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3909 Chung-Ju Wu <jasonwucj@gmail.com>
3911 * config/nds32/iterators.md (shift_rotate): New code iterator.
3912 (shift): New code attribute.
3913 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3914 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3915 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3916 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3917 bit-wise operations.
3918 (andsi3, *andsi3): Ditto.
3919 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3920 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3921 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3922 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3923 nds32_ior_operand, nds32_xor_operand): New predicates.
3925 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3927 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3928 (addsi3, subsi3): ... this.
3930 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3932 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3934 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3936 * config/nds32/nds32.md: Adjust indention.
3938 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3940 * config/nds32/nds32.md (feature): New attribute.
3942 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3944 * config/nds32/nds32.md (subtype): New attribute.
3946 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3949 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3950 expansion to perform a bitwise AND of the argument followed by a
3951 boolean negation of the result.
3953 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3955 PR rtl-optimization/84878
3956 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3957 the basic block. Assert the use reference is not artificial and that
3958 it has an associated insn.
3960 2018-04-04 Michael Matz <matz@suse.de>
3962 * builtins.c (compute_objsize): Pass correct operand
3963 to array_at_struct_end_p.
3965 2018-04-04 Richard Biener <rguenther@suse.de>
3968 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3969 from contexts for DINFO_LEVEL_TERSE and below.
3971 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3973 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3974 resiter_operand condition.
3975 * config/nds32/nds32.md (*move<mode>): Ditto.
3977 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3978 Monk Chiang <sh.chiang04@gmail.com>
3980 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3982 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3984 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3986 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3987 Kito Cheng <kito.cheng@gmail.com>
3989 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3990 nds32_cond_code_str, output_cond_branch,
3991 output_cond_branch_compare_zero, nds32_expand_cbranch,
3992 nds32_expand_cstore, nds32_expand_movcc,
3993 nds32_output_cbranchsi4_equality_zero,
3994 nds32_output_cbranchsi4_equality_reg,
3995 nds32_output_cbranchsi4_equality_reg_or_const_int,
3996 nds32_output_cbranchsi4_greater_less_zero: New functions.
3997 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3998 nds32_expand_cstore, nds32_expand_movcc,
3999 nds32_output_cbranchsi4_equality_zero,
4000 nds32_output_cbranchsi4_equality_reg,
4001 nds32_output_cbranchsi4_equality_reg_or_const_int,
4002 nds32_output_cbranchsi4_greater_less_zero): Declare.
4003 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4004 nds32_rimm11s_operand): New predicates.
4005 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4006 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4009 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4011 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4012 * config/nds32/nds32.md: Ditto.
4013 * config/nds32/pipelines.md: Ditto.
4015 2018-04-04 Richard Biener <rguenther@suse.de>
4017 PR tree-optimization/85168
4018 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4019 propagating abnormals.
4021 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4023 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4025 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4026 Kito Cheng <kito.cheng@gmail.com>
4028 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4029 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4030 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4031 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4032 * config/nds32/nds32.md (sibcall_internal): New.
4033 (sibcall_register): Remove.
4034 (sibcall_immediate): Remove.
4035 (sibcall_value_internal): New.
4036 (sibcall_value_register): Remove.
4037 (sibcall_value_immediate): Remove.
4038 * config/nds32/predicates.md (nds32_general_register_operand): New.
4039 (nds32_call_address_operand): New.
4041 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4043 PR rtl-optimization/85167
4044 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4045 bb_defs if *split_p, instead preinitialize it to NULL.
4047 PR tree-optimization/85156
4048 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4049 evaluating the argument multiple times.
4051 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4053 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4055 (_mm_cvtpd_ps): Likewise.
4056 (_mm_cvttpd_epi32): Likewise.
4057 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4058 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4059 vector, pixel, and bool following altivec.h include.
4061 2018-04-03 Martin Sebor <msebor@redhat.com>
4063 * doc/extend.texi (Common Function Attributes): Clarify.
4064 (const attribute): Likewise.
4065 (pure attribute): Likewise.
4067 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4070 * config/i386/i386.c (ix86_expand_vector_set): Use
4071 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4073 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4075 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4076 instructions when changing rounding bits to preserve precision bits
4077 in the x87 control word.
4079 2018-04-03 Martin Liska <mliska@suse.cz>
4081 PR tree-optimization/82491
4082 * rtl.h (strip_offset_and_add): Replace += suboffset with
4083 poly_uint64 () + suboffset.
4085 2018-03-29 Martin Liska <mliska@suse.cz>
4086 Martin Jambor <mjambor@suse.cz>
4089 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4090 param_type is not an integral or pointer type.
4092 2018-04-03 Richard Biener <rguenther@suse.de>
4094 * sese.h (recompute_all_dominators): Remove.
4096 2018-04-02 Martin Sebor <msebor@redhat.com>
4098 * doc/invoke.texi (-Wrestrict): Fix typos.
4100 2018-04-02 Jim Wilson <jimw@sifive.com>
4102 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4103 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4104 (<optab>di3, <optab>si3_extend): Likewise.
4105 (<optab>si3_mask, <optab>si3_mask_1): New.
4106 (<optab>di3_mask, <optab>di3_mask_1): New.
4107 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4108 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4109 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4111 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4113 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4116 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4118 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4119 (nds32_canonicalize_comparison): New function.
4121 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4122 Kito Cheng <kito.cheng@gmail.com>
4123 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4125 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4126 * config/nds32/constants.md (unspec_volatile_element): Add
4127 UNSPEC_VOLATILE_RELAX_GROUP.
4128 * config/nds32/nds32-relax-opt.c: New file.
4129 * config/nds32/nds32-predicates.c
4130 (nds32_symbol_load_store_p): New function.
4131 * config/nds32/nds32-protos.h
4132 (nds32_symbol_load_store_p): Declare function.
4133 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4134 * config/nds32/nds32.c
4135 (nds32_register_pass): New function to register pass.
4136 (nds32_register_passes): New function to register passes.
4137 * config/nds32/nds32.md (relax_group): New pattern.
4138 * config/nds32/nds32.opt (mrelax-hint): New option.
4139 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4141 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4143 * config/nds32/t-nds32: Modify files dependency.
4145 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4147 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4148 (PROFILE_HOOK): Define its implementation.
4150 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4152 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4153 type and 32-bit size.
4155 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4158 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4159 (V_128_256): New mode iterator.
4160 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4161 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4162 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4164 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4165 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4167 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4170 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4171 NaN inputs correctly.
4173 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4176 * config/rs6000/vsx.md (??r): New mode attribute.
4177 (*vsx_mov<mode>_64bit): Use it.
4178 (*vsx_mov<mode>_32bit): Likewise.
4180 2018-03-30 Martin Sebor <msebor@redhat.com>
4182 PR tree-optimization/84818
4183 * builtins.c (check_access): Use warning_n.
4185 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4188 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4190 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4193 2018-03-30 Julia Koval <julia.koval@intel.com>
4196 * x86-tune.def (movx, partial_reg_dependency): Enable for
4199 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4202 * lra-constraints.c (process_alt_operands): Move setting
4203 this_alternative_matches below.
4205 2018-03-29 Martin Liska <mliska@suse.cz>
4208 * doc/invoke.texi: Document how LTO works with debug info.
4209 Describe auto-load support of binutils. Mention 'x86-64'
4210 as valid option value of -march option.
4212 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4214 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4217 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4218 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4219 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4222 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4225 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4226 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4227 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4228 for RS6000_BTM_POWERPC64.
4229 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4230 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4231 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4236 2018-03-28 Carl Love <cel@us.ibm.com>
4239 2017-09-27 Carl Love <cel@us.ibm.com>
4241 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4242 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4243 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4246 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4248 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4249 instead of __vector bool.
4250 (_mm_max_pu8): Likewise.
4251 (_mm_min_pi16): Likewise.
4253 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4256 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4257 (DIVWEUO): Likewise.
4259 (DIVDEUO): Likewise.
4260 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4261 DIVWEUO and DIVDEUO.
4262 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4263 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4264 (div_extend): Likewise.
4265 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4267 (__builtin_divweuo): Likewise.
4268 (__builtin_divdeo): Likewise.
4269 (__builtin_divdeuo): Likewise.
4271 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4274 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4275 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4277 PR tree-optimization/82004
4278 * gimple-match-head.c (optimize_pow_to_exp): New function.
4279 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4280 Don't fold to exp if optimize_pow_to_exp is false.
4282 2018-03-28 Martin Liska <mliska@suse.cz>
4285 * calls.c (initialize_argument_information): Fix trailing space.
4286 * common.opt: Fix typo and provide better explanation for
4287 -fsanitize-coverage option.
4288 * config/i386/i386.opt: Fix typo.
4290 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4291 Martin Liska <mliska@suse.cz>
4294 * gimplify.c (asan_poison_variable): Don't do the check for
4295 gimplify_omp_ctxp here.
4296 (gimplify_decl_expr): Do it here.
4297 (gimplify_target_expr): Likewise.
4299 2018-03-28 Martin Liska <mliska@suse.cz>
4302 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4303 chkp_type_bounds_count if MPX is not enabled.
4305 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4307 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4309 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4312 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4313 function to create the function decl for complex long double
4314 multiply and divide for -mabi=ieeelongdouble.
4315 (init_float128_ieee): Call it.
4317 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4320 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4321 -fcf-protection=branch -mibt.
4322 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4324 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4327 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4329 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4332 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4333 extern array declarations.
4335 2018-03-27 Richard Biener <rguenther@suse.de>
4338 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4339 explicit single_use checks.
4341 2018-03-27 Richard Biener <rguenther@suse.de>
4343 PR tree-optimization/85082
4344 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4347 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4349 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4350 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4351 Turn on fasynchronous-unwind-tables and funwind-tables.
4353 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4356 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4357 (*bmi_blsr_<mode>_ccz): Ditto.
4359 2018-03-26 Tom de Vries <tom@codesourcery.com>
4361 PR tree-optimization/85063
4362 * omp-general.c (offloading_function_p): New function. Factor out
4364 * omp-offload.c (pass_omp_target_link::gate): ... here.
4365 * omp-general.h (offloading_function_p): Declare.
4366 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4367 with attribute omp declare target for offloading functions.
4369 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4371 PR tree-optimization/84005
4372 * tree-data-ref.h (get_base_for_alignment): Declare.
4373 * tree-data-ref.c (get_base_for_alignment_1): New function.
4374 (get_base_for_alignment): Likewise.
4375 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4376 get_base_for_alignment to find a suitable base object, instead
4377 of always using drb->base_address.
4379 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4382 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4383 known size by default.
4385 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4388 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4389 and non BLKmode operands.
4391 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4394 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4395 Clean up attributes.
4397 2018-03-23 Richard Biener <rguenther@suse.de>
4400 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4401 we are going to emit early debug for LTO.
4403 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4406 * function.c (match_asm_constraints_1): Don't optimize if input
4407 doesn't satisfy general_operand predicate for output's mode.
4410 * alias.c (write_dependence_p): Don't require for x_canonicalized
4411 non-VOIDmode if x has VOIDmode.
4414 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4415 just don't try to optimize it rather than assert it never happens.
4417 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4419 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4420 macro expansions for definition of ST_INTERNAL_<mode> and
4421 LD_INTERNAL_<mode> builtins.
4422 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4424 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4426 (altivec_expand_st_builtin): Likewise.
4427 (altivec_expand_builtin): Remove calls to deleted functions.
4428 (rs6000_address_for_altivec): Delete this function.
4429 * config/rs6000/vector.md: Remove expands for
4430 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4432 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4435 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4436 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4437 re-computing once computed.
4438 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4439 (arm_init_machine_status): Initialize
4440 machine->static_chain_stack_bytes.
4442 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4445 * doc/extend.texi: Add four new prototypes for vec_ld.
4446 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4447 definitions for more logical presentation.
4448 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4449 entries for V1TI variants of __builtin_altivec_ld builtin.
4450 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4451 handling of V1TI variant of LVX icode pattern.
4452 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4453 (rs6000_gimple_fold_builtin): Likewise.
4454 (altivec_init_builtins): Add code to define
4455 __builtin_altivec_lvx_v1ti function.
4457 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4460 * function.c (match_asm_constraints_1): Don't do the optimization
4461 if input isn't a REG, SUBREG, MEM or constant.
4463 2018-03-22 Tom de Vries <tom@codesourcery.com>
4465 PR tree-optimization/84956
4466 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4467 bb_has_abnormal_pred.
4469 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4472 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4473 DECL_INITIAL (decl) to decl at the end.
4474 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4477 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4479 * doc/extend.texi (__builtin_tgmath): Document when complex
4480 integer types are treated as _Complex _Float64.
4482 2018-03-21 Tom de Vries <tom@codesourcery.com>
4484 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4486 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4488 PR tree-optimization/84960
4489 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4490 if it is ENTRY block, move them into single succ of ENTRY in that case.
4492 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4494 PR tree-optimization/84811
4495 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4496 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4497 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4498 (known_subrange_p): Update accordingly. Cast each value involved
4499 in the size comparison, rather than casting the result of the
4502 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4504 PR tree-optimization/84982
4505 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4506 by flipping the least significant bit rather than all bits from
4507 bitpos to bitpos + bitsize - 1.
4509 2018-03-21 Nathan Sidwell <nathan@acm.org>
4511 * doc/extend.texi (Deprecated Features): Remove mention of
4512 long-deleted deprecations.
4514 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4517 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4518 * configure: Regenerate.
4520 2018-03-21 Tom de Vries <tom@codesourcery.com>
4522 PR tree-optimization/83126
4523 * tree-parloops.c (num_phis): New function.
4524 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4526 2018-03-21 Nathan Sidwell <nathan@acm.org>
4528 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4529 mention anon-struct/union members and trailing attributes.
4531 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4533 PR tree-optimization/84969
4534 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4535 builtin memset partitions if they set different rhs values.
4537 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4539 PR rtl-optimization/84989
4540 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4541 VEC_DUPLICATE with scalar result mode.
4543 2018-03-21 Martin Liska <mliska@suse.cz>
4546 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4547 not intended return statement.
4549 2018-03-21 Martin Liska <mliska@suse.cz>
4552 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4553 (chkp_find_bound_slots_1): Limit number of iterations.
4555 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4558 * Minor grammar fixes for x86 options.
4560 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4563 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4564 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4566 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4569 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4570 (altivec_lvx_<mode>_1op): Likewise.
4571 (altivec_stvx_<mode>_2op): Likewise.
4572 (altivec_stvx_<mode>_1op): Likewise.
4573 (altivec_lvx_<VM2:mode>): New define_expand.
4574 (altivec_stvx_<VM2:mode>): Likewise.
4575 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4576 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4577 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4578 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4579 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4580 (rs6000_gen_lvx): Likewise.
4581 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4582 (altivec_expand_stv_builtin): Likewise.
4583 (altivec_expand_builtin): Likewise.
4584 * config/rs6000/vector.md: Likewise.
4586 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4589 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4592 2018-03-20 Richard Biener <rguenther@suse.de>
4595 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4596 sign-conversions as zero, fall back to standard scalar_stmt
4599 2018-03-20 Martin Liska <mliska@suse.cz>
4602 * predict.c (rebuild_frequencies): Handle case when we have
4603 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4605 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4608 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4609 flag_section_anchors.
4610 * varasm.c (use_blocks_for_decl_p): Remove hack for
4611 dw2_force_const_mem.
4614 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4616 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4617 be created, use lowpart_subreg of operands[0] rather than operands[0]
4619 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4620 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4621 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4622 and n constraint instead of aarch64_shift_imm_di and Usd.
4623 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4624 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4626 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4629 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4630 to favor GPR over NEON registers.
4631 (<shift>di3_neon): Likewise.
4633 2018-03-20 Tom de Vries <tom@codesourcery.com>
4636 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4637 (nvptx_process_pars): Emit bar.sync asap and alap.
4639 2018-03-20 Tom de Vries <tom@codesourcery.com>
4642 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4643 seen_label if seen_label is already set.
4645 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4648 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4649 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4650 Use 1U instead of 1. Formatting fixes.
4653 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4654 instead of TREE_TYPE (s1) for the return value.
4656 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4658 PR tree-optimization/84946
4659 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4660 bitsize + bitsize in poly_uint64 rather than poly_int64.
4663 * dwarf2asm.c: Include fold-const.c.
4664 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4665 of decl rather than decl itself.
4667 PR rtl-optimization/84643
4668 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4670 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4673 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4674 calling assemble_variable.
4676 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4679 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4680 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4682 2018-03-19 Jim Wilson <jimw@sifive.com>
4685 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4686 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4687 (riscv_first_stack_step): Likewise.
4688 (riscv_option_override): Use STACK_BOUNDARY instead of
4690 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4692 (BIGGEST_ALIGNMENT): Set to 128.
4693 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4694 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4697 2018-03-19 Richard Biener <rguenther@suse.de>
4699 PR tree-optimization/84933
4700 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4701 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4703 2018-03-19 Richard Biener <rguenther@suse.de>
4705 PR tree-optimization/84859
4706 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4707 (cond_if_else_store_replacement): Perform sinking operation on
4708 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4709 Generalize what a BB with a single eligible store is.
4711 2018-03-19 Richard Biener <rguenther@suse.de>
4713 PR tree-optimization/84929
4714 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4715 chrec_is_positive against non-chrec arg.
4717 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4720 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4722 2018-03-18 Martin Liska <mliska@suse.cz>
4724 PR rtl-optimization/84635
4725 * regrename.c (build_def_use): Use matches_mode only when
4728 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4730 PR tree-optimization/84913
4731 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4732 vectorize chains of COND_EXPRs.
4734 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4736 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4738 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4740 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4742 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4744 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4746 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4747 Kito Cheng <kito.cheng@gmail.com>
4749 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4750 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4751 (nds32_adjust_reg_alloc_order): New function.
4752 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4754 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4756 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4757 nds32_print_operand, nds32_print_operand_address): Use
4758 HOST_WIDE_INT_PRINT_DEC instead.
4760 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4762 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4764 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4767 * config/i386/i386.c (initial_ix86_tune_features,
4768 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4770 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4771 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4772 rather than 1u << ix86_tune. Formatting fix.
4773 (ix86_option_override_internal): Change ix86_arch_mask from
4774 unsigned int to unsigned HOST_WIDE_INT, initialize to
4775 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4776 (ix86_function_specific_restore): Likewise.
4778 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4781 * postreload.c (reload_combine_recognize_pattern): Perform
4782 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4783 truncate_int_for_mode the result for the destination's mode.
4786 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4787 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4789 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4792 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4795 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4796 regno_allocno_class_array and sorted_pseudos.
4797 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4798 insns where regno is used.
4800 2018-03-16 Martin Liska <mliska@suse.cz>
4803 * multiple_target.c (create_dispatcher_calls): Redirect
4804 reference in the symbol table.
4806 2018-03-16 Martin Liska <mliska@suse.cz>
4809 * multiple_target.c (create_dispatcher_calls): Redirect also
4812 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4816 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4817 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4818 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4820 2018-03-16 Julia Koval <julia.koval@intel.com>
4822 * doc/invoke.texi (Skylake Server): Add CLWB.
4823 Cannonlake): Remove CLWB.
4825 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4827 PR tree-optimization/84841
4828 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4830 (FLOAT_ONE_CONST_TYPE): Define.
4831 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4832 (sort_by_operand_rank): Put entries with higher constant_type last
4833 rather than first to match comments.
4835 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4837 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4840 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4843 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4844 check and use address of target always.
4846 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4849 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4850 (indirect_thunk_bnd_needed): Likewise.
4851 (indirect_thunks_used): Likewise.
4852 (indirect_thunks_bnd_used): Likewise.
4853 (indirect_return_needed): New.
4854 (indirect_return_bnd_needed): Likewise.
4855 (output_indirect_thunk_function): Add a bool argument for
4857 (output_indirect_thunk_function): Don't generate alias for
4858 function return thunk.
4859 (ix86_code_end): Call output_indirect_thunk_function to generate
4860 function return thunks.
4861 (ix86_output_function_return): Set indirect_return_bnd_needed
4862 and indirect_return_needed instead of indirect_thunk_bnd_needed
4863 and indirect_thunk_needed.
4865 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4867 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4868 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4869 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4871 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4872 Paul Hua <paul.hua.gm@gmail.com>
4875 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4877 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4879 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4880 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4883 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4886 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4887 instead of GET_MODE_SIZE when comparing Units.
4889 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4892 * varasm.c (hash_section): Return an unchangeble hash value
4893 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4894 Return !aarch64_can_use_per_function_literal_pools_p ().
4896 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4899 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4900 rather than address of cmode as last argument to prepare_cmp_insn.
4902 2018-03-15 Julia Koval <julia.koval@intel.com>
4904 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4905 F_AVX512VNNI, F_AVX512BITALG): New.
4907 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4910 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4911 insn for floating-point loads and stores.
4913 2018-03-14 Carl Love <cel@us.ibm.com>
4915 * config/rs6000/rs6000-c.c: Add macro definitions for
4916 ALTIVEC_BUILTIN_VEC_PERMXOR.
4917 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4918 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4919 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4920 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4922 * config/doc/extend.texi: Add prototypes for vec_permxor.
4924 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4927 * diagnostic-show-locus.c (class layout_point): Convert m_line
4928 from int to linenum_type.
4929 (line_span::comparator): Use linenum "compare" function when
4930 comparing line numbers.
4931 (test_line_span): New function.
4932 (layout_range::contains_point): Convert param "row" from int to
4934 (layout_range::intersects_line_p): Likewise.
4935 (layout::will_show_line_p): Likewise.
4936 (layout::print_source_line): Likewise.
4937 (layout::should_print_annotation_line_p): Likewise.
4938 (layout::print_annotation_line): Likewise.
4939 (layout::print_leading_fixits): Likewise.
4940 (layout::annotation_line_showed_range_p): Likewise.
4941 (struct line_corrections): Likewise for field m_row.
4942 (line_corrections::line_corrections): Likewise for param "row".
4943 (layout::print_trailing_fixits): Likewise.
4944 (layout::get_state_at_point): Likewise.
4945 (layout::get_x_bound_for_row): Likewise.
4946 (layout::print_line): Likewise.
4947 (diagnostic_show_locus): Likewise for locals "last_line" and
4949 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4950 * input.c (selftest::test_linenum_comparisons): New function.
4951 (selftest::input_c_tests): Call it.
4952 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4953 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4954 * selftest.h (ASSERT_GT): New macro.
4955 (ASSERT_GT_AT): New macro.
4956 (ASSERT_LT): New macro.
4957 (ASSERT_LT_AT): New macro.
4959 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4961 PR rtl-optimization/84780
4962 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4964 2018-03-14 Martin Liska <mliska@suse.cz>
4966 * tree.c (record_node_allocation_statistics): Use
4967 get_stats_node_kind.
4968 (get_stats_node_kind): New function extracted from
4969 record_node_allocation_statistics.
4970 (free_node): Use get_stats_node_kind.
4972 2018-03-14 Richard Biener <rguenther@suse.de>
4974 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4975 that the value-set of ANTIC_IN doesn't grow.
4978 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4980 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4981 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4983 2018-03-14 Julia Koval <julia.koval@intel.com>
4985 * config.gcc (icelake-client, icelake-server): New.
4987 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4988 (initial_ix86_arch_features): Ditto.
4989 (PTA_SKYLAKE): Add SGX.
4990 (PTA_ICELAKE): Remove.
4991 (PTA_ICELAKE_CLIENT): New.
4992 (PTA_ICELAKE_SERVER): New.
4993 (ix86_option_override_internal): Split up icelake on icelake client and
4995 (get_builtin_code_for_version): Ditto.
4996 (fold_builtin_cpu): Ditto.
4997 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4998 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4999 * config/i386/i386.h (processor_type): Ditto.
5000 * doc/invoke.texi: Ditto.
5002 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5005 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5006 INTEGER_CST offset, add it together with bitpos / 8 and
5007 sign extend based on POINTER_SIZE.
5011 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5014 * config/i386/constraints.md (Yc): New register constraint.
5015 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5016 Use Yc constraint for alternative 2 of operand 0. Remove
5017 preferred_for_speed attribute.
5019 2018-03-14 Richard Biener <rguenther@suse.de>
5021 PR tree-optimization/84830
5022 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5023 with the old one to avoid oscillations.
5025 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5028 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5030 (assign_by_spills): Return a flag of reload assignment failure.
5031 Do not process the reload assignment failures. Do not spill other
5032 reload pseudos if they has the same reg class. Update n if
5034 (lra_assign): Add a return arg. Set up from the result of
5035 assign_by_spills call.
5036 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5037 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5038 usage_insns if it is not NULL.
5039 (spill_hard_reg_in_range): New function.
5040 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5041 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5042 function prototypes.
5043 (lra_assign): Change prototype.
5044 * lra.c (lra): Add code to deal with fails by splitting hard reg
5047 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5049 * config/riscv/riscv.opt (mrelax): New option.
5050 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5051 "norelax" when riscv_mrelax is disabled.
5052 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5054 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5057 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5058 reassociation for int modes.
5060 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5062 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5063 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5065 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5066 * config/aarch64/aarch64-sve.md
5067 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5068 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5069 (*extend<mode><Vwide>2): Rename to...
5070 (aarch64_sve_extend<mode><Vwide>2): ...this.
5071 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5072 renaming the old pattern to...
5073 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5075 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5076 define_expand, renaming the old pattern to...
5077 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5078 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5079 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5080 account when deciding which SVE instruction the optab should use.
5081 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5083 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5085 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5086 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5087 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5088 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5089 (tlsdesc_small_<mode>): Turn a define_expand and use
5090 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5091 (tlsdesc_small_advsimd_<mode>): ...this.
5092 (tlsdesc_small_sve_<mode>): New pattern.
5094 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5096 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5097 (UNSPEC_UMUL_HIGHPART): New constants.
5098 (MUL_HIGHPART): New int iteraor.
5099 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5100 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5102 (*<su>mul<mode>3_highpart): New define_insn.
5104 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5107 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5110 2018-03-13 Martin Liska <mliska@suse.cz>
5113 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5115 (sem_item_optimizer::~sem_item_optimizer): Release it.
5116 (sem_item_optimizer::merge_classes): Register variable aliases.
5117 (sem_item_optimizer::fixup_pt_set): New function.
5118 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5119 * ipa-icf.h: Declare new variables and functions.
5121 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5124 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5125 integer_pow2p@2 and test integer_pow2p in condition.
5126 (A < 0 ? C : 0): Similarly for @1.
5129 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5130 characters starting at p contain '\0' character, don't look beyond
5134 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5135 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5138 * reg-stack.c (change_stack): Change update_end var from int to
5139 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5140 also call set_block_for_insn on the newly added insns and rescan.
5143 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5144 on the last operand.
5147 * tree.c (stabilize_reference_1): Return save_expr (e) for
5148 STATEMENT_LIST even if it doesn't have side-effects.
5150 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5152 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5154 2018-03-12 Renlin Li <renlin.li@arm.com>
5156 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5157 aarch64_output_scalar_simd_mov_immediate.
5159 2018-03-12 Martin Sebor <msebor@redhat.com>
5161 PR tree-optimization/83456
5162 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5163 for perfectly overlapping calls to memcpy.
5164 (gimple_fold_builtin_memory_chk): Same.
5165 (gimple_fold_builtin_strcpy): Handle no-warning.
5166 (gimple_fold_builtin_stxcpy_chk): Same.
5167 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5169 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5171 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5172 parameter. Use it for SFmode.
5173 (rs6000_function_arg_advance_1): Adjust.
5174 (rs6000_function_arg): Adjust.
5175 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5177 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5179 PR rtl-optimization/84169
5180 PR rtl-optimization/84780
5181 * combine.c (can_combine_p): Check for a 2-insn combination whether
5182 the destination register is used between the two insns, too.
5184 2018-03-12 Richard Biener <rguenther@suse.de>
5186 PR tree-optimization/84803
5187 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5188 for refs DR analysis didn't process.
5190 2018-03-12 Richard Biener <rguenther@suse.de>
5192 PR tree-optimization/84777
5193 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5194 force-vectorize loops ignore whether we are optimizing for size.
5196 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5198 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5199 (TARGET_MD_ASM_ADJUST): Define.
5201 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5202 Kito Cheng <kito.cheng@gmail.com>
5203 Chung-Ju Wu <jasonwucj@gmail.com>
5205 * config/nds32/nds32.c (nds32_compute_stack_frame,
5206 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5207 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5208 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5209 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5210 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5211 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5212 * config/nds32/nds32.md (prologue, epilogue): Use macro
5213 NDS32_V3PUSH_AVAILABLE_P to do checking.
5215 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5218 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5219 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5220 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5221 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5222 addition of most attributes on !orig_type_die or the attribute not
5223 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5225 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5226 Chung-Ju Wu <jasonwucj@gmail.com>
5228 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5230 * config/nds32/nds32.opt (mvh): New option.
5232 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5233 Chung-Ju Wu <jasonwucj@gmail.com>
5235 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5237 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5238 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5241 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5242 Chung-Ju Wu <jasonwucj@gmail.com>
5244 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5246 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5247 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5249 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5250 Kito Cheng <kito.cheng@gmail.com>
5251 Chung-Ju Wu <jasonwucj@gmail.com>
5253 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5254 UNSPEC_FFMISM and UNSPEC_FLMISM.
5255 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5256 for ffb, ffmism and flmism.
5257 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5258 (unspec_ffmism): Ditto.
5259 (unspec_flmism): Ditto.
5260 (nds32_expand_builtin_impl): Check if string extension is available.
5261 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5262 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5264 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5267 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5270 * lra-assigns.c (assign_by_spills): Return a flag of reload
5271 assignment failure. Do not process the reload assignment
5272 failures. Do not spill other reload pseudos if they has the same
5274 (lra_assign): Add a return arg. Set up from the result of
5275 assign_by_spills call.
5276 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5277 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5278 usage_insns if it is not NULL.
5279 (spill_hard_reg_in_range): New function.
5280 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5281 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5282 function prototypes.
5283 (lra_assign): Change prototype.
5284 * lra.c (lra): Add code to deal with fails by splitting hard reg
5287 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5290 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5292 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5295 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5296 (dw_val_node): Add val_symbolic_view.
5297 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5298 (symview_upper_bound): New.
5299 (new_line_info_table): Initialize symviews_since_reset.
5300 (dwarf2out_source_line): Count symviews_since_reset and set
5301 symview_upper_bound.
5302 (dw_val_equal_p): Handle symview.
5303 (add_AT_symview): New.
5304 (print_dw_val): Handle symview.
5305 (attr_checksum, attr_checksum_ordered): Likewise.
5306 (same_dw_val_p, size_of_die): Likewise.
5307 (value_format, output_die): Likewise.
5308 (add_high_low_attributes): Use add_AT_symview for entry_view.
5309 (dwarf2out_finish): Reset symview_upper_bound, clear
5312 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5315 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5316 Add strict argument and use it.
5317 (rs6000_split_multireg_move): Update for new strict argument.
5318 (mem_operand_gpr): Disallow all non-offsettable addresses.
5319 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5321 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5324 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5325 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5326 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5329 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5330 decl, use remap_type if we want to use the type.
5332 2018-03-09 Martin Sebor <msebor@redhat.com>
5334 PR tree-optimization/84526
5335 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5337 (builtin_access::generic_overlap): Be prepared to handle non-array
5340 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5342 PR rtl-optimization/84682
5343 * lra-constraints.c (process_address_1): Check is_address flag
5344 for address constraints.
5345 (process_alt_operands): Likewise.
5346 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5347 preprocess_constraints.
5348 * recog.h (preprocess_constraints): Add oploc parameter.
5350 * recog.c (preprocess_constraints): Test address_operand for
5351 CT_ADDRESS constraints.
5353 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5356 * lra-assigns.c (assign_by_spills): Return a flag of reload
5357 assignment failure. Do not process the reload assignment
5358 failures. Do not spill other reload pseudos if they has the same
5360 (lra_assign): Add a return arg. Set up from the result of
5361 assign_by_spills call.
5362 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5363 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5364 usage_insns if it is not NULL.
5365 (spill_hard_reg_in_range): New function.
5366 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5367 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5368 function prototypes.
5369 (lra_assign): Change prototype.
5370 * lra.c (lra): Add code to deal with fails by splitting hard reg
5373 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5376 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5377 Accept complain bool parameter. Only emit errors if it is true.
5378 (arm_parse_cpu_option_name): Likewise.
5379 (arm_target_thumb_only): Adjust callers of the above.
5380 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5381 prototype to take a default true bool parameter.
5382 (arm_parse_arch_option_name): Likewise.
5384 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5385 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5389 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5390 * configure: Regenerate.
5391 * configure.ac ("linker --version-script option"): New.
5392 ("linker soname option"): New.
5394 2018-03-09 Richard Biener <rguenther@suse.de>
5396 PR tree-optimization/84775
5397 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5398 immediate uses of predicate stmts and mark them modified.
5401 PR tree-optimization/84178
5402 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5404 (version_loop_for_if_conversion): Delay update_ssa call.
5405 (tree_if_conversion): Delay update_ssa until after predicate
5408 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5411 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5412 when the function accesses prior frames.
5414 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5417 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5418 gen_llsym, otherwise call maybe_gen_llsym.
5421 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5422 has ',' character inside of it.
5424 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5427 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5428 as clobbering CC_REGNUM.
5430 2018-03-08 Richard Biener <rguenther@suse.de>
5433 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5434 (follow_copies_to_constant): Do not follow SSA names registered
5437 2018-03-08 Richard Biener <rguenther@suse.de>
5439 PR tree-optimization/84178
5440 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5442 (version_loop_for_if_conversion): Delay update_ssa call.
5443 (tree_if_conversion): Delay update_ssa until after predicate
5446 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5448 PR tree-optimization/84178
5449 * tree-if-conv.c (release_bb_predicate): Remove the
5450 the assertion that the stmts have NULL use_ops.
5451 Discard the statements, asserting that they haven't
5452 yet been added to a BB.
5454 2018-03-08 Richard Biener <rguenther@suse.de>
5456 PR tree-optimization/84746
5457 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5458 (phi_translate): Pass in destination ANTIC_OUT set.
5459 (phi_translate_1): Likewise. For a simplified result lookup
5460 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5461 (phi_translate_set): Adjust.
5462 (do_pre_regular_insertion): Likewise.
5463 (do_pre_partial_partial_insertion): Likewise.
5465 2018-03-08 Martin Liska <mliska@suse.cz>
5467 PR gcov-profile/84735
5468 * doc/gcov.texi: Document usage of profile files.
5469 * gcov-io.h: Document changes in the format.
5471 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5475 * dwarf2out.c (struct dw_line_info_table): Update comments for
5477 (FORCE_RESET_NEXT_VIEW): New.
5478 (FORCE_RESETTING_VIEW_P): New.
5479 (RESETTING_VIEW_P): Check for -1 too.
5480 (ZERO_VIEW_P): Likewise.
5481 (new_line_info_table): Force-reset next view.
5482 (dwarf2out_begin_function): Likewise.
5483 (dwarf2out_source_line): Simplify zero_view_p initialization.
5484 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5485 view directly. Omit view when omitting .loc at line 0.
5487 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5489 PR tree-optimization/84740
5490 * tree-switch-conversion.c (process_switch): Call build_constructors
5491 only if info.phi_count is non-zero.
5493 PR tree-optimization/84739
5494 * tree-tailcall.c (find_tail_calls): Check call arguments against
5495 DECL_ARGUMENTS (current_function_decl) rather than
5496 DECL_ARGUMENTS (func) when checking for tail recursion.
5498 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5500 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5501 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5502 Volker Reichelt's entry and add entries for people that perform
5503 GCC fuzzy testing and report numerous bugs.
5505 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5508 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5509 readonly data in sdata, if that is disabled.
5510 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5511 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5512 -mreadonly-in-sdata option.
5514 2018-03-07 Martin Sebor <msebor@redhat.com>
5516 PR tree-optimization/84468
5517 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5518 basic block when looking for nul assignment.
5520 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5523 * except.h (output_function_exception_table): Adjust prototype.
5524 * except.c (output_function_exception_table): Remove FNNAME parameter
5525 and add SECTION parameter. Ouput one part of the table at a time.
5526 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5527 the first part of the exception table and emit unwind directives.
5528 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5529 (i386_pe_seh_cold_init): Likewise.
5530 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5531 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5532 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5533 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5534 * config/i386/winnt.c: Include except.h.
5535 (struct seh_frame_state): Add reg_offset, after_prologue and
5536 in_cold_section fields.
5537 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5538 (i386_pe_seh_cold_init): New function.
5539 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5540 to seh->in_cold_section.
5541 (seh_emit_push): Record the offset of the push.
5542 (seh_emit_save): Record the offet of the save.
5543 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5544 Test seh->after_prologue to disregard the epilogue.
5545 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5546 (i386_pe_end_cold_function): New function.
5548 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5551 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5552 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5555 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5557 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5558 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5561 * multiple_target.c: Include tree-inline.h and intl.h.
5562 (expand_target_clones): Diagnose and fail if node->definition and
5563 !tree_versionable_function_p (node->decl).
5565 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5567 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5569 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5570 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5571 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5573 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5576 * combine.c (try_combine): Use reg_or_subregno instead of handling
5577 just paradoxical SUBREGs and REGs.
5579 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5581 * config/arc/arc.c (arc_finalize_pic): Remove function.
5582 (arc_must_save_register): We use single base PIC register, remove
5583 checks to save/restore the PIC register.
5584 (arc_expand_prologue): Likewise.
5585 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5587 (arc_verify_short): Likewise.
5588 (arc_attr_type): Likewise.
5589 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5590 (walk_stores): Likewise.
5591 (arc_address_cost): Make it static.
5592 (arc_verify_short): Likewise.
5593 (branch_dest): Likewise.
5594 (arc_attr_type): Likewise.
5595 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5596 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5597 (arc_final_prescan_insn): Remove inserting the nops due to
5598 hardware hazards. It is done in reorg step.
5599 (insn_length_variant_t): Remove.
5600 (insn_length_parameters_t): Likewise.
5601 (arc_insn_length_parameters): Likewise.
5602 (arc_get_insn_variants): Likewise.
5603 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5605 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5608 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5611 PR tree-optimization/84687
5612 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5614 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5616 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5618 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5619 Rename to ppc_speculation_barrier.
5620 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5621 __builtin_ppc_speculation_barrier.
5623 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5626 * combine.c (combine_simplify_rtx): Don't try to simplify if
5627 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5630 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5632 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5633 to 32 bytes when compiling for POWER9.
5635 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5638 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5639 regparm >= 3 with no arg reg available also for calls with
5640 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5643 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5645 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5647 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5650 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5652 2018-03-05 Richard Biener <rguenther@suse.de>
5654 PR tree-optimization/84486
5655 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5656 When inserting a __builtin_assume_aligned call set the LHS
5657 SSA name alignment info accordingly.
5659 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5661 PR tree-optimization/84114
5662 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5663 Avoid reassociation of FLOAT_MODE addition.
5665 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5667 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5668 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5669 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5670 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5671 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5672 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5673 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5675 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5676 __builtin_ia32_wbinvd): New builtins.
5677 (SPECIAL_ARGS2): New.
5678 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5679 (SPECIAL_ARGS2): New.
5680 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5681 (ix86_valid_target_attribute_inner_p): Ditto.
5682 (ix86_init_mmx_sse_builtins): Add special_args2.
5683 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5684 TARGET_WBNOINVD_P): New.
5685 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5686 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5687 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5688 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5689 * config/i386/pconfigintrin.h: New file.
5690 * config/i386/wbnoinvdintrin.h: Ditto.
5691 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5692 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5694 2018-03-05 Richard Biener <rguenther@suse.de>
5696 PR tree-optimization/84670
5697 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5699 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5700 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5701 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5702 doesn't grow if all successors (recursively) were visited at least
5705 2018-03-05 Richard Biener <rguenther@suse.de>
5707 PR tree-optimization/84650
5708 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5709 if executed in the loop pipeline.
5711 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5713 * doc/configfiles.texi (Configuration Files): Move info about
5714 conditionalizing $target-protos.h to...
5715 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5716 differs from $target-protos.h.
5718 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5719 Chung-Ju Wu <jasonwucj@gmail.com>
5721 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5722 * config/nds32/nds32-multiple.md (setmemsi): Define.
5723 * config/nds32/nds32-memory-manipulation.c
5724 (nds32_gen_dup_4_byte_to_word_value): New.
5725 (emit_setmem_word_loop): New.
5726 (emit_setmem_byte_loop): New.
5727 (nds32_expand_setmem_loop): New.
5728 (nds32_expand_setmem_loop_v3m): New.
5729 (nds32_expand_setmem_unroll): New.
5730 (nds32_expand_setmem): New.
5732 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5733 Chung-Ju Wu <jasonwucj@gmail.com>
5735 * config/nds32/nds32-memory-manipulation.c
5736 (nds32_emit_load_store): New.
5737 (nds32_emit_post_inc_load_store): New.
5738 (nds32_emit_mem_move): New.
5739 (nds32_emit_mem_move_block): New.
5740 (nds32_expand_movmemsi_loop_unknown_size): New.
5741 (nds32_expand_movmemsi_loop_known_size): New.
5742 (nds32_expand_movmemsi_loop): New.
5743 (nds32_expand_movmemsi_unroll): New.
5744 (nds32_expand_movmemqi): Rename ...
5745 (nds32_expand_movmemsi): ... to this.
5746 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5747 (movmemsi): ... to this.
5748 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5749 (nds32_expand_movmemsi): ... to this.
5751 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5752 Monk Chiang <sh.chiang04@gmail.com>
5753 Chung-Ju Wu <jasonwucj@gmail.com>
5755 * config/nds32/nds32-protos.h
5756 (nds32_expand_load_multiple): New arguments.
5757 (nds32_expand_store_multiple): Ditto.
5758 (nds32_valid_multiple_load_store): Rename ...
5759 (nds32_valid_multiple_load_store_p): ... to this.
5760 * config/nds32/nds32-memory-manipulation.c
5761 (nds32_expand_load_multiple): Refine implementation.
5762 (nds32_expand_store_multiple): Ditto.
5763 * config/nds32/nds32-multiple.md
5764 (load_multiple): Update nds32_expand_load_multiple interface.
5765 (store_multiple): Update nds32_expand_store_multiple interface.
5766 * config/nds32/nds32-predicates.c
5767 (nds32_valid_multiple_load_store): Rename ...
5768 (nds32_valid_multiple_load_store_p): ... to this and refine
5770 * config/nds32/predicates.md
5771 (nds32_load_multiple_and_update_address_operation): New predicate.
5772 (nds32_store_multiple_and_update_address_operation): New predicate.
5774 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5775 Chung-Ju Wu <jasonwucj@gmail.com>
5777 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5778 (combo): New attribute.
5779 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5781 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5783 * config/nds32/nds32.opt: Change -mcmodel= default value.
5785 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5786 Monk Chiang <sh.chiang04@gmail.com>
5787 Chung-Ju Wu <jasonwucj@gmail.com>
5789 * config/nds32/constants.md (unspec_element): New enum.
5790 * config/nds32/constraints.md (Umw): New constraint.
5791 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5792 * config/nds32/nds32-intrinsic.md: Likewise.
5793 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5794 (nds32_valid_smw_lwm_base_p): New.
5795 (nds32_output_smw_single_word): New.
5796 (nds32_output_lmw_single_word): New.
5797 (nds32_expand_unaligned_load): New.
5798 (nds32_expand_unaligned_store): New.
5799 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5800 (nds32_output_smw_single_word): Declare.
5801 (nds32_output_lmw_single_word): Declare.
5802 (nds32_expand_unaligned_load): Declare.
5803 (nds32_expand_unaligned_store): Declare.
5804 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5805 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5806 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5807 NDS32_BUILTIN_UASTORE_DW.
5808 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5811 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5812 Kito Cheng <kito.cheng@gmail.com>
5813 Chung-Ju Wu <jasonwucj@gmail.com>
5815 * config/nds32/nds32-intrinsic.c
5816 (nds32_expand_builtin_null_ftype_reg): Delete.
5817 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5818 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5819 (nds32_read_argument): New.
5820 (nds32_legitimize_target): Ditto.
5821 (nds32_legitimize_argument): Ditto.
5822 (nds32_check_constant_argument): Ditto.
5823 (nds32_expand_unop_builtin): Ditto.
5824 (nds32_expand_unopimm_builtin): Ditto.
5825 (nds32_expand_binop_builtin): Ditto.
5826 (nds32_builtin_decl_impl): Ditto.
5827 (builtin_description): Ditto.
5828 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5829 (nds32_init_builtins_impl): Ditto.
5830 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5831 (nds32_builtin_decl): New.
5832 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5833 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5835 2018-03-02 Jeff Law <law@redhat.com>
5837 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5838 (redundant_insn, fill_simple_delay_slots): Likewise.
5839 (fill_slots_from_thread): Likewise.
5840 * resource.c (mark_referenced_resources): Likewise.
5841 (mark_set_resources, find_dead_or_set_registers): Likewise.
5843 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5845 * substring-locations.h (format_warning_va): Formatting fix for
5847 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5849 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5850 * substring-locations.c: Include intl.h.
5851 (format_warning_va): Turned into small wrapper around
5852 format_warning_n_va, renamed to ...
5853 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5854 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5856 (format_warning_at_substring_n): New function.
5857 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5858 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5859 format_warning_at_substring with just a shorter name instead of
5860 const function pointer.
5861 (fmtwarn_n): New function.
5862 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5863 appropriate, get rid of all the fmtstr temporaries, move conditionals
5864 with G_() wrapped string literals directly into fmtwarn arguments,
5865 cast dir.len to (int), formatting fixes.
5867 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5869 * doc/invoke.texi: Remove "Cilk Plus" references.
5871 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5872 Richard Biener <rguenther@suse.de>
5875 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5876 for error or warning attributes if CALL_FROM_THUNK_P is set.
5879 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5882 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5883 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5886 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5887 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5889 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5890 instead of -1U in last predictors element's probability member.
5892 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5895 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5896 arguments if they are comparable.
5898 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5900 PR tree-optimization/84634
5901 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5902 masks and masked_loop_p with a single loop_masks, making sure it's
5903 null for bb vectorization.
5905 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5907 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5908 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5909 loop->force_vectorize to check whether there is no alias.
5911 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5914 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5916 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5918 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5919 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5920 instead of a loop around prev_real_insn.
5921 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5925 * config/i386/i386.c (ix86_print_operand): Use conditional
5926 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5929 2018-03-02 Richard Biener <rguenther@suse.de>
5931 PR tree-optimization/84427
5932 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5933 (bitmap_set_subtract_values): Rewrite to handle multiple
5936 (prune_clobbered_mems): Likewise.
5937 (phi_translate): Take edge instead of pred/phiblock.
5938 (phi_translate_1): Likewise.
5939 (phi_translate_set): Likewise. Insert all translated
5940 exprs for a value into the set, keeping possibly multiple
5941 expressions per value.
5942 (compute_antic_aux): Adjust for phi_translate changes.
5943 When intersecting union the expressions and prune those
5944 not in the final value set, keeping possibly multiple
5945 expressions per value. Do not use value-insertion
5946 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5947 all expressions. Add verification that the value-sets
5948 only shrink during iteration.
5949 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5950 (do_pre_regular_insertion): Likewise.
5951 (do_pre_partial_partial_insertion): Likewise.
5953 2018-03-02 Richard Biener <rguenther@suse.de>
5956 * config/darwin.c (saved_debug_info_level): New static global.
5957 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5958 (darwin_asm_lto_end): Restore debug info generation settings.
5960 2018-03-01 Martin Liska <mliska@suse.cz>
5963 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5966 2018-03-01 Richard Biener <rguenther@suse.de>
5969 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5970 type annotation with LTO when debug was disabled at compile-time.
5972 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5974 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5976 (mips_final_postscan_insn): Likewise.
5978 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5980 PR rtl-optimization/84528
5981 * alias.c (init_alias_target): Add commentary.
5982 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5983 a unique base value if the frame pointer is not eliminated
5984 to the stack pointer.
5986 2018-03-01 Tom de Vries <tom@codesourcery.com>
5988 PR rtl-optimization/83327
5989 * lra-int.h (hard_regs_spilled_into): Declare.
5990 * lra.c (hard_regs_spilled_into): Define.
5991 (init_reg_info): Init hard_regs_spilled_into.
5992 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5993 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5994 (process_bb_lives): Handle hard_regs_spilled_into.
5995 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5996 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5998 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6000 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6001 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6002 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6003 * config/rs6000/aix72.h: New file.
6005 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6007 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6008 instead of warning_at with conditional singular and plural messages
6012 * stor-layout.c (update_alignment_for_field): For
6013 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6014 && !DECL_PACKED (field), do the alignment update, just use
6015 only desired_align instead of MAX (type_align, desired_align)
6017 (place_field): Don't do known_align < desired_align handling
6018 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6019 is non-NULL, instead do it after rli->prev_field handling and
6020 only if not within a bitfield word. For DECL_PACKED (field)
6021 use type_align of BITS_PER_UNIT.
6023 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6025 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6026 superfluous parentheses and trailing spaces.
6028 2018-02-28 Richard Biener <rguenther@suse.de>
6030 PR tree-optimization/84584
6031 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6032 SCoPs with fake exit edge.
6034 2018-02-28 Martin Liska <mliska@suse.cz>
6037 * timevar.c (timer::print): Fix format to properly print 100%
6040 2018-02-28 Richard Biener <rguenther@suse.de>
6043 * genmatch.c (capture_info::walk_match): Do not mark
6044 captured expressions without operands as expr_p given
6045 they act more like predicates and should be subject to
6046 "lost tail" side-effect preserving.
6048 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6050 PR rtl-optimization/81611
6051 * auto-inc-dec.c (attempt_change): Move dead note from
6052 mem_insn if it's the next use of regno
6053 (find_address): Take address use of reg holding
6054 non-incremented value. Add parm to limit search to the named
6056 (merge_in_block): Attempt to use a mem insn that is the next
6057 use of the original regno.
6059 2018-02-27 Martin Sebor <msebor@redhat.com>
6062 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6063 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6065 2018-02-27 Martin Sebor <msebor@redhat.com>
6067 PR translation/84207
6068 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6069 n argument to unsigned HOST_WIDE_INT.
6070 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6071 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6072 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6073 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6075 2018-02-27 Richard Biener <rguenther@suse.de>
6077 PR tree-optimization/84512
6078 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6079 Do not use the estimate returned from record_stmt_cost for
6080 the scalar iteration cost but sum properly using add_stmt_cost.
6082 2018-02-27 Richard Biener <rguenther@suse.de>
6084 PR tree-optimization/84466
6085 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6086 Adjust last change to less strictly validate use operands.
6088 2018-02-27 Martin Liska <mliska@suse.cz>
6090 PR gcov-profile/84548
6091 * gcov.c (process_file): Allow partial overlap and consider it
6092 also as group functions.
6093 (output_lines): Properly calculate range of lines for a group.
6095 2018-02-27 Martin Liska <mliska@suse.cz>
6097 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6098 'ggc' suffixes. Change first column width.
6099 (timer::print): Fix formatting of the column.
6101 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6103 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6104 preserve inline entry blocks for the sake of debug inline
6105 entry point markers alone.
6106 (remove_unused_locals): Suggest in comments a better place to
6107 force the preservation of inline entry blocks that are
6108 otherwise unused, but do not preserve them.
6110 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6112 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6114 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6117 * config/i386/constraints.md (Bs): Replace
6118 ix86_indirect_branch_register with
6119 TARGET_INDIRECT_BRANCH_REGISTER.
6121 * config/i386/i386.md (indirect_jump): Likewise.
6122 (tablejump): Likewise.
6123 (*sibcall_memory): Likewise.
6124 (*sibcall_value_memory): Likewise.
6125 Peepholes of indirect call and jump via memory: Likewise.
6126 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6127 (*sibcall_value_GOT_32): Likewise.
6128 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6129 (GOT_memory_operand): Likewise.
6130 (call_insn_operand): Likewise.
6131 (sibcall_insn_operand): Likewise.
6132 (GOT32_symbol_operand): Likewise.
6133 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6135 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6137 PR rtl-optimization/83496
6138 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6139 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6140 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6141 redundant insn, if any.
6142 (relax_delay_slots): Likewise.
6143 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6145 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6147 PR tree-optimization/83965
6148 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6149 that grouped statements are part of a reduction chain. Return
6150 true if the statement is not marked as a reduction itself but
6152 (vect_recog_dot_prod_pattern): Don't check whether the statement
6153 is part of a group here.
6154 (vect_recog_sad_pattern): Likewise.
6155 (vect_recog_widen_sum_pattern): Likewise.
6157 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6160 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6162 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6165 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6167 (ix86_output_indirect_function_return): New prototype.
6168 (ix86_split_simple_return_pop_internal): Likewise.
6169 * config/i386/i386.c (indirect_return_via_cx): New.
6170 (indirect_return_via_cx_bnd): Likewise.
6171 (indirect_thunk_name): Handle return va CX_REG.
6172 (output_indirect_thunk_function): Create alias for
6173 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6174 (ix86_output_indirect_jmp): Remove the bool argument.
6175 (ix86_output_indirect_function_return): New function.
6176 (ix86_split_simple_return_pop_internal): Likewise.
6177 * config/i386/i386.md (*indirect_jump): Don't pass false
6178 to ix86_output_indirect_jmp.
6179 (*tablejump_1): Likewise.
6180 (simple_return_pop_internal): Change it to define_insn_and_split.
6181 Call ix86_split_simple_return_pop_internal to split it for
6183 (simple_return_indirect_internal): Call
6184 ix86_output_indirect_function_return instead of
6185 ix86_output_indirect_jmp.
6187 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6190 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6191 memset and value initialization afterwards.
6193 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6195 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6197 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6200 * common/config/aarch64/aarch64-common.c
6201 (aarch_option_optimization_table[]): Switch
6202 off fomit-frame-pointer
6204 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6205 Chung-Ju Wu <jasonwucj@gmail.com>
6207 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6209 (store_multiple): Ditto.
6211 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6213 * config.gcc: Add --with-cpu support for nds32 target.
6214 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6215 * config/nds32/nds32.opt: Add -mcpu= option.
6217 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6219 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6220 isel=yes): Warn for these deprecated options.
6222 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6224 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6225 ISA_2_5_MASKS_EMBEDDED.
6227 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6229 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6230 p->max as pointers rather than using iterative_hash_expr.
6232 2018-02-23 Carl Love <cel@us.ibm.com>
6234 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6235 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6237 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6238 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6239 P8V_BUILTIN_VEC_VUNSIGNED2.
6241 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6244 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6245 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6247 * lra-constraints.c (curr_insn_transform): Set up
6248 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6250 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6251 * lra-eliminations.c (spill_pseudos): Ditto.
6252 (process_insn_for_elimination): Ditto.
6253 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6254 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6255 LRA_NON_CLOBBERED_ALT.
6257 2018-02-22 Martin Sebor <msebor@redhat.com>
6259 PR tree-optimization/84480
6260 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6261 to maybe_diag_stxncpy_trunc. Call it.
6262 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6263 from gimple_fold_builtin_strcpy. Print inlining stack.
6264 (handle_builtin_stxncpy): Print inlining stack.
6265 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6267 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6270 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6271 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6272 and -fcheck-pointer-bounds are used together.
6273 (indirect_thunk_prefix): New enum.
6274 (indirect_thunk_need_prefix): New function.
6275 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6276 "_nt" instead of "_bnd" for NOTRACK prefix.
6277 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6278 (output_indirect_thunk_function): Likewise.
6280 (ix86_code_end): Update output_indirect_thunk_function calls.
6281 (ix86_output_indirect_branch_via_reg): Replace
6282 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6283 (ix86_output_indirect_branch_via_push): Likewise.
6284 (ix86_output_function_return): Likewise.
6285 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6286 incompatible with -fcf-protection=branch and
6287 -fcheck-pointer-bounds.
6289 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6292 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6293 Change gcc_assert call to output_operand_lossage.
6295 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6297 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6299 2018-02-22 DJ Delorie <dj@redhat.com>
6300 Sebastian Perta <sebastian.perta@renesas.com>
6301 Oleg Endo <olegendo@gcc.gnu.org>
6303 * config/rx/rx.c (rx_rtx_costs): New function.
6304 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6306 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6308 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6310 2018-02-22 Martin Liska <mliska@suse.cz>
6313 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6314 Add "native" as a possible value.
6316 2018-02-22 Martin Liska <mliska@suse.cz>
6319 * config/i386/i386.c (ix86_option_override_internal):
6320 Add "native" as a possible value for -march and -mtune.
6322 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6325 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6326 to all type variants.
6328 PR tree-optimization/84503
6329 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6330 width as info->bitpos + info->bitsize - start.
6331 (merged_store_group::merge_overlapping): Simplify width computation.
6332 (check_no_overlap): New function.
6333 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6334 start + width and last_order of the group, fail if check_no_overlap
6336 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6337 to group if check_no_overlap fails.
6339 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6341 * config/rs6000/altivec.md: Delete contraint arguments to
6342 define_expand, define_split, and define_peephole2, and in
6343 define_insn_and_split if always unused.
6344 * config/rs6000/darwin.md: Ditto.
6345 * config/rs6000/dfp.md: Ditto.
6346 * config/rs6000/rs6000.md: Ditto.
6347 * config/rs6000/sync.md: Ditto.
6348 * config/rs6000/vector.md: Ditto.
6349 * config/rs6000/vsx.md: Ditto.
6351 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6353 * config/rs6000/altivec.md: Write output control strings as braced
6354 blocks instead of double-quoted strings.
6355 * config/rs6000/darwin.md: Ditto.
6356 * config/rs6000/rs6000.md: Ditto.
6357 * config/rs6000/vector.md: Ditto.
6358 * config/rs6000/vsx.md: Ditto.
6360 2018-02-21 Jason Merrill <jason@redhat.com>
6362 PR c++/84314 - ICE with templates and fastcall attribute.
6363 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6365 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6367 * ipa-cp.c (determine_versionability): Fix comment typos.
6369 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6372 * ipa-cp.c (determine_versionability): Do not version functions caling
6375 2018-02-21 Martin Liska <mliska@suse.cz>
6378 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6379 Add "native" as a possible value.
6380 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6381 the macro when native cpu detection is available.
6383 2018-02-21 Martin Liska <mliska@suse.cz>
6386 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6387 Add "native" as a possible value.
6388 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6389 when native cpu detection is available.
6391 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6392 Martin Sebor <msebor@redhat.com>
6394 PR tree-optimization/84478
6395 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6397 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6398 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6399 support which is conservatively correct, for 2 only stay conservative
6400 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6401 argument to the 2 argument get_range_strlen, adjust 6 arg
6402 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6404 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6405 (gimple_fold_builtin_strlen): Pass true as last argument to
6408 2018-02-20 Martin Sebor <msebor@redhat.com>
6411 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6412 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6413 (builtin_memref::builtin_memref): Factor out parts into
6414 set_base_and_offset and call it.
6416 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6419 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6420 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6421 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6422 search at the associated MODE_INT.
6424 2018-02-20 Jeff Law <law@redhat.com>
6427 PR tree-optimization/81592
6429 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6430 for range data rather than using global data.
6431 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6432 range data rather than using global data.
6433 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6434 pass it to children as needed.
6435 (struct directive::fmtresult): Similarly.
6436 (struct directive::set_width): Similarly.
6437 (struct directive::set_precision): Similarly.
6438 (format_integer, format_directive, parse_directive): Similarly.
6439 (format_none): Accept unnamed vr_values parameter.
6440 (format_percent, format_floating, format_character): Similarly.
6441 (format_string, format_plain): Similarly.
6442 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6443 the EVRP range analyzer for range data rather than using global data.
6444 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6445 gimple-ssa-evrp-analyze.h
6446 (class sprintf_dom_walker): Add after_dom_children member function.
6447 Add evrp_range_analyzer member.
6448 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6449 range analyzer as needed.
6450 (sprintf_dom_walker::after_dom_children): New member function.
6451 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6453 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6454 (evrp_range_analyzer::pop_to_marker): Likewise.
6456 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6458 PR tree-optimization/84419
6459 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6460 with the required type if its current type is compatible but
6463 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6466 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6467 after vectorization.
6469 2018-02-20 Martin Liska <mliska@suse.cz>
6472 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6473 possible values if we don't have a hint.
6475 2018-02-20 Martin Liska <mliska@suse.cz>
6479 * final.c (shorten_branches): Build align_tab array with one
6481 * opts.c (finish_options): Add alignment option limit check.
6482 (MAX_CODE_ALIGN): Likewise.
6483 (MAX_CODE_ALIGN_VALUE): Likewise.
6484 * doc/invoke.texi: Document maximum allowed option value for
6485 all -falign-* options.
6487 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6490 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6491 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6492 * var-tracking.c (emit_note_insn_var_location): Remove all references
6493 to NOTE_INSN_CALL_ARG_LOCATION.
6494 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6495 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6496 Use copy_rtx_if_shared.
6497 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6498 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6499 (dwarf2out_var_location): Remove handling of
6500 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6502 * final.c (final_scan_insn): Remove all references to
6503 NOTE_INSN_CALL_ARG_LOCATION.
6504 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6505 before dumping final insns.
6506 * except.c (emit_note_eh_region_end): Remove all references to
6507 NOTE_INSN_CALL_ARG_LOCATION.
6508 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6509 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6510 * config/arc/arc.c (hwloop_optimize): Likewise.
6511 * config/arm/arm.c (create_fix_barrier): Likewise.
6512 * config/s390/s390.c (s390_chunkify_start): Likewise.
6513 * config/sh/sh.c (find_barrier): Likewise.
6514 * config/i386/i386.c (rest_of_insert_endbranch,
6515 ix86_seh_fixup_eh_fallthru): Likewise.
6516 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6517 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6518 * config/frv/frv.c (frv_function_prologue): Likewise.
6519 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6521 (note_outside_basic_block_p): Remove all references to
6522 NOTE_INSN_CALL_ARG_LOCATION.
6523 * gengtype.c (adjust_field_rtx_def): Likewise.
6524 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6526 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6527 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6530 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6533 PR tree-optimization/84452
6534 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6535 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6538 2018-02-19 Martin Liska <mliska@suse.cz>
6541 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6543 2018-02-19 Martin Liska <mliska@suse.cz>
6544 Richard Sandiford <richard.sandiford@linaro.org>
6546 PR tree-optimization/82491
6547 * gimple-fold.c (get_base_constructor): Make earlier bail out
6550 2018-02-19 Carl Love <cel@us.ibm.com>
6552 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6553 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6555 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6556 P8V_BUILTIN_VEC_NEG.
6558 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6560 * config/rl78/rl78.md (movdf): New define expand.
6562 2018-02-19 Martin Liska <mliska@suse.cz>
6565 * doc/invoke.texi: Fix typo.
6566 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6568 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6570 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6571 handle rs6000_single_float and rs6000_double_float specially for
6574 2018-02-16 Jeff Law <law@redhat.com>
6576 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6579 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6582 * ipa-inline.c (inline_small_functions): Fix a typo.
6584 2018-02-16 Nathan Sidwell <nathan@acm.org>
6586 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6588 2018-02-16 Carl Love <cel@us.ibm.com>
6590 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6591 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6592 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6593 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6594 expansion to P8V_BUILTIN_VEC_FLOAT2.
6596 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6598 PR rtl-optimization/70023
6599 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6600 src_regno into account.
6602 2018-02-16 Carl Love <cel@us.ibm.com>
6604 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6605 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6606 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6607 * config/rs6000/rs6000.c: Remove case statements for
6608 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6609 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6610 and P9V_BUILTIN_VEC_VINSERT4B.
6611 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6612 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6613 * config/rs6000/vsx.md:
6614 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6617 2018-02-16 Carl Love <cel@us.ibm.com>
6619 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6621 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6623 * config/rs6000/rs6000-c.c: Add the definitions for
6624 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6625 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6626 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6627 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6628 definition for insert4b and define insn *insert3b_internal.
6629 * doc/extend.texi: Add documentation for vec_extract4b.
6631 2018-02-16 Nathan Sidwell <nathan@acm.org>
6633 * doc/extend.texi (Backwards Compatibility): Mention friend
6634 injection. Note for-scope is deprecated.
6635 * doc/invoke.texi (-ffriend-injection): Deprecate.
6637 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6639 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6640 that moved to I2, also allow destinations that are a paradoxical
6641 subreg (instead of a normal reg).
6643 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6646 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6649 2018-02-16 Richard Biener <rguenther@suse.de>
6651 PR tree-optimization/84037
6652 PR tree-optimization/84016
6654 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6655 Adjust vec_construct for the fact we need additional higher latency
6656 128bit inserts for AVX256 and AVX512 vector builds.
6657 (ix86_add_stmt_cost): Scale vector construction cost for
6660 2018-02-16 Richard Biener <rguenther@suse.de>
6662 PR tree-optimization/84417
6663 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6664 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6665 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6667 2018-02-16 Martin Liska <mliska@suse.cz>
6670 * internal-fn.def (ASAN_CHECK): Set proper flags.
6671 (ASAN_MARK): Likewise.
6673 2018-02-16 Julia Koval <julia.koval@intel.com>
6675 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6676 from PTA_CANNONLAKE.
6678 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6681 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6682 Use ++iter rather than iter++ for std::list iterators.
6683 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6684 defer deleting them until all nodes in the forest are processed. Do
6685 free even leaf nodes. Change to_process into auto_vec.
6688 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6689 * vec.h (vec_default_construct): Use memset instead of placement new
6690 if BROKEN_VALUE_INITIALIZATION is defined.
6691 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6692 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6695 PR rtl-optimization/83723
6696 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6697 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6698 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6700 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6702 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6704 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6706 PR rtl-optimization/81443
6707 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6708 from inner REGs to paradoxical SUBREGs.
6710 2018-02-16 Richard Biener <rguenther@suse.de>
6712 PR tree-optimization/84399
6713 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6714 For operands we can analyze at their definition make sure we can
6715 analyze them at each use as well.
6717 2018-02-16 Richard Biener <rguenther@suse.de>
6719 PR tree-optimization/84190
6720 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6721 volatile accesses if the decl isn't volatile.
6723 2018-02-15 Jason Merrill <jason@redhat.com>
6725 PR c++/84314 - ICE with templates and fastcall attribute.
6726 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6727 TYPE_CANONICAL on an existing type.
6729 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6731 PR tree-optimization/84383
6732 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6733 dstoff nor call operand_equal_p if dstbase is NULL.
6735 PR tree-optimization/84334
6736 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6737 also a CONSTANT_CLASS_P, punt.
6739 2018-02-14 Jim Wilson <jimw@sifive.com>
6741 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6742 first SMALL_OPERAND check. New local min_second_step. Move assert
6743 to where locals are set. Add TARGET_RVC support.
6744 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6746 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6748 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6750 2018-02-14 Martin Sebor <msebor@redhat.com>
6752 PR tree-optimization/83698
6753 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6754 arrays constrain the offset range to their bounds.
6755 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6756 (builtin_access::overlap): Avoid setting the size of overlap if it's
6758 (maybe_diag_overlap): Also consider arrays when deciding what values
6759 of offsets to include in diagnostics.
6761 2018-02-14 Martin Sebor <msebor@redhat.com>
6764 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6765 that correspond to the kind of a declaration.
6767 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6770 * config/pa/pa.md: Load address of PIC label using the linkage table
6771 if the label is nonlocal.
6773 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6775 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6776 warning message if user requests -maltivec=be.
6777 * doc/invoke.texi: Document deprecation of -maltivec=be.
6779 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6782 * config/rs6000/rs6000-c.c: Update definitions for
6783 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6784 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6786 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6789 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6790 add _get_ssp intrinsics. Remove argument from
6791 __builtin_ia32_rdssp[d|q].
6792 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6793 * config/i386/i386-builtin.def: Remove argument from
6794 __builtin_ia32_rdssp[d|q].
6795 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6796 ix86_expand_special_args_builtin for _rdssp[d|q].
6797 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6798 Clear register before usage.
6799 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6800 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6802 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6804 PR tree-optimization/84357
6805 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6806 operand 1 of an ARRAY_REF too.
6808 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6811 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6812 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6814 (set_of_reg): New struct.
6815 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6816 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6817 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6819 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6820 Split into bitclr, bitset, bitinvert patterns if appropriate.
6821 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6822 use rx_fuse_in_memory_bitop.
6823 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6824 to named insn, correct maximum insn length.
6826 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6829 * machmode.def: Define a complex mode for PARTIAL_INT.
6830 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6832 * doc/rtl.texi: Document CSPImode.
6833 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6835 (msp430_hard_regno_nregs_with_padding): Likewise.
6837 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6840 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6842 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6844 PR rtl-optimization/84169
6845 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6846 we generated a parallel as new i3 and we split that to new i2 and i3
6847 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6848 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6849 those to i2, not i1. Partially rewrite this scan code.
6851 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6854 * stor-layout.c (place_field): For variable length fields, adjust
6855 offset_align afterwards not just based on the field's alignment,
6856 but also on the size.
6859 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6860 of exps and logs in the use_exp2 case.
6862 2018-02-13 Jeff Law <law@redhat.com>
6864 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6867 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6870 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6874 * common.opt (gas-loc-support, gas-locview-support): New.
6875 (ginline-points, ginternal-reset-location-views): New.
6876 * doc/invoke.texi: Document them. Use @itemx where intended.
6877 (gvariable-location-views): Adjust.
6878 * target.def (reset_location_view): New.
6879 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6880 (TARGET_RESET_LOCATION_VIEW): New.
6881 * doc/tm.texi: Rebuilt.
6882 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6883 (dwarf2out_default_as_locview_support): New.
6884 (output_asm_line_debug_info): Use option variables.
6885 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6886 (output_loc_list): Likewise.
6887 (add_high_low_attributes): Check option variables.
6888 Don't output entry view attribute in strict mode.
6889 (gen_inlined_subroutine_die): Check option variables.
6890 (dwarf2out_inline_entry): Likewise.
6891 (init_sections_and_labels): Likewise.
6892 (dwarf2out_early_finish): Likewise.
6893 (maybe_reset_location_view): New, from...
6894 (dwarf2out_var_location): ... here. Call it.
6895 * debug.h (dwarf2out_default_as_loc_support): Declare.
6896 (dwarf2out_default_as_locview_support): Declare.
6897 * hooks.c (hook_int_rtx_insn_0): New.
6898 * hooks.h (hook_int_rtx_insn_0): Declare.
6899 * toplev.c (process_options): Take -gas-loc-support and
6900 -gas-locview-support from dwarf2out. Enable
6901 -gvariable-location-views by default only with locview
6902 assembler support. Enable -ginternal-reset-location-views by
6903 default only if the target defines the corresponding hook.
6904 Enable -ginline-points by default if location views are
6905 enabled; force it disabled if statement frontiers are
6907 * tree-inline.c (expand_call_inline): Check option variables.
6908 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6910 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6912 PR tree-optimization/84321
6913 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6914 handling. Also check whether the anti-range contains any values
6915 that satisfy the mask; switch to a VR_RANGE if not.
6917 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6920 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6922 2018-02-13 Martin Jambor <mjambor@suse.cz>
6925 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6926 of call statements, also set location of a load to a temporary.
6928 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6930 * config/rl78/rl78.c (add_vector_labels): New function.
6931 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6932 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6933 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6934 which checks that no arguments are passed.
6935 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6936 * doc/extend.texi: Documentation for the new attribute.
6938 2018-02-13 Andreas Schwab <schwab@suse.de>
6940 * config/riscv/linux.h (CPP_SPEC): Define.
6942 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6945 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6946 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6947 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6948 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6949 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6950 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6951 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6952 temporarily for AES and PCLMUL builtins.
6954 PR tree-optimization/84339
6955 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6956 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6960 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6961 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6962 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6964 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6966 * omp-simd-clone.h: New file.
6967 * omp-simd-clone.c: Include omp-simd-clone.h.
6968 (expand_simd_clones): No longer static.
6969 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6970 cgraph.h and omp-simd-clone.h.
6971 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6972 (vect_recog_widen_shift_pattern): Formatting fix.
6973 (vect_pattern_recog_1): Don't check optab for calls.
6976 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6977 operands[2] into a REG before using gen_lowpart on it.
6979 2018-02-12 Jeff Law <law@redhat.com>
6982 * config/sh/sh.c (find_barrier): Consider a sibling call
6985 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6986 successfully back substituting a reg.
6988 2018-02-12 Richard Biener <rguenther@suse.de>
6990 PR tree-optimization/84037
6991 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6992 parameter, move visited init to caller.
6993 (vect_slp_analyze_operations): Separate cost from validity
6994 check, initialize visited once for all instances.
6995 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6997 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6999 (vect_model_promotion_demotion_cost): Likewise.
7000 (vectorizable_bswap): Guard cost modeling with !slp_node
7001 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7003 (vectorizable_call): Likewise.
7004 (vectorizable_conversion): Likewise.
7005 (vectorizable_assignment): Likewise.
7006 (vectorizable_shift): Likewise.
7007 (vectorizable_operation): Likewise.
7008 (vectorizable_store): Likewise.
7009 (vectorizable_load): Likewise.
7010 (vectorizable_condition): Likewise.
7011 (vectorizable_comparison): Likewise.
7013 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7016 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7017 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7018 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7020 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7023 * params.def (inline-min-speedup): Increase from 8 to 15.
7024 (max-inline-insns-auto): Decrease from 40 to 30.
7025 * ipa-split.c (consider_split): Add some buffer for function to
7026 be considered inlining candidate.
7027 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7030 2018-02-12 Richard Biener <rguenther@suse.de>
7032 PR tree-optimization/84037
7033 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7034 matched stmts if we cannot swap the non-matched ones.
7036 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7038 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7039 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7040 _mm_maskz_scalef_round_ss): New intrinsics.
7041 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7042 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7043 __builtin_ia32_scalefss_round): Remove.
7044 (__builtin_ia32_scalefsd_mask_round,
7045 __builtin_ia32_scalefss_mask_round): New intrinsics.
7046 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7047 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7048 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7049 "<round_constraint>")): Changed to ...
7050 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7051 "<round_scalar_constraint>")): ... this.
7052 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7053 %0, %1, %2<round_op3>}"): Changed to ...
7054 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7055 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7056 %2<round_scalar_mask_op3>}"): ... this.
7057 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7059 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7061 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7062 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7063 (_mm_maskz_sqrt_round_ss): New intrinsics.
7064 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7065 (__builtin_ia32_sqrtsd_mask_round)
7066 (__builtin_ia32_sqrtss_mask_round): New builtins.
7067 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7068 (__builtin_ia32_sqrtss_round): Remove.
7069 (__builtin_ia32_sqrtsd_mask_round)
7070 (__builtin_ia32_sqrtss_mask_round): New builtins.
7071 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7072 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7073 ((match_operand:VF_128 1 "vector_operand"
7074 "xBm,<round_constraint>")): Changed to ...
7075 ((match_operand:VF_128 1 "vector_operand"
7076 "xBm,<round_scalar_constraint>")): ... this.
7077 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7078 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7079 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7080 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7081 %<iptr>1<round_scalar_mask_op3>}): ... this.
7082 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7083 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7085 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7088 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7089 Cast vec_cmpeq result to correct type.
7090 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7091 Cast vec_cmpgt result to correct type.
7093 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7095 * final.c (final_scan_insn_1): Renamed from...
7096 (final_scan_insn): ... this. New wrapper, to recover
7097 seen from the outermost call in recursive ones.
7098 * config/sparc/sparc.c (output_return): Drop seen from call.
7099 (output_sibcall): Likewise.
7100 * config/visium/visium.c (output_branch): Likewise.
7102 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7104 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7107 2018-02-10 Alan Modra <amodra@gmail.com>
7110 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7111 Specify LR as an input.
7113 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7116 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7117 remove_member_access_dummy_vars): New functions.
7118 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7119 lower_omp_1, execute_lower_omp): Use them.
7121 PR rtl-optimization/84308
7122 * shrink-wrap.c (spread_components): Release todo vector.
7124 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7126 PR rtl-optimization/57193
7127 * ira-color.c (struct allocno_color_data): Add member
7128 conflict_allocno_hard_prefs.
7129 (update_conflict_allocno_hard_prefs): New.
7130 (bucket_allocno_compare_func): Add a preference based on
7131 conflict_allocno_hard_prefs.
7132 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7133 (color_allocnos): Remove a dead code. Initiate
7134 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7136 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7139 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7140 constraint from =wa to wa. Avoid a subreg on the output operand,
7141 instead use a pseudo and subreg it in a move.
7142 (p9_xxbrd_<mode>): Changed to ...
7143 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7144 (p9_xxbrd_v2df): New expander.
7145 (p9_xxbrw_<mode>): Changed to ...
7146 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7147 (p9_xxbrw_v4sf): New expander.
7149 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7151 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7153 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7156 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7157 multiply in 32-bit mode.
7158 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7159 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7162 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7164 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7165 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7166 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7167 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7169 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7172 * dwarf2out.c (is_trivial_indirect_ref): New function.
7173 (dwarf2out_late_global_decl): Do not generate a location
7174 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7175 and that are not defined in the current unit.
7177 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7179 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7180 instead of a libcall for UNORDERED.
7182 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7185 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7186 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7188 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7191 * config/s390/s390.c (s390_set_current_function): Invoke
7192 s390_indirect_branch_settings also if fndecl didn't change.
7194 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7196 * config/rs6000/rs6000.md (blockage): Set length to zero.
7198 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7200 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7202 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7205 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7206 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7210 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7211 PARALLEL incoming that failed vt_get_decl_and_offset check.
7214 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7215 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7217 (get_variable_section): For decls in named .bss* sections pass true as
7218 second argument to bss_initializer_p.
7220 2018-02-09 Marek Polacek <polacek@redhat.com>
7221 Jakub Jelinek <jakub@redhat.com>
7224 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7225 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7226 Sync some changes from cxx_fold_indirect_ref.
7228 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7230 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7232 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7233 (BLOCK_INLINE_ENTRY_LABEL): New.
7234 (dwarf2out_var_location): Disregard inline entry markers.
7235 (inline_entry_data): New struct.
7236 (inline_entry_data_hasher): New hashtable type.
7237 (inline_entry_data_hasher::hash): New.
7238 (inline_entry_data_hasher::equal): New.
7239 (inline_entry_data_table): New variable.
7240 (add_high_low_attributes): Add DW_AT_entry_pc and
7241 DW_AT_GNU_entry_view attributes if a pending entry is found
7242 in inline_entry_data_table. Add old entry_pc attribute only
7243 if debug nonbinding markers are disabled.
7244 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7245 markers are enabled.
7246 (block_within_block_p, dwarf2out_inline_entry): New.
7247 (dwarf2out_finish): Check that no entries remained in
7248 inline_entry_data_table.
7249 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7250 (final_scan_insn, notice_source_line): Likewise.
7251 (rest_of_clean_state): Skip inline entry markers.
7252 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7254 * gimple.c (gimple_build_debug_inline_entry): New.
7255 * gimple.h (enum gimple_debug_subcode): Add
7256 GIMPLE_DEBUG_INLINE_ENTRY.
7257 (gimple_build_debug_inline_entry): Declare.
7258 (gimple_debug_inline_entry_p): New.
7259 (gimple_debug_nonbind_marker_p): Adjust.
7260 * insn-notes.def (INLINE_ENTRY): New.
7261 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7262 inline entry marker notes.
7263 (print_insn): Likewise.
7264 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7265 (INSN_DEBUG_MARKER_KIND): Likewise.
7266 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7267 * tree-inline.c (expand_call_inline): Build and insert
7268 debug_inline_entry stmt.
7269 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7270 inline entry blocks early, if nonbind markers are enabled.
7271 (dump_scope_block): Dump fragment info.
7272 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7273 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7274 (gimple_build_debug_inline_entry): New.
7275 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7276 Enable/disable inline entry points too.
7277 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7278 (DEBUG_INSN): Describe inline entry markers.
7280 * common.opt (gvariable-location-views): New.
7281 (gvariable-location-views=incompat5): New.
7282 * config.in: Rebuilt.
7283 * configure: Rebuilt.
7284 * configure.ac: Test assembler for view support.
7285 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7286 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7287 * dwarf2out.c (var_loc_view): New typedef.
7288 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7289 (dwarf2out_locviews_in_attribute): New.
7290 (dwarf2out_locviews_in_loclist): New.
7291 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7292 (enum dw_line_info_opcode): Add LI_adv_address.
7293 (struct dw_line_info_table): Add view.
7294 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7295 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7296 (zero_view_p): New variable.
7297 (ZERO_VIEW_P): New macro.
7298 (output_asm_line_debug_info): New.
7299 (struct var_loc_node): Add view.
7300 (add_AT_view_list, AT_loc_list): New.
7301 (add_var_loc_to_decl): Add view param. Test it against last.
7302 (new_loc_list): Add view params. Record them.
7303 (AT_loc_list_ptr): Handle loc and view lists.
7304 (view_list_to_loc_list_val_node): New.
7305 (print_dw_val): Handle dw_val_class_view_list.
7306 (size_of_die): Likewise.
7307 (value_format): Likewise.
7308 (loc_list_has_views): New.
7309 (gen_llsym): Set vl_symbol too.
7310 (maybe_gen_llsym, skip_loc_list_entry): New.
7311 (dwarf2out_maybe_output_loclist_view_pair): New.
7312 (output_loc_list): Output view list or entries too.
7313 (output_view_list_offset): New.
7314 (output_die): Handle dw_val_class_view_list.
7315 (output_dwarf_version): New.
7316 (output_compilation_unit_header): Use it.
7317 (output_skeleton_debug_sections): Likewise.
7318 (output_rnglists, output_line_info): Likewise.
7319 (output_pubnames, output_aranges): Update version comments.
7320 (output_one_line_info_table): Output view numbers in asm comments.
7321 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7322 Call maybe_gen_llsym.
7323 (loc_list_from_tree_1): Adjust.
7324 (add_AT_location_description): Create view list attribute if
7325 needed, check it's absent otherwise.
7326 (convert_cfa_to_fb_loc_list): Adjust.
7327 (maybe_emit_file): Call output_asm_line_debug_info for test.
7328 (dwarf2out_var_location): Reset views as needed. Precompute
7329 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7330 attribute. Set view.
7331 (new_line_info_table): Reset next view.
7332 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7333 (dwarf2out_source_line): Likewise. Output view resets and labels to
7334 the assembler, or select appropriate line info opcodes.
7335 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7336 (optimize_string_length): Catch it. Adjust.
7337 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7338 dw_val_class_view_list, and remove it if no longer needed.
7339 (hash_loc_list): Hash view numbers.
7340 (loc_list_hasher::equal): Compare them.
7341 (optimize_location_lists): Check whether a view list symbol is
7342 needed, and whether the locview attribute is present, and
7343 whether they match. Remove the locview attribute if no longer
7345 (index_location_lists): Call skip_loc_list_entry for test.
7346 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7347 Use output_dwarf_version.
7348 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7349 (struct dw_val_node): Add val_view_list.
7350 * final.c (SEEN_NEXT_VIEW): New.
7351 (set_next_view_needed): New.
7352 (clear_next_view_needed): New.
7353 (maybe_output_next_view): New.
7354 (final_start_function): Rename to...
7355 (final_start_function_1): ... this. Take pointer to FIRST,
7356 add SEEN parameter. Emit param bindings in the initial view.
7357 (final_start_function): Reintroduce SEEN-less interface.
7358 (final): Rename to...
7359 (final_1): ... this. Take SEEN parameter. Output final pending
7360 next view at the end.
7361 (final): Reintroduce seen-less interface.
7362 (final_scan_insn): Output pending next view before switching
7363 sections or ending a block. Mark the next view as needed when
7364 outputting variable locations. Notify debug backend of section
7365 changes, and of location view changes.
7366 (rest_of_handle_final): Adjust.
7367 * toplev.c (process_options): Autodetect value for debug variable
7368 location views option. Warn on incompat5 without -gdwarf-5.
7369 * doc/invoke.texi (gvariable-location-views): New.
7370 (gvariable-location-views=incompat5): New.
7371 (gno-variable-location-views): New.
7373 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7375 PR tree-optimization/84136
7376 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7377 that the result of find_edge is non-NULL.
7379 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7382 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7383 storing integer register in SImode. Fix cost of 256 and 512
7384 byte aligned SSE register store.
7386 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7388 * config/i386/i386.c (ix86_multiplication_cost): Fix
7389 multiplication cost for TARGET_AVX512DQ.
7391 2018-02-08 Marek Polacek <polacek@redhat.com>
7393 PR tree-optimization/84238
7394 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7397 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7399 PR tree-optimization/84265
7400 * tree-vect-stmts.c (vectorizable_store): Don't treat
7401 VMAT_CONTIGUOUS accesses as grouped.
7402 (vectorizable_load): Likewise.
7404 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7406 PR tree-optimization/81635
7407 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7408 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7409 (test_round_for_mask): New functions.
7410 (wide_int_cc_tests): Call test_round_for_mask.
7411 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7412 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7413 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7414 range returned by get_range_info.
7416 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7419 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7420 * symtab.c: Include builtins.h
7421 (symtab_node::output_to_lto_symbol_table_p): Move here
7422 from lto-streamer-out.c:output_symbol_p.
7423 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7424 (output_symbol_p): Move all logic to symtab.c
7425 (produce_symtab): Update.
7427 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7429 * config/s390/s390-opts.h (enum indirect_branch): Define.
7430 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7431 (s390_indirect_branch_via_thunk)
7432 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7433 (enum s390_indirect_branch_type): Define.
7434 * config/s390/s390.c (struct s390_frame_layout, struct
7435 machine_function): Remove.
7436 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7437 (indirect_branch_table_label_no, indirect_branch_table_name):
7439 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7440 (enum s390_indirect_branch_option): Define.
7441 (s390_return_addr_from_memory): New function.
7442 (s390_handle_string_attribute): New function.
7443 (s390_attribute_table): Add new attribute handler.
7444 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7445 (s390_indirect_branch_via_thunk): New function.
7446 (s390_indirect_branch_via_inline_thunk): New function.
7447 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7448 sibling call optimization for non z10 compiles.
7449 (s390_emit_call): Force indirect branch target to be a single
7450 register. Add r1 clobber for non-z10 compiles.
7451 (s390_emit_epilogue): Emit return jump via return_use expander.
7452 (s390_reorg): Handle JUMP_INSNs as execute targets.
7453 (s390_option_override_internal): Perform validity checks for the
7454 new command line options.
7455 (s390_indirect_branch_attrvalue): New function.
7456 (s390_indirect_branch_settings): New function.
7457 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7458 (s390_output_indirect_thunk_function): New function.
7459 (s390_code_end): Implement target hook.
7460 (s390_case_values_threshold): Implement target hook.
7461 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7463 * config/s390/s390.h (struct s390_frame_layout)
7464 (struct machine_function): Move here from s390.c.
7465 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7466 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7467 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7468 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7469 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7470 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7471 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7472 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7473 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7474 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7475 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7476 (mnemonic attribute): Add values which aren't recognized
7478 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7479 pattern for branch conversion. Fix mnemonic attribute.
7480 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7481 indirect branch via thunk if requested.
7482 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7483 ("*indirect_jump"): Disable for branch conversion using out of
7485 ("indirect_jump_via_thunk<mode>_z10")
7486 ("indirect_jump_via_thunk<mode>")
7487 ("indirect_jump_via_inlinethunk<mode>_z10")
7488 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7489 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7490 ("casesi_jump_via_inlinethunk<mode>_z10")
7491 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7492 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7493 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7494 ("*indirect2_jump"): Disable for branch conversion.
7495 ("casesi_jump"): Turn into expander and expand patterns for branch
7497 ("return_use"): New expander.
7498 ("*return"): Emit return via thunk and rename it to ...
7499 ("*return<mode>"): ... this one.
7500 * config/s390/s390.opt: Add new options and and enum for the
7503 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7505 * lra-constraints.c (match_reload): Unconditionally use
7506 gen_lowpart_SUBREG, rather than selecting between that
7507 and equivalent gen_rtx_SUBREG code.
7509 2018-02-08 Richard Biener <rguenther@suse.de>
7511 PR tree-optimization/84233
7512 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7513 changed flag instead of boguously re-using phi_inserted.
7515 2018-02-08 Martin Jambor <mjambor@suse.cz>
7517 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7518 static local variables.
7520 2018-02-08 Richard Biener <rguenther@suse.de>
7522 PR tree-optimization/84278
7523 * tree-vect-stmts.c (vectorizable_store): When looking for
7524 smaller vector types to perform grouped strided loads/stores
7525 make sure the mode is supported by the target.
7526 (vectorizable_load): Likewise.
7528 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7530 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7531 Increase LDP/STP opportunities by adding adjacent callee-saves.
7533 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7535 PR rtl-optimization/84068
7536 PR rtl-optimization/83459
7537 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7539 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7541 PR tree-optimization/84224
7542 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7543 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7546 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7549 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7550 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7551 count, remove one USE.
7553 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7555 * doc/install.texi (Configuration): Document the
7556 --with-long-double-format={ibm,ieee} PowerPC configuration
7560 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7561 Convert from define_expand to be define_insn_and_split. Rework
7562 float/double/_Float128 conversions to QI/HI/SImode to work with
7563 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7564 conversions to QI/HImode types did a store and then a load to
7565 truncate the value. For conversions to VSX registers, don't split
7566 the insn, instead emit the code directly. Use the code iterator
7567 any_fix to combine signed and unsigned conversions.
7568 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7569 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7570 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7571 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7572 (fix_<mode>di2_hw): Likewise.
7573 (fixuns_<mode>di2_hw): Likewise.
7574 (fix_<mode>si2_hw): Likewise.
7575 (fixuns_<mode>si2_hw): Likewise.
7576 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7577 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7578 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7579 fix<uns>_trunc<SFDF:mode>si2_p8.
7580 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7582 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7583 (fix<uns>_<mode>_mem): Likewise.
7584 (fctiw<u>z_<mode>_mem): Likewise.
7585 (fix<uns>_<mode>_mem): Likewise.
7586 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7587 the register allocator from doing a direct move to the GPRs to do
7588 a store, and instead use the ISA 3.0 store byte/half-word from
7589 vector register instruction. For IEEE 128-bit floating point,
7590 also optimize stores of 32-bit ints.
7591 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7593 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7595 * genextract.c (push_pathstr_operand): New function to support
7597 (walk_rtx): Call push_pathstr_operand.
7598 (print_path): Support [a-zA-Z].
7600 2018-02-07 Richard Biener <rguenther@suse.de>
7602 PR tree-optimization/84037
7603 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7604 (cse_and_gimplify_to_preheader): Declare.
7605 (vect_get_place_in_interleaving_chain): Likewise.
7606 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7608 (_loop_vec_info::~_loop_vec_info): Delete it.
7609 (cse_and_gimplify_to_preheader): New function.
7610 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7611 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7612 (vectorizable_load): Likewise. For grouped stores always base
7613 the IV on the first element.
7614 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7615 condition before gimplifying.
7617 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7619 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7620 *DIV_EXPR and *MOD_EXPR.
7622 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7625 * config/i386/i386.c (ix86_option_override_internal): Mask out
7626 the CF_SET bit when checking -fcf-protection.
7628 2018-02-07 Tom de Vries <tom@codesourcery.com>
7631 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7634 2018-02-07 Richard Biener <rguenther@suse.de>
7636 PR tree-optimization/84204
7637 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7640 PR tree-optimization/84205
7641 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7642 special-case isl_ast_op_zdiv_r.
7644 PR tree-optimization/84223
7645 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7646 Only add conditions from within the region.
7647 (gather_bbs::after_dom_children): Adjust.
7649 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7652 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7653 * config/avr/avr.md: Only post-reload split REG-REG moves if
7654 either register is GENERAL_REG_P.
7656 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7658 PR tree-optimization/84235
7659 * tree-ssa-scopedtables.c
7660 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7661 if the subtraction is performed in floating point type where NaNs are
7662 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7663 build 1. Formatting fix.
7665 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7668 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7669 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7670 and skip it regardless of bb boundaries. Use CALL_P macro,
7671 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7672 unnecessarily, formatting fix.
7674 2018-02-06 Michael Collison <michael.collison@arm.com>
7676 * config/arm/thumb2.md:
7677 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7678 (*thumb_mov_notscc): Ditto.
7680 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7683 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7684 unsigned_fix, not "s".
7686 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7688 * configure.ac (gcc_fn_eh_frame_ro): New function.
7689 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7690 correct .eh_frame permissions.
7691 * configure: Regenerate.
7693 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7695 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7698 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7700 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7701 Display warning message for -mno-speculate-indirect-jumps.
7703 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7705 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7707 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7709 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7711 PR tree-optimization/84225
7712 * tree-eh.c (find_trapping_overflow): Only call
7713 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7715 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7718 * config/i386/i386.c: Reimplement the check of possible options
7719 -mibt/-mshstk conbination. Change error messages.
7720 * doc/invoke.texi: Fix a typo: remove extra '='.
7722 2018-02-06 Marek Polacek <polacek@redhat.com>
7724 PR tree-optimization/84228
7725 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7727 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7730 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7731 emitted arch directives.
7732 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7733 __ARM_FEATURE_COPROC before changing architectures.
7735 2018-02-06 Richard Biener <rguenther@suse.de>
7737 * config/i386/i386.c (print_reg): Fix typo.
7738 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7740 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7742 * configure: Regenerate.
7744 2018-02-05 Martin Sebor <msebor@redhat.com>
7746 PR tree-optimization/83369
7747 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7750 2018-02-05 Martin Liska <mliska@suse.cz>
7752 * doc/invoke.texi: Cherry-pick upstream r323995.
7754 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7756 * ira.c (ira_init_register_move_cost): Adjust comment.
7758 2018-02-05 Martin Liska <mliska@suse.cz>
7760 PR gcov-profile/84137
7761 * doc/gcov.texi: Fix typo in documentation.
7763 2018-02-05 Martin Liska <mliska@suse.cz>
7765 PR gcov-profile/83879
7766 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7767 using dlopen functionality.
7769 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7771 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7772 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7773 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7774 _mm_maskz_range_round_ss): New intrinsics.
7775 (__builtin_ia32_rangesd128_round)
7776 (__builtin_ia32_rangess128_round): Remove.
7777 (__builtin_ia32_rangesd128_mask_round,
7778 __builtin_ia32_rangess128_mask_round): New builtins.
7779 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7780 __builtin_ia32_rangess128_round): Remove.
7781 (__builtin_ia32_rangesd128_mask_round,
7782 __builtin_ia32_rangess128_mask_round): New builtins.
7783 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7784 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7785 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7786 "<round_saeonly_constraint>")): Changed to ...
7787 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7788 "<round_saeonly_scalar_constraint>")): ... this.
7789 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7790 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7791 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7792 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7793 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7795 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7797 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7799 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7800 Remove all values except native, 8540 and 8548.
7802 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7804 * config/i386/i386.c (ix86_output_function_return): Pass
7805 INVALID_REGNUM, instead of -1, as invalid register number to
7806 indirect_thunk_name and output_indirect_thunk.
7808 2018-02-02 Julia Koval <julia.koval@intel.com>
7810 * config.gcc: Add -march=icelake.
7811 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7812 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7813 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7814 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7815 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7816 (processor_target_table): Add icelake.
7817 (ix86_option_override_internal): Handle new PTAs.
7818 (get_builtin_code_for_version): Handle icelake.
7819 (M_INTEL_COREI7_ICELAKE): New.
7820 (fold_builtin_cpu): Handle icelake.
7821 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7822 * doc/invoke.texi: Add -march=icelake.
7824 2018-02-02 Julia Koval <julia.koval@intel.com>
7826 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7827 to wide_int_bitmask.
7828 * wide-int-bitmask.h: New.
7830 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7833 * config/i386/i386.md: Replace Pmode with word_mode in
7834 builtin_setjmp_setup and builtin_longjmp to support x32.
7836 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7840 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7842 (rs6000_supported_cpu_names): New static variable.
7843 (linux_cpu_translation_table): Likewise.
7844 (elf_platform) <cpu>: Define new static variable and use it.
7845 Translate kernel AT_PLATFORM name to canonical name if needed.
7846 Error if platform name is unknown.
7848 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7851 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7853 2018-02-01 Jeff Law <law@redhat.com>
7856 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7857 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7858 the scratch if RELEASE_VIA_POP is false.
7859 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7860 If we have to save a temporary register, decrement SIZE appropriately.
7861 Pass new arguments to release_scratch_register_on_entry.
7862 (ix86_adjust_stack_and_probe): Likewise.
7863 (ix86_emit_probe_stack_range): Pass new arguments to
7864 release_scratch_register_on_entry.
7866 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7868 PR rtl-optimization/84157
7869 * combine.c (change_zero_ext): Use REG_P predicate in
7870 front of HARD_REGISTER_P predicate.
7872 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7874 * config/avr/avr.c (avr_option_override): Move disabling of
7875 -fdelete-null-pointer-checks to...
7876 * common/config/avr/avr-common.c (avr_option_optimization_table):
7879 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7881 PR tree-optimization/81635
7882 * tree-data-ref.c (split_constant_offset_1): For types that
7883 wrap on overflow, try to use range info to prove that wrapping
7886 2018-02-01 Renlin Li <renlin.li@arm.com>
7889 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7891 (aarch64_register_move_cost): Likewise.
7892 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7894 (REG_CLASS_NAMES): Likewise.
7895 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7896 TAILCALL_ADDR_REGS. Remove IP registers.
7897 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7899 2018-02-01 Richard Biener <rguenther@suse.de>
7901 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7902 for specifying RPO order and allow NULL for that.
7903 * domwalk.c (dom_walker::dom_walker): Likewise.
7904 (dom_walker::walk): Handle NULL RPO order.
7905 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7907 (rewrite_update_dom_walker): Likewise.
7908 (mark_def_dom_walker): Likewise.
7910 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7912 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7913 (aarch64_maybe_expand_sve_subreg_move): Declare.
7914 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7915 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7917 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7918 that are semantically a reverse operation.
7919 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7920 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7921 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7923 (aarch64_can_change_mode_class): For big-endian, forbid changes
7924 between two SVE modes if they have different element sizes.
7926 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7928 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7929 the TImode handling for big-endian targets.
7931 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7933 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7934 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7936 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7937 Remove BSWAP handing for big-endian targets and use the form of
7938 LD1RQ appropariate for the mode.
7940 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7942 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7943 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7946 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7949 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7950 check for operands that need to go through aarch64_sve_reload_be.
7952 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7954 PR tree-optimization/81661
7955 PR tree-optimization/84117
7956 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7957 * tree-eh.c: Include gimplify.h.
7958 (find_trapping_overflow, replace_trapping_overflow,
7959 rewrite_to_non_trapping_overflow): New functions.
7960 * tree-vect-loop.c: Include tree-eh.h.
7961 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7962 * tree-data-ref.c: Include tree-eh.h.
7963 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7965 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7967 PR rtl-optimization/84123
7968 * combine.c (change_zero_ext): Check if hard register satisfies
7969 can_change_dest_mode before calling gen_lowpart_SUBREG.
7971 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7974 * ira.c (ira_init_register_move_cost): Remove assert.
7976 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7978 PR rtl-optimization/84071
7979 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7980 * doc/tm.texi: Regenerate.
7982 2018-01-31 Richard Biener <rguenther@suse.de>
7984 PR tree-optimization/84132
7985 * tree-data-ref.c (analyze_miv_subscript): Properly
7986 check whether evolution_function_is_affine_multivariate_p
7987 before calling gcd_of_steps_may_divide_p.
7989 2018-01-31 Julia Koval <julia.koval@intel.com>
7992 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7993 * config/i386/i386.md (rdpid_rex64) New.
7994 (rdpid): Make 32bit only.
7996 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7999 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8000 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8002 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8005 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8007 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8009 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8011 PR rtl-optimization/84071
8012 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8013 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8015 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8017 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8018 (arc_attribute_table): Add 'aux' attribute.
8019 (arc_in_small_data_p): Consider aux like variables.
8020 (arc_is_aux_reg_p): New function.
8021 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8022 (arc_get_aux_arg): New function.
8023 (prepare_move_operands): Handle aux-register access.
8024 (arc_handle_aux_attribute): New function.
8025 * doc/extend.texi (ARC Variable attributes): Add subsection.
8027 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8029 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8030 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8031 (arc_attribute_table): Add 'uncached' attribute.
8032 (arc_print_operand): Print '.di' flag for uncached memory
8034 (arc_in_small_data_p): Do not consider for small data the uncached
8036 (arc_is_uncached_mem_p): New function.
8037 * config/arc/predicates.md (compact_store_memory_operand): Check
8038 for uncached memory accesses.
8039 (nonvol_nonimm_operand): Likewise.
8040 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8042 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8045 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8046 falign-loops=): Add Optimization flag.
8048 2018-01-30 Jeff Law <law@redhat.com>
8051 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8052 INT_REGISTERS_SAVED. Check it prior to calling
8053 get_scratch_register_on_entry.
8054 (ix86_adjust_stack_and_probe): Similarly.
8055 (ix86_emit_probe_stack_range): Similarly.
8056 (ix86_expand_prologue): Corresponding changes.
8058 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8061 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8062 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8064 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8067 * lra-constraints.c (curr_insn_transform): Process AND in the
8070 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8072 PR rtl-optimization/83986
8073 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8074 dependence against last_pending_memory_flush in addition to
8077 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8079 PR tree-optimization/81611
8080 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8083 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8086 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8089 2018-01-30 Richard Biener <rguenther@suse.de>
8090 Jakub Jelinek <jakub@redhat.com>
8092 PR tree-optimization/84111
8093 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8094 inner loops added during recursion, as they don't have up-to-date
8097 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8100 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8101 (can_inline_edge_by_limits_p): ... here.
8102 (can_early_inline_edge_p, check_callers,
8103 update_caller_keys, update_callee_keys, recursive_inlining,
8104 add_new_edges_to_heap, speculation_useful_p,
8105 inline_small_functions,
8106 inline_small_functions, flatten_function,
8107 inline_to_all_callers_1): Update.
8109 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8111 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8114 2018-01-30 Richard Biener <rguenther@suse.de>
8116 PR tree-optimization/83008
8117 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8118 invariant and constant vector uses in stmts when they need
8121 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8124 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8125 * configure: Regenerate.
8127 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8129 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8131 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8132 Use gen_rtx_REG rather than gen_lowpart.
8134 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8136 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8137 rather than 0 when creating partial subregs.
8139 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8141 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8144 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8147 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8148 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8149 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8150 flags. This restores the settings used before the 2017-07-24.
8151 Turning off pre increment/decrement/modify allows IVOPTS to
8152 optimize DF/SF loops where the index is an int.
8154 2018-01-29 Richard Biener <rguenther@suse.de>
8155 Kelvin Nilsen <kelvin@gcc.gnu.org>
8158 * tree-vect-stmts.c (vectorizable_call): Don't call
8159 targetm.vectorize_builtin_md_vectorized_function if callee is
8162 2018-01-22 Carl Love <cel@us.ibm.com>
8164 * doc/extend.tex: Fix typo in second arg in
8165 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8167 2018-01-29 Richard Biener <rguenther@suse.de>
8169 PR tree-optimization/84086
8170 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8171 (flush_ssaname_freelist): When SSA names were released reset
8172 the SCEV hash table.
8174 2018-01-29 Richard Biener <rguenther@suse.de>
8176 PR tree-optimization/84057
8177 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8178 removed paths when removing edges.
8180 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8182 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8183 -mfunction-return=@var{choice}.
8185 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8188 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8190 (layout::print_source_line): Likewise.
8191 (test_get_line_width_without_trailing_whitespace): Add test cases.
8193 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8196 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8199 2018-01-26 Jim Wilson <jimw@sifive.com>
8201 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8203 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8206 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8208 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8209 and CMP + SUB-immediate -> SUBS.
8211 2018-01-26 Martin Sebor <msebor@redhat.com>
8213 PR tree-optimization/83896
8214 * tree-ssa-strlen.c (get_string_len): Rename...
8215 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8216 Avoid assuming length is constant.
8217 (handle_char_store): Use HOST_WIDE_INT for string length.
8219 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8222 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8223 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8225 2018-01-26 Richard Biener <rguenther@suse.de>
8227 PR rtl-optimization/84003
8228 * dse.c (record_store): Only record redundant stores when
8229 the earlier store aliases at least all accesses the later one does.
8231 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8233 PR rtl-optimization/83985
8234 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8235 REG_CFA_RESTORE insns.
8236 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8237 REG_CFA_RESTORE insns here.
8240 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8241 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8243 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8245 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8246 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8247 (arc_init): Likewise.
8248 (arc_override_options): Likewise.
8249 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8251 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8253 * config/arc/arc.h (TARGET_DBNZ): Define.
8254 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8255 properly set the tune attribute.
8256 (dbnz): Use TARGET_DBNZ guard.
8257 * config/arc/arc.opt (mtune): Add core3 option.
8259 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8261 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8262 recognize new pic like addresses.
8263 (arc_delegitimize_address): Clean up.
8265 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8267 * config/arc/arc-arches.def: Option mrf16 valid for all
8269 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8270 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8271 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8272 * config/arc/arc-tables.opt: Regenerate.
8273 * config/arc/arc.c (arc_conditional_register_usage): Handle
8274 reduced register file case.
8275 (arc_file_start): Set must have build attributes.
8276 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8278 * config/arc/arc.opt (mrf16): Add new option.
8279 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8280 * config/arc/genmultilib.awk: Handle new mrf16 option.
8281 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8282 * config/arc/t-multilib: Regenerate.
8283 * doc/invoke.texi (ARC Options): Document mrf16 option.
8285 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8287 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8288 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8289 (arc_attribute_table): Add 'secure_call' attribute.
8290 (arc_print_operand): Print secure call operand.
8291 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8293 (arc_is_secure_call_p): New function. * config/arc/arc.md
8294 (call_i): Add support for sjli instruction.
8295 (call_value_i): Likewise.
8296 * config/arc/constraints.md (Csc): New constraint.
8298 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8299 John Eric Martin <John.Martin@emmicro-us.com>
8301 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8302 * config/arc/arc.c (_arc_jli_section): New struct.
8303 (arc_jli_section): New type.
8304 (rc_jli_sections): New static variable.
8305 (arc_handle_jli_attribute): New function.
8306 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8307 (arc_file_end): New function.
8308 (TARGET_ASM_FILE_END): Define.
8309 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8310 (arc_add_jli_section): New function.
8311 (jli_call_scan): Likewise.
8312 (arc_reorg): Call jli_call_scan.
8313 (arc_output_addsi): Remove 'S' from printing asm operand.
8314 (arc_is_jli_call_p): New function.
8315 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8317 (movhi_insn): Likewise.
8318 (movsi_insn): Likewise.
8319 (movsi_set_cc_insn): Likewise.
8320 (loadqi_update): Likewise.
8321 (load_zeroextendqisi_update): Likewise.
8322 (load_signextendqisi_update): Likewise.
8323 (loadhi_update): Likewise.
8324 (load_zeroextendhisi_update): Likewise.
8325 (load_signextendhisi_update): Likewise.
8326 (loadsi_update): Likewise.
8327 (loadsf_update): Likewise.
8328 (movsicc_insn): Likewise.
8329 (bset_insn): Likewise.
8330 (bxor_insn): Likewise.
8331 (bclr_insn): Likewise.
8332 (bmsk_insn): Likewise.
8333 (bicsi3_insn): Likewise.
8334 (cmpsi_cc_c_insn): Likewise.
8335 (movsi_ne): Likewise.
8336 (movsi_cond_exec): Likewise.
8337 (clrsbsi2): Likewise.
8348 (call_i): Remove 'S' asm letter, add jli instruction.
8349 (call_value_i): Likewise.
8350 * config/arc/arc.op (mjli-always): New option.
8351 * config/arc/constraints.md (Cji): New constraint.
8352 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8354 (subsf3_fpx): Likewise.
8355 (mulsf3_fpx): Likewise.
8356 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8358 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8360 * doc/invoke.texi (ARC): Document mjli-always option.
8362 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8364 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8365 avoid addition with 0 and use incw and decw where possible.
8367 2018-01-26 Richard Biener <rguenther@suse.de>
8369 PR tree-optimization/81082
8370 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8371 association if it requires casting to unsigned.
8372 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8373 from fold_plusminus_mult_expr to catch important cases late when
8374 range info is available.
8376 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8378 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8379 * configure.ac (hidden_linkonce): New test.
8380 * configure: Regenerate.
8381 * config.in: Regenerate.
8383 2018-01-26 Julia Koval <julia.koval@intel.com>
8385 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8386 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8387 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8388 _mm_mask_bitshuffle_epi64_mask): Fix type.
8389 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8390 USI_FTYPE_V4DI_V4DI_USI): Remove.
8391 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8392 __builtin_ia32_vpshufbitqmb256_mask,
8393 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8394 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8395 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8397 2018-01-26 Alan Modra <amodra@gmail.com>
8400 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8401 UNSPEC_VBPERMQ. Sort other unspecs.
8403 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8405 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8407 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8410 * predict.c (drop_profile): Do not push/pop cfun; update also
8412 (handle_missing_profiles): Fix logic looking for zero profiles.
8414 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8417 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8418 on functions with #pragma omp declare simd or functions with simd
8420 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8421 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8422 Remove trailing \n from warning_at calls.
8424 2018-01-25 Tom de Vries <tom@codesourcery.com>
8427 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8428 for neutered workers.
8430 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8433 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8434 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8436 2018-01-24 Jeff Law <law@redhat.com>
8439 * i386.c (get_probe_interval): Move to earlier point.
8440 (ix86_compute_frame_layout): If -fstack-clash-protection and
8441 the frame is larger than the probe interval, then use pushes
8442 to save registers rather than reg->mem moves.
8443 (ix86_expand_prologue): Remove conditional for int_registers_saved
8446 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8449 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8450 min/max for never referenced object.
8452 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8455 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8457 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8458 attributes from DECL_ATTRIBUTES (decl) without affecting
8459 DECL_ATTRIBUTES (current_function_decl).
8460 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8461 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8463 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8465 PR tree-optimization/83979
8466 * fold-const.c (fold_comparison): Use constant_boolean_node
8467 instead of boolean_{true,false}_node.
8469 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8471 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8474 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8476 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8477 Simplify the clause that sets the length attribute.
8478 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8479 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8480 clause that sets the length attribute.
8481 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8483 2018-01-24 Tom de Vries <tom@codesourcery.com>
8486 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8487 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8488 Add strict parameter.
8489 (prevent_branch_around_nothing): Insert dummy insn between branch to
8490 label and label with no ptx insn inbetween.
8491 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8493 2018-01-24 Tom de Vries <tom@codesourcery.com>
8496 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8497 for neutered threads in warp.
8498 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8500 2018-01-24 Richard Biener <rguenther@suse.de>
8502 PR tree-optimization/83176
8503 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8506 2018-01-24 Richard Biener <rguenther@suse.de>
8508 PR tree-optimization/82819
8509 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8510 code generating pluses that are no-ops in the target precision.
8512 2018-01-24 Richard Biener <rguenther@suse.de>
8515 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8517 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8519 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8520 to merge probabilities.
8521 * predict.c (probably_never_executed): Also mark as cold functions
8522 with global 0 profile and guessed local profile.
8523 * profile-count.c (profile_probability::combine_with_count): New
8525 * profile-count.h (profile_probability::operator*,
8526 profile_probability::operator*=, profile_probability::operator/,
8527 profile_probability::operator/=): Reduce precision to adjusted
8528 and set value to guessed on contradictory divisions.
8529 (profile_probability::combine_with_freq): Remove.
8530 (profile_probability::combine_wiht_count): Declare.
8531 (profile_count::force_nonzero):: Set to adjusted.
8532 (profile_count::probability_in):: Set quality to adjusted.
8533 * tree-ssa-tail-merge.c (replace_block_by): Use
8536 2018-01-23 Andrew Waterman <andrew@sifive.com>
8537 Jim Wilson <jimw@sifive.com>
8539 * config/riscv/riscv.c (riscv_stack_boundary): New.
8540 (riscv_option_override): Set riscv_stack_boundary. Handle
8541 riscv_preferred_stack_boundary_arg.
8542 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8543 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8544 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8545 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8546 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8547 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8549 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8552 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8553 of struct ix86_frame.
8554 (ix86_expand_epilogue): Likewise. Add a local variable for
8555 the reg_save_offset field in struct ix86_frame.
8557 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8559 PR tree-optimization/82604
8560 * tree-loop-distribution.c (enum partition_kind): New enum item
8561 PKIND_PARTIAL_MEMSET.
8562 (partition_builtin_p): Support above new enum item.
8563 (generate_code_for_partition): Ditto.
8564 (compute_access_range): Differentiate cases that equality can be
8565 proven at all loops, the innermost loops or no loops.
8566 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8567 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8568 (finalize_partitions, distribute_loop): Don't fuse partition of
8569 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8570 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8573 2018-01-23 Martin Liska <mliska@suse.cz>
8575 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8576 order to ignore the predictor.
8577 (PRED_POLYMORPHIC_CALL): Likewise.
8578 (PRED_RECURSIVE_CALL): Likewise.
8580 2018-01-23 Martin Liska <mliska@suse.cz>
8582 * tree-profile.c (tree_profiling): Print function header to
8583 aware reader which function we are working on.
8584 * value-prof.c (gimple_find_values_to_profile): Do not print
8585 not interesting value histograms.
8587 2018-01-23 Martin Liska <mliska@suse.cz>
8589 * profile-count.h (enum profile_quality): Add
8590 profile_uninitialized as the first value. Do not number values
8591 as they are zero based.
8592 (profile_count::verify): Update sanity check.
8593 (profile_probability::verify): Likewise.
8595 2018-01-23 Nathan Sidwell <nathan@acm.org>
8597 * doc/invoke.texi (ffor-scope): Deprecate.
8599 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8601 PR tree-optimization/83510
8602 * domwalk.c (set_all_edges_as_executable): New function.
8603 (dom_walker::dom_walker): Convert bool param
8604 "skip_unreachable_blocks" to enum reachability. Move setup of
8605 edge flags to set_all_edges_as_executable and only do it when
8606 reachability is REACHABLE_BLOCKS.
8607 * domwalk.h (enum dom_walker::reachability): New enum.
8608 (dom_walker::dom_walker): Convert bool param
8609 "skip_unreachable_blocks" to enum reachability.
8610 (set_all_edges_as_executable): New decl.
8611 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8612 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8614 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8615 but converting true to REACHABLE_BLOCKS.
8616 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8618 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8619 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8620 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8622 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8623 if check_all_array_refs will be called.
8625 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8627 * tree.c (selftest::test_location_wrappers): Add more test
8630 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8632 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8633 (selftest::test_bit_in_range): Likewise.
8635 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8638 * doc/sourcebuild.texi (vect_float): Say that the selector
8639 only describes the situation when -funsafe-math-optimizations is on.
8640 (vect_float_strict): Document.
8642 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8644 PR tree-optimization/83965
8645 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8646 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8647 instead of checking only for a reduction.
8648 (vect_recog_widen_sum_pattern): Likewise.
8650 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8652 * predict.c (probably_never_executed): Only use precise profile info.
8653 (compute_function_frequency): Skip after inlining hack since we now
8654 have quality checking.
8656 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8658 * profile-count.h (profile_probability::very_unlikely,
8659 profile_probability::unlikely, profile_probability::even): Set
8660 precision to guessed.
8662 2018-01-23 Richard Biener <rguenther@suse.de>
8664 PR tree-optimization/83963
8665 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8666 Properly terminate dominator walk when crossing the exit edge not
8667 when visiting its source block.
8669 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8672 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8673 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8675 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8677 PR tree-optimization/83957
8678 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8679 semicolon after for body surrounded by braces.
8681 PR tree-optimization/83081
8682 * profile-count.h (profile_probability::split): New method.
8683 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8684 Use profile_probability::split.
8685 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8686 when splitting a single conditional jump into 2.
8688 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8690 PR tree-optimization/69452
8691 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8694 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8696 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8697 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8698 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8700 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8702 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8703 * config/rl78/rl78.md (movdi): New define_expand.
8704 * config/rl78/rl78.c (rl78_split_movdi): New function.
8706 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8709 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8711 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8712 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8713 128-bit to produce an UNSPEC move to get the double word with the
8714 signbit and then a shift directly to do signbit.
8715 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8716 implementation with a new version that just does either a direct
8717 move or a regular move. Move memory interface to separate insns.
8718 Move insns so they are next to the expander.
8719 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8720 with signbit move. Split big and little endian case.
8721 (signbit<mode>2_dm_mem_le): Likewise.
8722 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8723 (signbit<mode>2_dm2): Likewise.
8725 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8727 * config/rl78/rl78.md (anddi3): New define_expand.
8729 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8731 * config/rl78/rl78.md (umindi3): New define_expand.
8733 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8735 * config/rl78/rl78.md (smindi3): New define_expand.
8737 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8739 * config/rl78/rl78.md (smaxdi3): New define_expand.
8741 2018-01-22 Carl Love <cel@us.ibm.com>
8743 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8744 LVX_V1TI): Add macro expansion.
8745 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8746 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8747 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8748 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8749 Change check to determine if the instruction is a byte reversing
8750 entry. Fix typo in comment.
8751 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8752 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8753 Add def_builtin calls for new builtins.
8754 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8755 Add define_insn expansion.
8757 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8759 * config/rl78/rl78.md (umaxdi3): New define_expand.
8761 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8763 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8764 for non-QImode registers.
8766 2018-01-22 Richard Biener <rguenther@suse.de>
8768 PR tree-optimization/83963
8769 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8770 including the loop exit block.
8771 (scop_detection::merge_sese): Likewise.
8772 (scop_detection::add_scop): Do it here instead.
8774 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8776 * doc/sourcebuild.texi (arm_softfloat): Document.
8778 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8781 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8782 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8783 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8785 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8786 David Edelsohn <dje.gcc@gmail.com>
8789 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8790 Change "crset eq" to "crset 2".
8791 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8792 (*call_indirect_aix<mode>_nospec): Likewise.
8793 (*call_value_indirect_aix<mode>_nospec): Likewise.
8794 (*call_indirect_elfv2<mode>_nospec): Likewise.
8795 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8796 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8797 change assembly output from . to $.
8798 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8799 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8800 (*tablejump<mode>_internal1_nospec): Likewise.
8802 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8805 * config/sh/sh_optimize_sett_clrt.cc:
8806 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8808 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8810 PR tree-optimization/83940
8811 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8812 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8813 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8814 use it to pass back the definition type.
8815 (vect_check_store_rhs): Likewise rhs_dt_out.
8816 (vect_build_gather_load_calls): Add a mask_dt argument and use
8817 it instead of a call to vect_is_simple_use.
8818 (vectorizable_store): Update calls to vect_check_load_store_mask
8819 and vect_check_store_rhs. Use the dt returned by the latter instead
8820 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8821 instead of calls to vect_is_simple_use. Pass the scalar rather
8822 than the vector operand to vect_is_simple_use when handling
8823 second and subsequent copies of an rhs value.
8824 (vectorizable_load): Update calls to vect_check_load_store_mask
8825 and vect_build_gather_load_calls. Use the cached mask_dt and
8826 gs_info.offset_dt instead of calls to vect_is_simple_use.
8828 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8831 * tree-emutls.c: Include gimplify.h.
8832 (lower_emutls_2): New function.
8833 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8834 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8835 it before further processing.
8838 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8839 UINTVAL (trueop1) instead of INTVAL (op1).
8841 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8845 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8846 INCOMING_FRAME_SP_OFFSET if not defined.
8847 (scan_trace): Add ENTRY argument. If true and
8848 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8849 emit a note to adjust the CFA offset.
8850 (create_cfi_notes): Adjust scan_trace callers.
8851 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8852 INCOMING_FRAME_SP_OFFSET in the CIE.
8853 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8854 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8856 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8857 * doc/tm.texi: Regenerated.
8859 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8861 PR rtl-optimization/83147
8862 * lra-constraints.c (remove_inheritance_pseudos): Use
8863 lra_substitute_pseudo_within_insn.
8865 2018-01-19 Tom de Vries <tom@codesourcery.com>
8866 Cesar Philippidis <cesar@codesourcery.com>
8869 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8871 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8874 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8875 spaces for function labels.
8877 2018-01-19 Martin Liska <mliska@suse.cz>
8879 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8880 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8881 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8882 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8883 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8884 (PRED_CONST_RETURN): Change from 69 to 65.
8885 (PRED_NULL_RETURN): Change from 91 to 71.
8886 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8887 (PRED_LOOP_GUARD): Change from 66 to 73.
8889 2018-01-19 Martin Liska <mliska@suse.cz>
8891 * predict.c (predict_insn_def): Add new assert.
8892 (struct branch_predictor): Change type to signed integer.
8893 (test_prediction_value_range): Amend test to cover
8895 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8896 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8897 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8898 (PRED_LOOP_IV_COMPARE): Likewise.
8899 * predict.h (PROB_UNINITIALIZED): Define new constant.
8901 2018-01-19 Martin Liska <mliska@suse.cz>
8903 * predict.c (dump_prediction): Add new format for
8904 analyze_brprob.py script which is enabled with -details
8906 * profile-count.h (precise_p): New function.
8908 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8910 PR tree-optimization/83922
8911 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8912 there are no statements that need masking.
8913 (vect_active_double_reduction_p): New function.
8914 (vect_analyze_loop_operations): Use it when handling phis that
8915 are not in the loop header.
8917 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8919 PR tree-optimization/83914
8920 * tree-vect-loop.c (vectorizable_induction): Don't convert
8921 init_expr or apply the peeling adjustment for inductions
8922 that are nested within the vectorized loop.
8924 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8926 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8929 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8933 * function.h (gimplify_parameters): Add gimple_seq * argument.
8934 * function.c: Include gimple.h and options.h.
8935 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8936 for the added local temporaries if needed.
8937 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8938 if there are any parameter cleanups, wrap whole body into a
8939 try/finally with the cleanups.
8941 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8944 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8945 Use GET_MODE_CLASS for scalar floating point.
8947 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8951 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8952 Fix call of call_cgraph_insertion_hooks.
8954 2018-01-18 Martin Sebor <msebor@redhat.com>
8956 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8958 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8961 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8964 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8967 * common.opt: (-ffile-prefix-map): New option.
8968 * opts.c (common_handle_option): Defer it.
8969 * opts-global.c (handle_common_deferred_options): Handle it.
8970 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8971 * file-prefix-map.h: New file.
8972 (remap_debug_filename, add_debug_prefix_map): ...here.
8973 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8974 * final.c (debug_prefix_map, add_debug_prefix_map
8975 remap_debug_filename): Move to...
8976 * file-prefix-map.c: New file.
8977 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8978 generalize, get rid of alloca(), use strrchr() instead of strchr().
8979 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8980 Implement in terms of add_prefix_map().
8981 (remap_macro_filename, remap_debug_filename): Implement in term of
8983 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8984 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8985 * dbxout.c: Include file-prefix-map.h.
8986 * varasm.c: Likewise.
8987 * vmsdbgout.c: Likewise.
8988 * xcoffout.c: Likewise.
8989 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8990 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8991 * doc/invoke.texi (-ffile-prefix-map): Document.
8992 (-fdebug-prefix-map): Update description.
8994 2018-01-18 Martin Liska <mliska@suse.cz>
8996 * config/i386/i386.c (indirect_thunk_name): Document that also
8998 (output_indirect_thunk): Document why both instructions
8999 (pause and lfence) are generated.
9001 2018-01-18 Richard Biener <rguenther@suse.de>
9003 PR tree-optimization/83887
9004 * graphite-scop-detection.c
9005 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9006 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9007 (scop_detection::merge_sese): Re-implement with a flood-fill
9008 algorithm that properly finds a SESE region if it exists.
9010 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9013 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9014 pointer_diff optimizations use view_convert instead of convert.
9016 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9018 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9019 Generate different code for -mno-speculate-indirect-jumps.
9020 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9021 (*call_indirect_aix<mode>): Disable for
9022 -mno-speculate-indirect-jumps.
9023 (*call_indirect_aix<mode>_nospec): New define_insn.
9024 (*call_value_indirect_aix<mode>): Disable for
9025 -mno-speculate-indirect-jumps.
9026 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9027 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9028 -mno-speculate-indirect-jumps.
9029 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9031 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9033 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9034 long double type, set the flags for noting the default long double
9035 type, even if we don't pass or return a long double type.
9037 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9040 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9043 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9045 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9046 support for merge[hl].
9047 (fold_mergehl_helper): New helper function.
9048 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9049 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9050 (altivec_vmrglw_direct): Add xxmrglw insn.
9052 2018-01-17 Andrew Waterman <andrew@sifive.com>
9054 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9055 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9057 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9060 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9061 call the lto_location_cache before reading the
9062 DECL_SOURCE_LOCATION of the types.
9064 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9065 Richard Sandiford <richard.sandiford@linaro.org>
9067 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9068 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9069 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9070 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9071 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9073 * config/aarch64/constraints.md (aarch64_movti_operand):
9075 * config/aarch64/predicates.md (Uti): Add new constraint.
9077 2018-01-17 Carl Love <cel@us.ibm.com>
9079 * config/rs6000/vsx.md (define_expand xl_len_r,
9080 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9081 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9083 (define_expand, define_insn): Move the shift left from the
9084 define_insn to the define_expand for lxvl and stxvl instructions.
9085 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9086 and XL_LEN_R definitions to PURE.
9088 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9090 * config/i386/i386.c (indirect_thunk_name): Declare regno
9091 as unsigned int. Compare regno with INVALID_REGNUM.
9092 (output_indirect_thunk): Ditto.
9093 (output_indirect_thunk_function): Ditto.
9094 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9095 in the call to output_indirect_thunk_function.
9097 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9100 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9101 rather than the size of inner_type to determine the stack slot size
9102 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9104 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9107 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9110 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9112 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9113 endian Linux systems to optionally enable multilibs for selecting
9114 the long double type if the user configured an explicit type.
9115 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9116 have no long double multilibs if not defined.
9117 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9118 warn if the user used -mabi={ieee,ibm}longdouble and we built
9119 multilibs for long double.
9120 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9121 appropriate multilib option.
9122 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9124 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9125 for building long double multilibs.
9126 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9128 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9130 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9133 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9135 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9138 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9141 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9144 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9146 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9147 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9149 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9151 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9152 different rtl trees depending on TARGET_64BIT.
9153 (rs6000_gen_lvx): Likewise.
9155 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9157 * config/visium/visium.md (nop): Tweak comment.
9158 (hazard_nop): Likewise.
9160 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9162 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9163 -mspeculate-indirect-jumps.
9164 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9165 for -mno-speculate-indirect-jumps.
9166 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9167 (*call_value_indirect_elfv2<mode>): Disable for
9168 -mno-speculate-indirect-jumps.
9169 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9170 (indirect_jump): Emit different RTL for
9171 -mno-speculate-indirect-jumps.
9172 (*indirect_jump<mode>): Disable for
9173 -mno-speculate-indirect-jumps.
9174 (*indirect_jump<mode>_nospec): New define_insn.
9175 (tablejump): Emit different RTL for
9176 -mno-speculate-indirect-jumps.
9177 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9178 (tablejumpsi_nospec): New define_expand.
9179 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9180 (tablejumpdi_nospec): New define_expand.
9181 (*tablejump<mode>_internal1): Disable for
9182 -mno-speculate-indirect-jumps.
9183 (*tablejump<mode>_internal1_nospec): New define_insn.
9184 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9187 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9189 * caller-save.c (insert_save): Drop unnecessary parameter. All
9192 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9193 Richard Biener <rguenth@suse.de>
9196 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9197 return early, inline manually is_gimple_sizepos. Make sure if we
9198 call gimplify_expr we don't end up with a gimple constant.
9199 * tree.c (variably_modified_type_p): Don't return true for
9200 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9201 * gimplify.h (is_gimple_sizepos): Remove.
9203 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9205 PR tree-optimization/83857
9206 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9207 vectorizable_live_operation for pure SLP statements.
9208 (vectorizable_live_operation): Handle PHIs.
9210 2018-01-16 Richard Biener <rguenther@suse.de>
9212 PR tree-optimization/83867
9213 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9214 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9216 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9219 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9220 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9221 If off is not INTEGER_CST, issue a may not be aligned warning
9222 rather than isn't aligned. Use isn%'t rather than isn't.
9223 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9225 <case MULT_EXPR>: Improve the case when bottom and one of the
9226 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9227 operand, in that case check if the other operand is multiple of
9228 bottom divided by the INTEGER_CST operand.
9230 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9233 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9234 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9235 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9236 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9237 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9238 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9239 (pa_function_arg_size): New function.
9241 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9243 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9244 in a separate statement.
9246 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9248 PR tree-optimization/83847
9249 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9250 group gathers and scatters.
9252 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9254 PR rtl-optimization/86620
9255 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9257 PR rtl-optimization/83213
9258 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9259 to last if both are JUMP_INSNs.
9261 PR tree-optimization/83843
9262 * gimple-ssa-store-merging.c
9263 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9264 store_immediate_info for bswap/nop orig_stores.
9266 2018-01-15 Andrew Waterman <andrew@sifive.com>
9268 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9270 <UDIV>: Increase cost if !TARGET_DIV.
9272 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9274 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9275 (define_attr "cr_logical_3op"): New.
9276 (cceq_ior_compare): Adjust.
9277 (cceq_ior_compare_complement): Adjust.
9278 (*cceq_rev_compare): Adjust.
9279 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9280 (is_cracked_insn): Adjust.
9281 (insn_must_be_first_in_group): Adjust.
9282 * config/rs6000/40x.md: Adjust.
9283 * config/rs6000/440.md: Adjust.
9284 * config/rs6000/476.md: Adjust.
9285 * config/rs6000/601.md: Adjust.
9286 * config/rs6000/603.md: Adjust.
9287 * config/rs6000/6xx.md: Adjust.
9288 * config/rs6000/7450.md: Adjust.
9289 * config/rs6000/7xx.md: Adjust.
9290 * config/rs6000/8540.md: Adjust.
9291 * config/rs6000/cell.md: Adjust.
9292 * config/rs6000/e300c2c3.md: Adjust.
9293 * config/rs6000/e500mc.md: Adjust.
9294 * config/rs6000/e500mc64.md: Adjust.
9295 * config/rs6000/e5500.md: Adjust.
9296 * config/rs6000/e6500.md: Adjust.
9297 * config/rs6000/mpc.md: Adjust.
9298 * config/rs6000/power4.md: Adjust.
9299 * config/rs6000/power5.md: Adjust.
9300 * config/rs6000/power6.md: Adjust.
9301 * config/rs6000/power7.md: Adjust.
9302 * config/rs6000/power8.md: Adjust.
9303 * config/rs6000/power9.md: Adjust.
9304 * config/rs6000/rs64.md: Adjust.
9305 * config/rs6000/titan.md: Adjust.
9307 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9309 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9310 ix86_indirect_branch_register logic.
9312 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9314 * config/i386/constraints.md (Bs): Update
9315 ix86_indirect_branch_register check. Don't check
9316 ix86_indirect_branch_register with GOT_memory_operand.
9318 * config/i386/predicates.md (GOT_memory_operand): Don't check
9319 ix86_indirect_branch_register here.
9320 (GOT32_symbol_operand): Likewise.
9322 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9324 * config/i386/predicates.md (constant_call_address_operand):
9325 Rewrite ix86_indirect_branch_register logic.
9326 (sibcall_insn_operand): Likewise.
9328 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9330 * config/i386/constraints.md (Bs): Replace
9331 ix86_indirect_branch_thunk_register with
9332 ix86_indirect_branch_register.
9334 * config/i386/i386.md (indirect_jump): Likewise.
9335 (tablejump): Likewise.
9336 (*sibcall_memory): Likewise.
9337 (*sibcall_value_memory): Likewise.
9338 Peepholes of indirect call and jump via memory: Likewise.
9339 * config/i386/i386.opt: Likewise.
9340 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9341 (GOT_memory_operand): Likewise.
9342 (call_insn_operand): Likewise.
9343 (sibcall_insn_operand): Likewise.
9344 (GOT32_symbol_operand): Likewise.
9346 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9349 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9350 type rather than type addr's type points to.
9351 (expand_omp_atomic_mutex): Likewise.
9352 (expand_omp_atomic): Likewise.
9354 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9357 * config/i386/i386.c (output_indirect_thunk_function): Use
9358 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9359 for __x86_return_thunk.
9361 2018-01-15 Richard Biener <rguenther@suse.de>
9364 * expmed.c (extract_bit_field_1): Fix typo.
9366 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9369 * config/arm/iterators.md (VF): New mode iterator.
9370 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9371 Remove integer-related logic from pattern.
9372 (neon_vabd<mode>_3): Likewise.
9374 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9377 * common.opt (fstrict-overflow): No longer an alias.
9378 (fwrapv-pointer): New option.
9379 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9380 also for pointer types based on flag_wrapv_pointer.
9381 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9382 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9383 opts->x_flag_wrapv got set.
9384 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9385 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9386 POINTER_TYPE_OVERFLOW_UNDEFINED.
9387 * match.pd: Likewise in address comparison pattern.
9388 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9390 2018-01-15 Richard Biener <rguenther@suse.de>
9393 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9394 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9395 Reset type names to their identifier if their TYPE_DECL doesn't
9396 have linkage (and thus is used for ODR and devirt).
9397 (save_debug_info_for_decl): Remove.
9398 (save_debug_info_for_type): Likewise.
9399 (add_tree_to_fld_list): Adjust.
9400 * tree-pretty-print.c (dump_generic_node): Make dumping of
9401 type names more robust.
9403 2018-01-15 Richard Biener <rguenther@suse.de>
9405 * BASE-VER: Bump to 8.0.1.
9407 2018-01-14 Martin Sebor <msebor@redhat.com>
9410 * builtins.c (check_access): Avoid warning when the no-warning bit
9413 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9415 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9416 * ira-color (allocno_hard_regs_compare): Likewise.
9418 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9421 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9422 Use .pushsection/.popsection.
9424 2018-01-14 Martin Sebor <msebor@redhat.com>
9427 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9429 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9431 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9432 entry from extra_headers.
9433 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9434 extra_headers, make the list bitwise identical to the i?86-*-* one.
9436 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9438 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9439 -mcmodel=large with -mindirect-branch=thunk,
9440 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9441 -mfunction-return=thunk-extern.
9442 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9443 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9444 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9446 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9448 * config/i386/i386.c (print_reg): Print the name of the full
9449 integer register without '%'.
9450 (ix86_print_operand): Handle 'V'.
9451 * doc/extend.texi: Document 'V' modifier.
9453 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9455 * config/i386/constraints.md (Bs): Disallow memory operand for
9456 -mindirect-branch-register.
9458 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9459 (GOT_memory_operand): Likewise.
9460 (call_insn_operand): Likewise.
9461 (sibcall_insn_operand): Likewise.
9462 (GOT32_symbol_operand): Likewise.
9463 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9464 for -mindirect-branch-register.
9465 (tablejump): Likewise.
9466 (*sibcall_memory): Likewise.
9467 (*sibcall_value_memory): Likewise.
9468 Disallow peepholes of indirect call and jump via memory for
9469 -mindirect-branch-register.
9470 (*call_pop): Replace m with Bw.
9471 (*call_value_pop): Likewise.
9472 (*sibcall_pop_memory): Replace m with Bs.
9473 * config/i386/i386.opt (mindirect-branch-register): New option.
9474 * doc/invoke.texi: Document -mindirect-branch-register option.
9476 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9478 * config/i386/i386-protos.h (ix86_output_function_return): New.
9479 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9480 set function_return_type.
9481 (indirect_thunk_name): Add ret_p to indicate thunk for function
9483 (output_indirect_thunk_function): Pass false to
9484 indirect_thunk_name.
9485 (ix86_output_indirect_branch_via_reg): Likewise.
9486 (ix86_output_indirect_branch_via_push): Likewise.
9487 (output_indirect_thunk_function): Create alias for function
9488 return thunk if regno < 0.
9489 (ix86_output_function_return): New function.
9490 (ix86_handle_fndecl_attribute): Handle function_return.
9491 (ix86_attribute_table): Add function_return.
9492 * config/i386/i386.h (machine_function): Add
9493 function_return_type.
9494 * config/i386/i386.md (simple_return_internal): Use
9495 ix86_output_function_return.
9496 (simple_return_internal_long): Likewise.
9497 * config/i386/i386.opt (mfunction-return=): New option.
9498 (indirect_branch): Mention -mfunction-return=.
9499 * doc/extend.texi: Document function_return function attribute.
9500 * doc/invoke.texi: Document -mfunction-return= option.
9502 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9504 * config/i386/i386-opts.h (indirect_branch): New.
9505 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9506 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9507 with local indirect jump when converting indirect call and jump.
9508 (ix86_set_indirect_branch_type): New.
9509 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9510 (indirectlabelno): New.
9511 (indirect_thunk_needed): Likewise.
9512 (indirect_thunk_bnd_needed): Likewise.
9513 (indirect_thunks_used): Likewise.
9514 (indirect_thunks_bnd_used): Likewise.
9515 (INDIRECT_LABEL): Likewise.
9516 (indirect_thunk_name): Likewise.
9517 (output_indirect_thunk): Likewise.
9518 (output_indirect_thunk_function): Likewise.
9519 (ix86_output_indirect_branch_via_reg): Likewise.
9520 (ix86_output_indirect_branch_via_push): Likewise.
9521 (ix86_output_indirect_branch): Likewise.
9522 (ix86_output_indirect_jmp): Likewise.
9523 (ix86_code_end): Call output_indirect_thunk_function if needed.
9524 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9526 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9527 (ix86_attribute_table): Add indirect_branch.
9528 * config/i386/i386.h (machine_function): Add indirect_branch_type
9529 and has_local_indirect_jump.
9530 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9532 (tablejump): Likewise.
9533 (*indirect_jump): Use ix86_output_indirect_jmp.
9534 (*tablejump_1): Likewise.
9535 (simple_return_indirect_internal): Likewise.
9536 * config/i386/i386.opt (mindirect-branch=): New option.
9537 (indirect_branch): New.
9540 (thunk-inline): Likewise.
9541 (thunk-extern): Likewise.
9542 * doc/extend.texi: Document indirect_branch function attribute.
9543 * doc/invoke.texi: Document -mindirect-branch= option.
9545 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9548 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9550 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9552 * ipa-inline.c (want_inline_small_function_p): Return false if
9553 inlining has already failed with CIF_FINAL_ERROR.
9554 (update_caller_keys): Call want_inline_small_function_p before
9556 (update_callee_keys): Likewise.
9558 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9560 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9562 (rs6000_quadword_masked_address_p): Likewise.
9563 (quad_aligned_load_p): Likewise.
9564 (quad_aligned_store_p): Likewise.
9565 (const_load_sequence_p): Add comment to describe the outer-most loop.
9566 (mimic_memory_attributes_and_flags): New function.
9567 (rs6000_gen_stvx): Likewise.
9568 (replace_swapped_aligned_store): Likewise.
9569 (rs6000_gen_lvx): Likewise.
9570 (replace_swapped_aligned_load): Likewise.
9571 (replace_swapped_load_constant): Capitalize argument name in
9572 comment describing this function.
9573 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9574 and stores that access quad-word aligned addresses and replace
9575 with stvx or lvx instructions when appropriate.
9576 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9577 New function prototype.
9578 (rs6000_quadword_masked_address_p): Likewise.
9579 (rs6000_gen_lvx): Likewise.
9580 (rs6000_gen_stvx): Likewise.
9581 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9582 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9583 when memory address is aligned.
9584 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9585 this split to select lvx instruction when memory address is aligned.
9586 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9587 instruction when memory address is aligned.
9588 (*vsx_le_perm_load_v16qi): Likewise.
9589 (four unnamed splitters): Modify to select the stvx instruction
9590 when memory is aligned.
9592 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9594 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9595 which appears in the queue multiple times.
9597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9598 Alan Hayward <alan.hayward@arm.com>
9599 David Sherwood <david.sherwood@arm.com>
9601 * tree-vectorizer.h (vec_lower_bound): New structure.
9602 (_loop_vec_info): Add check_nonzero and lower_bounds.
9603 (LOOP_VINFO_CHECK_NONZERO): New macro.
9604 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9605 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9606 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9607 fields. Make seg_len the distance travelled, not including the
9609 (dr_direction_indicator): Declare.
9610 (dr_zero_step_indicator): Likewise.
9611 (dr_known_forward_stride_p): Likewise.
9612 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9614 (runtime_alias_check_p): Allow runtime alias checks with
9616 (operator ==): Compare access_size and align.
9617 (prune_runtime_alias_test_list): Rework for new distinction between
9618 the access_size and seg_len.
9619 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9621 (get_segment_min_max): New function.
9622 (create_intersect_range_checks): Use it.
9623 (dr_step_indicator): New function.
9624 (dr_direction_indicator): Likewise.
9625 (dr_zero_step_indicator): Likewise.
9626 (dr_known_forward_stride_p): Likewise.
9627 * tree-loop-distribution.c (data_ref_segment_size): Return
9628 DR_STEP * (niters - 1).
9629 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9631 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9632 (vect_preserves_scalar_order_p): New function, split out from...
9633 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9634 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9635 (vect_vfa_access_size): New function.
9636 (vect_vfa_align): Likewise.
9637 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9638 (dump_lower_bound): New function.
9639 (vect_check_lower_bound): Likewise.
9640 (vect_small_gap_p): Likewise.
9641 (vectorizable_with_step_bound_p): Likewise.
9642 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9643 depencies if the vectorization factor is 1. Convert the checks
9644 for nonzero steps into checks on the bounds of DR_STEP. Try using
9645 a bunds check for variable steps if the minimum required step is
9646 relatively small. Update calls to the dr_with_seg_len
9647 constructor and to vect_compile_time_alias.
9648 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9650 (vect_loop_versioning): Call it.
9651 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9653 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9656 Alan Hayward <alan.hayward@arm.com>
9657 David Sherwood <david.sherwood@arm.com>
9659 * doc/sourcebuild.texi (vect_scatter_store): Document.
9660 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9662 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9664 * genopinit.c (main): Add supports_vec_scatter_store and
9665 supports_vec_scatter_store_cached to target_optabs.
9666 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9667 IFN_MASK_SCATTER_STORE.
9668 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9670 * internal-fn.h (internal_store_fn_p): Declare.
9671 (internal_fn_stored_value_index): Likewise.
9672 * internal-fn.c (scatter_store_direct): New macro.
9673 (expand_scatter_store_optab_fn): New function.
9674 (direct_scatter_store_optab_supported_p): New macro.
9675 (internal_store_fn_p): New function.
9676 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9677 IFN_MASK_SCATTER_STORE.
9678 (internal_fn_mask_index): Likewise.
9679 (internal_fn_stored_value_index): New function.
9680 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9682 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9683 * optabs-query.c (supports_vec_scatter_store_p): New function.
9684 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9685 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9686 true for scatter stores.
9687 (vect_gather_scatter_fn_p): Handle scatter stores too.
9688 (vect_check_gather_scatter): Consider using scatter stores if
9689 supports_vec_scatter_store_p.
9690 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9692 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9693 internal_fn_stored_value_index.
9694 (check_load_store_masking): Handle scatter stores too.
9695 (vect_get_store_rhs): Make public.
9696 (vectorizable_call): Use internal_store_fn_p.
9697 (vectorizable_store): Handle scatter store internal functions.
9698 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9699 when deciding whether the end of the group has been reached.
9700 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9701 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9702 (mask_scatter_store<mode>): New insns.
9704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9705 Alan Hayward <alan.hayward@arm.com>
9706 David Sherwood <david.sherwood@arm.com>
9708 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9709 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9710 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9712 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9713 Use vect_truncate_gather_scatter_offset if we can't treat the
9714 operation as a normal gather load or scatter store.
9715 (get_group_load_store_type): Take the gather_scatter_info
9716 as argument. Try using a gather load or scatter store for
9717 single-element groups.
9718 (get_load_store_type): Update calls to get_group_load_store_type
9719 and vect_use_strided_gather_scatters_p.
9721 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9722 Alan Hayward <alan.hayward@arm.com>
9723 David Sherwood <david.sherwood@arm.com>
9725 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9726 optional tree argument.
9727 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9729 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9730 but continue to use the current value as a fallback.
9731 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9732 to compare the updates.
9733 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9734 (get_load_store_type): Use it when handling a strided access.
9735 (vect_get_strided_load_store_ops): New function.
9736 (vect_get_data_ptr_increment): Likewise.
9737 (vectorizable_load): Handle strided gather loads. Always pass
9738 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9740 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9741 Alan Hayward <alan.hayward@arm.com>
9742 David Sherwood <david.sherwood@arm.com>
9744 * doc/md.texi (gather_load@var{m}): Document.
9745 (mask_gather_load@var{m}): Likewise.
9746 * genopinit.c (main): Add supports_vec_gather_load and
9747 supports_vec_gather_load_cached to target_optabs.
9748 * optabs-tree.c (init_tree_optimization_optabs): Use
9749 ggc_cleared_alloc to allocate target_optabs.
9750 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9751 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9753 * internal-fn.h (internal_load_fn_p): Declare.
9754 (internal_gather_scatter_fn_p): Likewise.
9755 (internal_fn_mask_index): Likewise.
9756 (internal_gather_scatter_fn_supported_p): Likewise.
9757 * internal-fn.c (gather_load_direct): New macro.
9758 (expand_gather_load_optab_fn): New function.
9759 (direct_gather_load_optab_supported_p): New macro.
9760 (direct_internal_fn_optab): New function.
9761 (internal_load_fn_p): Likewise.
9762 (internal_gather_scatter_fn_p): Likewise.
9763 (internal_fn_mask_index): Likewise.
9764 (internal_gather_scatter_fn_supported_p): Likewise.
9765 * optabs-query.c (supports_at_least_one_mode_p): New function.
9766 (supports_vec_gather_load_p): Likewise.
9767 * optabs-query.h (supports_vec_gather_load_p): Declare.
9768 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9769 and memory_type field.
9770 (NUM_PATTERNS): Bump to 15.
9771 * tree-vect-data-refs.c: Include internal-fn.h.
9772 (vect_gather_scatter_fn_p): New function.
9773 (vect_describe_gather_scatter_call): Likewise.
9774 (vect_check_gather_scatter): Try using internal functions for
9775 gather loads. Recognize existing calls to a gather load function.
9776 (vect_analyze_data_refs): Consider using gather loads if
9777 supports_vec_gather_load_p.
9778 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9779 (vect_get_gather_scatter_offset_type): Likewise.
9780 (vect_convert_mask_for_vectype): Likewise.
9781 (vect_add_conversion_to_patterm): Likewise.
9782 (vect_try_gather_scatter_pattern): Likewise.
9783 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9784 (vect_vect_recog_func_ptrs): Add it.
9785 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9786 internal_fn_mask_index and internal_gather_scatter_fn_p.
9787 (check_load_store_masking): Take the gather_scatter_info as an
9788 argument and handle gather loads.
9789 (vect_get_gather_scatter_ops): New function.
9790 (vectorizable_call): Check internal_load_fn_p.
9791 (vectorizable_load): Likewise. Handle gather load internal
9793 (vectorizable_store): Update call to check_load_store_masking.
9794 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9795 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9796 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9797 (aarch64_gather_scale_operand_d): New predicates.
9798 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9799 (mask_gather_load<mode>): New insns.
9801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9802 Alan Hayward <alan.hayward@arm.com>
9803 David Sherwood <david.sherwood@arm.com>
9805 * optabs.def (fold_left_plus_optab): New optab.
9806 * doc/md.texi (fold_left_plus_@var{m}): Document.
9807 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9808 * internal-fn.c (fold_left_direct): Define.
9809 (expand_fold_left_optab_fn): Likewise.
9810 (direct_fold_left_optab_supported_p): Likewise.
9811 * fold-const-call.c (fold_const_fold_left): New function.
9812 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9813 * tree-parloops.c (valid_reduction_p): New function.
9814 (gather_scalar_reductions): Use it.
9815 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9816 (vect_finish_replace_stmt): Declare.
9817 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9818 (needs_fold_left_reduction_p): New function, split out from...
9819 (vect_is_simple_reduction): ...here. Accept reductions that
9820 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9821 (vect_force_simple_reduction): Also store the reduction type in
9822 the assignment's STMT_VINFO_REDUC_TYPE.
9823 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9824 (merge_with_identity): New function.
9825 (vect_expand_fold_left): Likewise.
9826 (vectorize_fold_left_reduction): Likewise.
9827 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9828 scalar phi in place for it. Check for target support and reject
9829 cases that would reassociate the operation. Defer the transform
9830 phase to vectorize_fold_left_reduction.
9831 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9832 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9833 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9837 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9838 call to ifc_temp_var.
9840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9841 Alan Hayward <alan.hayward@arm.com>
9842 David Sherwood <david.sherwood@arm.com>
9844 * target.def (legitimize_address_displacement): Take the original
9845 offset as a poly_int.
9846 * targhooks.h (default_legitimize_address_displacement): Update
9848 * targhooks.c (default_legitimize_address_displacement): Likewise.
9849 * doc/tm.texi: Regenerate.
9850 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9851 as an argument, moving assert of ad->disp == ad->disp_term to...
9852 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9853 Try calling targetm.legitimize_address_displacement before expanding
9854 the address rather than afterwards, and adjust for the new interface.
9855 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9856 Match the new hook interface. Handle SVE addresses.
9857 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9860 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9862 * Makefile.in (OBJS): Add early-remat.o.
9863 * target.def (select_early_remat_modes): New hook.
9864 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9865 * doc/tm.texi: Regenerate.
9866 * targhooks.h (default_select_early_remat_modes): Declare.
9867 * targhooks.c (default_select_early_remat_modes): New function.
9868 * timevar.def (TV_EARLY_REMAT): New timevar.
9869 * passes.def (pass_early_remat): New pass.
9870 * tree-pass.h (make_pass_early_remat): Declare.
9871 * early-remat.c: New file.
9872 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9874 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9877 Alan Hayward <alan.hayward@arm.com>
9878 David Sherwood <david.sherwood@arm.com>
9880 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9881 vfm1 with a bound_epilog parameter.
9882 (vect_do_peeling): Update calls accordingly, and move the prologue
9883 call earlier in the function. Treat the base bound_epilog as 0 for
9884 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9885 this base when peeling for gaps.
9886 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9887 with fully-masked loops.
9888 (vect_estimate_min_profitable_iters): Handle the single peeled
9889 iteration in that case.
9891 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9892 Alan Hayward <alan.hayward@arm.com>
9893 David Sherwood <david.sherwood@arm.com>
9895 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9896 single-element interleaving even if the size is not a power of 2.
9897 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9898 accesses for single-element interleaving if the group size is
9901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9902 Alan Hayward <alan.hayward@arm.com>
9903 David Sherwood <david.sherwood@arm.com>
9905 * doc/md.texi (fold_extract_last_@var{m}): Document.
9906 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9907 * optabs.def (fold_extract_last_optab): New optab.
9908 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9909 * internal-fn.c (fold_extract_direct): New macro.
9910 (expand_fold_extract_optab_fn): Likewise.
9911 (direct_fold_extract_optab_supported_p): Likewise.
9912 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9913 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9914 EXTRACT_LAST_REDUCTION.
9915 (get_initial_def_for_reduction): Do not create an initial vector
9916 for EXTRACT_LAST_REDUCTION reductions.
9917 (vectorizable_reduction): Leave the scalar phi in place for
9918 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9919 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9920 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9921 transform phase to vectorizable_condition.
9922 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9924 (vect_finish_stmt_generation): ...here.
9925 (vect_finish_replace_stmt): New function.
9926 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9927 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9929 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9932 Alan Hayward <alan.hayward@arm.com>
9933 David Sherwood <david.sherwood@arm.com>
9935 * doc/md.texi (extract_last_@var{m}): Document.
9936 * optabs.def (extract_last_optab): New optab.
9937 * internal-fn.def (EXTRACT_LAST): New internal function.
9938 * internal-fn.c (cond_unary_direct): New macro.
9939 (expand_cond_unary_optab_fn): Likewise.
9940 (direct_cond_unary_optab_supported_p): Likewise.
9941 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9942 loops using EXTRACT_LAST.
9943 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9944 (extract_last_<mode>): ...this optab.
9945 (vec_extract<mode><Vel>): Update accordingly.
9947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9948 Alan Hayward <alan.hayward@arm.com>
9949 David Sherwood <david.sherwood@arm.com>
9951 * target.def (empty_mask_is_expensive): New hook.
9952 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9953 * doc/tm.texi: Regenerate.
9954 * targhooks.h (default_empty_mask_is_expensive): Declare.
9955 * targhooks.c (default_empty_mask_is_expensive): New function.
9956 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9957 if the target says that empty masks are expensive.
9958 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9960 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9963 Alan Hayward <alan.hayward@arm.com>
9964 David Sherwood <david.sherwood@arm.com>
9966 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9967 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9968 (vect_use_loop_mask_for_alignment_p): New function.
9969 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9970 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9971 niters_skip argument. Make sure that the first niters_skip elements
9972 of the first iteration are inactive.
9973 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9974 Update call to vect_set_loop_masks_directly.
9975 (get_misalign_in_elems): New function, split out from...
9976 (vect_gen_prolog_loop_niters): ...here.
9977 (vect_update_init_of_dr): Take a code argument that specifies whether
9978 the adjustment should be added or subtracted.
9979 (vect_update_init_of_drs): Likewise.
9980 (vect_prepare_for_masked_peels): New function.
9981 (vect_do_peeling): Skip prologue peeling if we're using a mask
9982 instead. Update call to vect_update_inits_of_drs.
9983 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9985 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9986 alignment. Do not include the number of peeled iterations in
9987 the minimum threshold in that case.
9988 (vectorizable_induction): Adjust the start value down by
9989 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9990 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9991 Take the number of skipped iterations into account when calculating
9993 * tree-vect-stmts.c (vect_gen_while_not): New function.
9995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9996 Alan Hayward <alan.hayward@arm.com>
9997 David Sherwood <david.sherwood@arm.com>
9999 * doc/sourcebuild.texi (vect_fully_masked): Document.
10000 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10001 default value to 0.
10002 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10004 (vect_analyze_loop_2): ...here. Don't check the vectorization
10005 factor against the number of loop iterations if the loop is
10008 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10009 Alan Hayward <alan.hayward@arm.com>
10010 David Sherwood <david.sherwood@arm.com>
10012 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10013 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10014 (dump_groups): Update accordingly.
10015 (iv_use::mem_type): New member variable.
10016 (address_p): New function.
10017 (record_use): Add a mem_type argument and initialize the new
10019 (record_group_use): Add a mem_type argument. Use address_p.
10020 Remove obsolete null checks of base_object. Update call to record_use.
10021 (find_interesting_uses_op): Update call to record_group_use.
10022 (find_interesting_uses_cond): Likewise.
10023 (find_interesting_uses_address): Likewise.
10024 (get_mem_type_for_internal_fn): New function.
10025 (find_address_like_use): Likewise.
10026 (find_interesting_uses_stmt): Try find_address_like_use before
10027 calling find_interesting_uses_op.
10028 (addr_offset_valid_p): Use the iv mem_type field as the type
10029 of the addressed memory.
10030 (add_autoinc_candidates): Likewise.
10031 (get_address_cost): Likewise.
10032 (split_small_address_groups_p): Use address_p.
10033 (split_address_groups): Likewise.
10034 (add_iv_candidate_for_use): Likewise.
10035 (autoinc_possible_for_pair): Likewise.
10036 (rewrite_groups): Likewise.
10037 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10038 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10039 (get_alias_ptr_type_for_ptr_address): New function.
10040 (rewrite_use_address): Rewrite address uses in calls that were
10041 identified by find_address_like_use.
10043 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10044 Alan Hayward <alan.hayward@arm.com>
10045 David Sherwood <david.sherwood@arm.com>
10047 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10049 * gimple-expr.h (is_gimple_addressable: Likewise.
10050 * gimple-expr.c (is_gimple_address): Likewise.
10051 * internal-fn.c (expand_call_mem_ref): New function.
10052 (expand_mask_load_optab_fn): Use it.
10053 (expand_mask_store_optab_fn): Likewise.
10055 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10056 Alan Hayward <alan.hayward@arm.com>
10057 David Sherwood <david.sherwood@arm.com>
10059 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10060 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10061 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10062 (cond_umax@var{mode}): Document.
10063 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10064 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10065 (cond_umin_optab, cond_umax_optab): New optabs.
10066 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10067 (COND_IOR, COND_XOR): New internal functions.
10068 * internal-fn.h (get_conditional_internal_fn): Declare.
10069 * internal-fn.c (cond_binary_direct): New macro.
10070 (expand_cond_binary_optab_fn): Likewise.
10071 (direct_cond_binary_optab_supported_p): Likewise.
10072 (get_conditional_internal_fn): New function.
10073 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10074 Cope with reduction statements that are vectorized as calls rather
10076 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10077 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10078 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10079 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10080 (UNSPEC_COND_EOR): New unspecs.
10081 (optab): Add mappings for them.
10082 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10083 (sve_int_op, sve_fp_op): New int attributes.
10085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10086 Alan Hayward <alan.hayward@arm.com>
10087 David Sherwood <david.sherwood@arm.com>
10089 * optabs.def (while_ult_optab): New optab.
10090 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10091 * internal-fn.def (WHILE_ULT): New internal function.
10092 * internal-fn.h (direct_internal_fn_supported_p): New override
10093 that takes two types as argument.
10094 * internal-fn.c (while_direct): New macro.
10095 (expand_while_optab_fn): New function.
10096 (convert_optab_supported_p): Likewise.
10097 (direct_while_optab_supported_p): New macro.
10098 * wide-int.h (wi::udiv_ceil): New function.
10099 * tree-vectorizer.h (rgroup_masks): New structure.
10100 (vec_loop_masks): New typedef.
10101 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10102 and fully_masked_p.
10103 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10104 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10105 (vect_max_vf): New function.
10106 (slpeel_make_loop_iterate_ntimes): Delete.
10107 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10108 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10109 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10110 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10111 internal-fn.h, stor-layout.h and optabs-query.h.
10112 (vect_set_loop_mask): New function.
10113 (add_preheader_seq): Likewise.
10114 (add_header_seq): Likewise.
10115 (interleave_supported_p): Likewise.
10116 (vect_maybe_permute_loop_masks): Likewise.
10117 (vect_set_loop_masks_directly): Likewise.
10118 (vect_set_loop_condition_masked): Likewise.
10119 (vect_set_loop_condition_unmasked): New function, split out from
10120 slpeel_make_loop_iterate_ntimes.
10121 (slpeel_make_loop_iterate_ntimes): Rename to..
10122 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10123 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10124 (vect_do_peeling): Update call accordingly.
10125 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10127 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10128 mask_compare_type, can_fully_mask_p and fully_masked_p.
10129 (release_vec_loop_masks): New function.
10130 (_loop_vec_info): Use it to free the loop masks.
10131 (can_produce_all_loop_masks_p): New function.
10132 (vect_get_max_nscalars_per_iter): Likewise.
10133 (vect_verify_full_masking): Likewise.
10134 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10135 retries, and free the mask rgroups before retrying. Check loop-wide
10136 reasons for disallowing fully-masked loops. Make the final decision
10137 about whether use a fully-masked loop or not.
10138 (vect_estimate_min_profitable_iters): Do not assume that peeling
10139 for the number of iterations will be needed for fully-masked loops.
10140 (vectorizable_reduction): Disable fully-masked loops.
10141 (vectorizable_live_operation): Likewise.
10142 (vect_halve_mask_nunits): New function.
10143 (vect_double_mask_nunits): Likewise.
10144 (vect_record_loop_mask): Likewise.
10145 (vect_get_loop_mask): Likewise.
10146 (vect_transform_loop): Handle the case in which the final loop
10147 iteration might handle a partial vector. Call vect_set_loop_condition
10148 instead of slpeel_make_loop_iterate_ntimes.
10149 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10150 (check_load_store_masking): New function.
10151 (prepare_load_store_mask): Likewise.
10152 (vectorizable_store): Handle fully-masked loops.
10153 (vectorizable_load): Likewise.
10154 (supportable_widening_operation): Use vect_halve_mask_nunits for
10156 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10157 (vect_gen_while): New function.
10158 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10159 (aarch64_uqdec<mode>): New insn.
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 (reduc_and_scal_optab, reduc_ior_scal_optab)
10166 (reduc_xor_scal_optab): New optabs.
10167 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10168 (reduc_xor_scal_@var{m}): Document.
10169 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10170 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10171 internal functions.
10172 * fold-const-call.c (fold_const_call): Handle them.
10173 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10174 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10175 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10176 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10177 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10178 (UNSPEC_XORV): New unspecs.
10179 (optab): Add entries for them.
10180 (BITWISEV): New int iterator.
10181 (bit_reduc_op): New int attributes.
10183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10184 Alan Hayward <alan.hayward@arm.com>
10185 David Sherwood <david.sherwood@arm.com>
10187 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10188 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10189 * optabs.def (vec_shl_insert_optab): New optab.
10190 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10191 (duplicate_and_interleave): Likewise.
10192 * tree-vect-loop.c: Include internal-fn.h.
10193 (neutral_op_for_slp_reduction): New function, split out from
10194 get_initial_defs_for_reduction.
10195 (get_initial_def_for_reduction): Handle option 2 for variable-length
10196 vectors by loading the neutral value into a vector and then shifting
10197 the initial value into element 0.
10198 (get_initial_defs_for_reduction): Replace the code argument with
10199 the neutral value calculated by neutral_op_for_slp_reduction.
10200 Use gimple_build_vector for constant-length vectors.
10201 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10202 but the first group_size elements have a neutral value.
10203 Use duplicate_and_interleave otherwise.
10204 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10205 Update call to get_initial_defs_for_reduction. Handle SLP
10206 reductions for variable-length vectors by creating one vector
10207 result for each scalar result, with the elements associated
10208 with other scalar results stubbed out with the neutral value.
10209 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10210 Require IFN_VEC_SHL_INSERT for double reductions on
10211 variable-length vectors, or SLP reductions that have
10212 a neutral value. Require can_duplicate_and_interleave_p
10213 support for variable-length unchained SLP reductions if there
10214 is no neutral value, such as for MIN/MAX reductions. Also require
10215 the number of vector elements to be a multiple of the number of
10216 SLP statements when doing variable-length unchained SLP reductions.
10217 Update call to vect_create_epilog_for_reduction.
10218 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10219 and remove initial values.
10220 (duplicate_and_interleave): Make public.
10221 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10222 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10225 Alan Hayward <alan.hayward@arm.com>
10226 David Sherwood <david.sherwood@arm.com>
10228 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10229 (can_duplicate_and_interleave_p): New function.
10230 (vect_get_and_check_slp_defs): Take the vector of statements
10231 rather than just the current one. Remove excess parentheses.
10232 Restriction rejectinon of vect_constant_def and vect_external_def
10233 for variable-length vectors to boolean types, or types for which
10234 can_duplicate_and_interleave_p is false.
10235 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10236 (duplicate_and_interleave): New function.
10237 (vect_get_constant_vectors): Use gimple_build_vector for
10238 constant-length vectors and suitable variable-length constant
10239 vectors. Use duplicate_and_interleave for other variable-length
10240 vectors. Don't defer the update when inserting new statements.
10242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10243 Alan Hayward <alan.hayward@arm.com>
10244 David Sherwood <david.sherwood@arm.com>
10246 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10247 min_profitable_iters doesn't go negative.
10249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10250 Alan Hayward <alan.hayward@arm.com>
10251 David Sherwood <david.sherwood@arm.com>
10253 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10254 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10255 * optabs.def (vec_mask_load_lanes_optab): New optab.
10256 (vec_mask_store_lanes_optab): Likewise.
10257 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10258 (MASK_STORE_LANES): Likewise.
10259 * internal-fn.c (mask_load_lanes_direct): New macro.
10260 (mask_store_lanes_direct): Likewise.
10261 (expand_mask_load_optab_fn): Handle masked operations.
10262 (expand_mask_load_lanes_optab_fn): New macro.
10263 (expand_mask_store_optab_fn): Handle masked operations.
10264 (expand_mask_store_lanes_optab_fn): New macro.
10265 (direct_mask_load_lanes_optab_supported_p): Likewise.
10266 (direct_mask_store_lanes_optab_supported_p): Likewise.
10267 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10269 (vect_load_lanes_supported): Likewise.
10270 * tree-vect-data-refs.c (strip_conversion): New function.
10271 (can_group_stmts_p): Likewise.
10272 (vect_analyze_data_ref_accesses): Use it instead of checking
10273 for a pair of assignments.
10274 (vect_store_lanes_supported): Take a masked_p parameter.
10275 (vect_load_lanes_supported): Likewise.
10276 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10277 vect_store_lanes_supported and vect_load_lanes_supported.
10278 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10279 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10280 parameter. Don't allow gaps for masked accesses.
10281 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10282 and vect_load_lanes_supported.
10283 (get_load_store_type): Take a masked_p parameter and update
10284 call to get_group_load_store_type.
10285 (vectorizable_store): Update call to get_load_store_type.
10286 Handle IFN_MASK_STORE_LANES.
10287 (vectorizable_load): Update call to get_load_store_type.
10288 Handle IFN_MASK_LOAD_LANES.
10290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10291 Alan Hayward <alan.hayward@arm.com>
10292 David Sherwood <david.sherwood@arm.com>
10294 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10296 * config/aarch64/aarch64-protos.h
10297 (aarch64_sve_struct_memory_operand_p): Declare.
10298 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10299 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10300 (VPRED, vpred): Handle SVE structure modes.
10301 * config/aarch64/constraints.md (Utx): New constraint.
10302 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10303 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10304 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10305 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10306 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10307 structure modes. Split into pieces after RA.
10308 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10309 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10311 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10312 SVE structure modes.
10313 (aarch64_classify_address): Likewise.
10314 (sizetochar): Move earlier in file.
10315 (aarch64_print_operand): Handle SVE register lists.
10316 (aarch64_array_mode): New function.
10317 (aarch64_sve_struct_memory_operand_p): Likewise.
10318 (TARGET_ARRAY_MODE): Redefine.
10320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10321 Alan Hayward <alan.hayward@arm.com>
10322 David Sherwood <david.sherwood@arm.com>
10324 * target.def (array_mode): New target hook.
10325 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10326 * doc/tm.texi: Regenerate.
10327 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10328 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10329 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10330 targetm.array_mode.
10331 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10335 Alan Hayward <alan.hayward@arm.com>
10336 David Sherwood <david.sherwood@arm.com>
10338 * fold-const.c (fold_binary_loc): Check the argument types
10339 rather than the result type when testing for a vector operation.
10341 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10343 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10344 * doc/tm.texi: Regenerate.
10346 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10347 Alan Hayward <alan.hayward@arm.com>
10348 David Sherwood <david.sherwood@arm.com>
10350 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10351 (sve): Document new AArch64 extension.
10352 * doc/md.texi (w): Extend the description of the AArch64
10353 constraint to include SVE vectors.
10354 (Upl, Upa): Document new AArch64 predicate constraints.
10355 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10357 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10358 (msve-vector-bits=): New option.
10359 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10360 SVE when these are disabled.
10361 (sve): New extension.
10362 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10363 modes. Adjust their number of units based on aarch64_sve_vg.
10364 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10365 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10366 aarch64_addr_query_type.
10367 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10368 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10369 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10370 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10371 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10372 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10373 (aarch64_simd_imm_zero_p): Delete.
10374 (aarch64_check_zero_based_sve_index_immediate): Declare.
10375 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10376 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10377 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10378 (aarch64_sve_float_mul_immediate_p): Likewise.
10379 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10380 rather than an rtx.
10381 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10382 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10383 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10384 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10385 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10386 (aarch64_regmode_natural_size): Likewise.
10387 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10388 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10390 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10391 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10392 for VG and the SVE predicate registers.
10393 (V_ALIASES): Add a "z"-prefixed alias.
10394 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10395 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10396 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10397 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10398 (REG_CLASS_NAMES): Add entries for them.
10399 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10400 and the predicate registers.
10401 (aarch64_sve_vg): Declare.
10402 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10403 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10404 (REGMODE_NATURAL_SIZE): Define.
10405 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10407 * config/aarch64/aarch64.c: Include cfgrtl.h.
10408 (simd_immediate_info): Add a constructor for series vectors,
10409 and an associated step field.
10410 (aarch64_sve_vg): New variable.
10411 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10412 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10413 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10414 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10415 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10416 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10417 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10418 (aarch64_get_mask_mode): New functions.
10419 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10420 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10421 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10422 predicate modes and predicate registers. Explicitly restrict
10423 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10424 to store a vector mode if it is recognized by
10425 aarch64_classify_vector_mode.
10426 (aarch64_regmode_natural_size): New function.
10427 (aarch64_hard_regno_caller_save_mode): Return the original mode
10429 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10430 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10431 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10432 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10434 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10435 does not overlap dest if the function is frame-related. Handle
10437 (aarch64_split_add_offset): New function.
10438 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10439 them aarch64_add_offset.
10440 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10441 and update call to aarch64_sub_sp.
10442 (aarch64_add_cfa_expression): New function.
10443 (aarch64_expand_prologue): Pass extra temporary registers to the
10444 functions above. Handle the case in which we need to emit new
10445 DW_CFA_expressions for registers that were originally saved
10446 relative to the stack pointer, but now have to be expressed
10447 relative to the frame pointer.
10448 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10450 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10451 IP0 and IP1 values for SVE frames.
10452 (aarch64_expand_vec_series): New function.
10453 (aarch64_expand_sve_widened_duplicate): Likewise.
10454 (aarch64_expand_sve_const_vector): Likewise.
10455 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10456 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10457 into the register, rather than emitting a SET directly.
10458 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10459 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10460 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10461 (offset_9bit_signed_scaled_p): New functions.
10462 (aarch64_replicate_bitmask_imm): New function.
10463 (aarch64_bitmask_imm): Use it.
10464 (aarch64_cannot_force_const_mem): Reject expressions involving
10465 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10466 (aarch64_classify_index): Handle SVE indices, by requiring
10467 a plain register index with a scale that matches the element size.
10468 (aarch64_classify_address): Handle SVE addresses. Assert that
10469 the mode of the address is VOIDmode or an integer mode.
10470 Update call to aarch64_classify_symbol.
10471 (aarch64_classify_symbolic_expression): Update call to
10472 aarch64_classify_symbol.
10473 (aarch64_const_vec_all_in_range_p): New function.
10474 (aarch64_print_vector_float_operand): Likewise.
10475 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10476 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10477 and the FP immediates 1.0 and 0.5.
10478 (aarch64_print_address_internal): Handle SVE addresses.
10479 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10480 (aarch64_regno_regclass): Handle predicate registers.
10481 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10483 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10484 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10485 (aarch64_convert_sve_vector_bits): New function.
10486 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10487 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10488 rather than an rtx.
10489 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10490 Handle SVE vector and predicate modes. Accept VL-based constants
10491 that need only one temporary register, and VL offsets that require
10492 no temporary registers.
10493 (aarch64_conditional_register_usage): Mark the predicate registers
10494 as fixed if SVE isn't available.
10495 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10496 Return true for SVE vector and predicate modes.
10497 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10498 rather than an unsigned int. Handle SVE modes.
10499 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10501 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10503 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10504 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10505 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10506 (aarch64_sve_float_mul_immediate_p): New functions.
10507 (aarch64_sve_valid_immediate): New function.
10508 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10509 Explicitly reject structure modes. Check for INDEX constants.
10510 Handle PTRUE and PFALSE constants.
10511 (aarch64_check_zero_based_sve_index_immediate): New function.
10512 (aarch64_simd_imm_zero_p): Delete.
10513 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10514 vector modes. Accept constants in the range of CNT[BHWD].
10515 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10516 ask for an Advanced SIMD mode.
10517 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10518 (aarch64_simd_vector_alignment): Handle SVE predicates.
10519 (aarch64_vectorize_preferred_vector_alignment): New function.
10520 (aarch64_simd_vector_alignment_reachable): Use it instead of
10522 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10523 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10525 (MAX_VECT_LEN): Delete.
10526 (expand_vec_perm_d): Add a vec_flags field.
10527 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10528 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10529 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10531 (aarch64_evpc_rev): Rename to...
10532 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10533 (aarch64_evpc_rev_global): New function.
10534 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10535 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10537 (aarch64_evpc_sve_tbl): New function.
10538 (aarch64_expand_vec_perm_const_1): Update after rename of
10539 aarch64_evpc_rev. Handle SVE permutes too, trying
10540 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10541 than aarch64_evpc_tbl.
10542 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10543 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10544 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10545 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10546 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10547 (aarch64_expand_sve_vcond): New functions.
10548 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10549 of aarch64_vector_mode_p.
10550 (aarch64_dwarf_poly_indeterminate_value): New function.
10551 (aarch64_compute_pressure_classes): Likewise.
10552 (aarch64_can_change_mode_class): Likewise.
10553 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10554 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10555 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10556 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10557 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10558 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10559 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10560 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10562 (Dn, Dl, Dr): Accept const as well as const_vector.
10563 (Dz): Likewise. Compare against CONST0_RTX.
10564 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10565 of "vector" where appropriate.
10566 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10567 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10568 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10569 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10570 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10571 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10572 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10573 (v_int_equiv): Extend to SVE modes.
10574 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10576 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10577 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10578 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10579 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10580 (SVE_COND_FP_CMP): New int iterators.
10581 (perm_hilo): Handle the new unpack unspecs.
10582 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10584 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10585 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10586 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10587 (aarch64_equality_operator, aarch64_constant_vector_operand)
10588 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10589 (aarch64_sve_nonimmediate_operand): Likewise.
10590 (aarch64_sve_general_operand): Likewise.
10591 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10592 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10593 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10594 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10595 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10596 (aarch64_sve_float_arith_immediate): Likewise.
10597 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10598 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10599 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10600 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10601 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10602 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10603 (aarch64_sve_float_arith_operand): Likewise.
10604 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10605 (aarch64_sve_float_mul_operand): Likewise.
10606 (aarch64_sve_vec_perm_operand): Likewise.
10607 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10608 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10609 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10610 as well as const_vector.
10611 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10612 in file. Use CONST0_RTX and CONSTM1_RTX.
10613 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10614 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10615 Use aarch64_simd_imm_zero.
10616 * config/aarch64/aarch64-sve.md: New file.
10617 * config/aarch64/aarch64.md: Include it.
10618 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10619 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10620 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10621 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10622 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10623 (sve): New attribute.
10624 (enabled): Disable instructions with the sve attribute unless
10626 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10627 aarch64_expand_mov_immediate.
10628 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10629 CNT[BHSD] immediates.
10630 (movti): Split CONST_POLY_INT moves into two halves.
10631 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10632 Split additions that need a temporary here if the destination
10633 is the stack pointer.
10634 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10635 (*add<mode>3_poly_1): New instruction.
10636 (set_clobber_cc): New expander.
10638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10640 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10641 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10642 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10643 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10644 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10645 Change innermode from fixed_mode_size to machine_mode.
10646 (simplify_subreg): Update call accordingly. Handle a constant-sized
10647 subreg of a variable-length CONST_VECTOR.
10649 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10650 Alan Hayward <alan.hayward@arm.com>
10651 David Sherwood <david.sherwood@arm.com>
10653 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10654 (add_offset_to_base): New function, split out from...
10655 (create_mem_ref): ...here. When handling a scale other than 1,
10656 check first whether the address is valid without the offset.
10657 Add it into the base if so, leaving the index and scale as-is.
10659 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10662 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10663 fold_for_warn before checking if arg2 is INTEGER_CST.
10665 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10667 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10668 (store_multiple_operation): Delete.
10669 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10670 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10671 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10672 guarded by TARGET_STRING.
10673 (rs6000_output_load_multiple): Delete.
10674 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10675 OPTION_MASK_STRING / TARGET_STRING handling.
10676 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10677 (const rs6000_opt_masks) <"string">: Change mask to 0.
10678 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10679 (MASK_STRING): Delete.
10680 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10682 (load_multiple): Delete.
10689 (store_multiple): Delete.
10696 (movmemsi_8reg): Delete.
10697 (corresponding unnamed define_insn): Delete.
10698 (movmemsi_6reg): Delete.
10699 (corresponding unnamed define_insn): Delete.
10700 (movmemsi_4reg): Delete.
10701 (corresponding unnamed define_insn): Delete.
10702 (movmemsi_2reg): Delete.
10703 (corresponding unnamed define_insn): Delete.
10704 (movmemsi_1reg): Delete.
10705 (corresponding unnamed define_insn): Delete.
10706 * config/rs6000/rs6000.opt (mno-string): New.
10707 (mstring): Replace by deprecation warning stub.
10708 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10710 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10712 * regrename.c (regrename_do_replace): If replacing the same
10713 reg multiple times, try to reuse last created gen_raw_REG.
10716 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10717 main to workaround a bug in GDB.
10719 2018-01-12 Tom de Vries <tom@codesourcery.com>
10722 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10724 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10726 PR rtl-optimization/80481
10727 * ira-color.c (get_cap_member): New function.
10728 (allocnos_conflict_by_live_ranges_p): Use it.
10729 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10730 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10732 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10735 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10736 (*saddl_se_1): Ditto.
10737 (*ssubsi_1): Ditto.
10738 (*ssubl_se_1): Ditto.
10740 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10742 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10743 rather than wi::to_widest for DR_INITs.
10744 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10745 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10746 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10748 (vect_analyze_group_access_1): Note that here.
10750 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10752 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10753 polynomial type sizes.
10755 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10757 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10758 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10759 (gimple_add_tmp_var): Likewise.
10761 2018-01-12 Martin Liska <mliska@suse.cz>
10763 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10764 (gimple_alloc_sizes): Likewise.
10765 (dump_gimple_statistics): Use PRIu64 in printf format.
10766 * gimple.h: Change uint64_t to int.
10768 2018-01-12 Martin Liska <mliska@suse.cz>
10770 * tree-core.h: Use uint64_t instead of int.
10771 * tree.c (tree_node_counts): Likewise.
10772 (tree_node_sizes): Likewise.
10773 (dump_tree_statistics): Use PRIu64 in printf format.
10775 2018-01-12 Martin Liska <mliska@suse.cz>
10777 * Makefile.in: As qsort_chk is implemented in vec.c, add
10778 vec.o to linkage of gencfn-macros.
10779 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10780 passing the info to record_node_allocation_statistics.
10781 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10783 * ggc-common.c (struct ggc_usage): Add operator== and use
10784 it in operator< and compare function.
10785 * mem-stats.h (struct mem_usage): Likewise.
10786 * vec.c (struct vec_usage): Remove operator< and compare
10787 function. Can be simply inherited.
10789 2018-01-12 Martin Jambor <mjambor@suse.cz>
10792 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10793 * tree-ssa-math-opts.c: Include domwalk.h.
10794 (convert_mult_to_fma_1): New function.
10795 (fma_transformation_info): New type.
10796 (fma_deferring_state): Likewise.
10797 (cancel_fma_deferring): New function.
10798 (result_of_phi): Likewise.
10799 (last_fma_candidate_feeds_initial_phi): Likewise.
10800 (convert_mult_to_fma): Added deferring logic, split actual
10801 transformation to convert_mult_to_fma_1.
10802 (math_opts_dom_walker): New type.
10803 (math_opts_dom_walker::after_dom_children): New method, body moved
10804 here from pass_optimize_widening_mul::execute, added deferring logic
10806 (pass_optimize_widening_mul::execute): Moved most of code to
10807 math_opts_dom_walker::after_dom_children.
10808 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10809 * config/i386/i386.c (ix86_option_override_internal): Added
10810 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10812 2018-01-12 Richard Biener <rguenther@suse.de>
10815 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10816 inline instance vars.
10818 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10821 * config/rx/rx.c (rx_is_restricted_memory_address):
10822 Handle SUBREG case.
10824 2018-01-12 Richard Biener <rguenther@suse.de>
10826 PR tree-optimization/80846
10827 * target.def (split_reduction): New target hook.
10828 * targhooks.c (default_split_reduction): New function.
10829 * targhooks.h (default_split_reduction): Declare.
10830 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10831 target requests first reduce vectors by combining low and high
10833 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10834 (get_vectype_for_scalar_type_and_size): Export.
10835 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10836 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10837 * doc/tm.texi: Regenerate.
10838 * config/i386/i386.c (ix86_split_reduction): Implement
10839 TARGET_VECTORIZE_SPLIT_REDUCTION.
10841 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10844 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10845 in PIC mode except for TARGET_VXWORKS_RTP.
10846 * config/sparc/sparc.c: Include cfgrtl.h.
10847 (TARGET_INIT_PIC_REG): Define.
10848 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10849 (sparc_pic_register_p): New predicate.
10850 (sparc_legitimate_address_p): Use it.
10851 (sparc_legitimize_pic_address): Likewise.
10852 (sparc_delegitimize_address): Likewise.
10853 (sparc_mode_dependent_address_p): Likewise.
10854 (gen_load_pcrel_sym): Remove 4th parameter.
10855 (load_got_register): Adjust call to above. Remove obsolete stuff.
10856 (sparc_expand_prologue): Do not call load_got_register here.
10857 (sparc_flat_expand_prologue): Likewise.
10858 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10859 (sparc_use_pseudo_pic_reg): New function.
10860 (sparc_init_pic_reg): Likewise.
10861 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10862 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10864 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10866 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10867 Add item for branch_cost.
10869 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10871 PR rtl-optimization/83565
10872 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10873 not extend the result to a larger mode for rotate operations.
10874 (num_sign_bit_copies1): Likewise.
10876 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10879 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10881 Use values-Xc.o for -pedantic.
10882 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10884 2018-01-12 Martin Liska <mliska@suse.cz>
10887 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10889 (possible_polymorphic_call_targets): Use it.
10890 (ipa_devirt): Likewise.
10892 2018-01-12 Martin Liska <mliska@suse.cz>
10894 * profile-count.h (enum profile_quality): Use 0 as invalid
10895 enum value of profile_quality.
10897 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10899 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10900 -mext-string options.
10902 2018-01-12 Richard Biener <rguenther@suse.de>
10904 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10905 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10906 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10908 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10910 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10912 * configure.ac (--with-long-double-format): Add support for the
10913 configuration option to change the default long double format on
10915 * config.gcc (powerpc*-linux*-*): Likewise.
10916 * configure: Regenerate.
10917 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10918 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10919 used without modification.
10921 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10923 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10924 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10925 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10926 MISC_BUILTIN_SPEC_BARRIER.
10927 (rs6000_init_builtins): Likewise.
10928 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10930 (speculation_barrier): New define_insn.
10931 * doc/extend.texi: Document __builtin_speculation_barrier.
10933 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10936 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10937 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10938 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10940 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10941 integral modes instead of "ss" and "sd".
10942 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10943 vectors with 32-bit and 64-bit elements.
10944 (vecdupssescalarmodesuffix): New mode attribute.
10945 (vec_dup<mode>): Use it.
10947 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10950 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10951 frame if argument is passed on stack.
10953 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10956 * ree.c (combine_reaching_defs): Optimize also
10957 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10958 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10960 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10962 PR middle-end/83189
10963 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10965 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10967 PR middle-end/83718
10968 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10969 after they are computed.
10971 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10973 PR tree-optimization/83695
10974 * gimple-loop-linterchange.cc
10975 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10976 reset cached scev information after interchange.
10977 (pass_linterchange::execute): Remove call to scev_reset_htab.
10979 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10981 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10982 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10983 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10984 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10985 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10986 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10987 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10988 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10989 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10990 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10991 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10992 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10993 (V_lane_reg): Likewise.
10994 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10996 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10997 (vfmal_lane_low<mode>_intrinsic,
10998 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10999 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11000 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11001 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11002 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11003 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11005 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11007 * config/arm/arm-cpus.in (fp16fml): New feature.
11008 (ALL_SIMD): Add fp16fml.
11009 (armv8.2-a): Add fp16fml as an option.
11010 (armv8.3-a): Likewise.
11011 (armv8.4-a): Add fp16fml as part of fp16.
11012 * config/arm/arm.h (TARGET_FP16FML): Define.
11013 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11015 * config/arm/arm-modes.def (V2HF): Define.
11016 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11017 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11018 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11019 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11020 vfmsl_low, vfmsl_high): New set of builtins.
11021 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11022 (vfml_op): New code attribute.
11023 (VFMLHALVES): New int iterator.
11024 (VFML, VFMLSEL): New mode attributes.
11025 (V_reg): Define mapping for V2HF.
11026 (V_hi, V_lo): New mode attributes.
11027 (VF_constraint): Likewise.
11028 (vfml_half, vfml_half_selector): New int attributes.
11029 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11031 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11032 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11034 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11035 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11036 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11037 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11039 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11040 Document new effective target and option set.
11042 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11044 * config/arm/arm-cpus.in (armv8_4): New feature.
11045 (ARMv8_4a): New fgroup.
11046 (armv8.4-a): New arch.
11047 * config/arm/arm-tables.opt: Regenerate.
11048 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11049 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11050 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11051 Add matching rules for -march=armv8.4-a and extensions.
11052 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11054 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11057 * config/rx/rx.md (BW): New mode attribute.
11058 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11060 2018-01-11 Richard Biener <rguenther@suse.de>
11062 PR tree-optimization/83435
11063 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11064 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11065 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11067 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11068 Alan Hayward <alan.hayward@arm.com>
11069 David Sherwood <david.sherwood@arm.com>
11071 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11073 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11074 (aarch64_print_address_internal): Use it to check for a zero offset.
11076 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11077 Alan Hayward <alan.hayward@arm.com>
11078 David Sherwood <david.sherwood@arm.com>
11080 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11081 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11082 Return a poly_int64 rather than a HOST_WIDE_INT.
11083 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11084 rather than a HOST_WIDE_INT.
11085 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11086 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11087 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11088 final_offset from HOST_WIDE_INT to poly_int64.
11089 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11090 to_constant when getting the number of units in an Advanced SIMD
11092 (aarch64_builtin_vectorized_function): Check for a constant number
11094 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11096 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11097 attribute instead of GET_MODE_NUNITS.
11098 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11099 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11100 GET_MODE_SIZE for fixed-size registers.
11101 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11102 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11103 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11104 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11105 (aarch64_print_operand, aarch64_print_address_internal)
11106 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11107 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11108 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11109 Handle polynomial GET_MODE_SIZE.
11110 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11111 wider than SImode without modification.
11112 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11113 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11114 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11115 passing and returning SVE modes.
11116 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11117 rather than GEN_INT.
11118 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11119 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11120 (aarch64_allocate_and_probe_stack_space): Likewise.
11121 (aarch64_layout_frame): Cope with polynomial offsets.
11122 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11123 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11124 polynomial offsets.
11125 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11126 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11127 poly_int64 rather than a HOST_WIDE_INT.
11128 (aarch64_get_separate_components, aarch64_process_components)
11129 (aarch64_expand_prologue, aarch64_expand_epilogue)
11130 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11131 (aarch64_anchor_offset): New function, split out from...
11132 (aarch64_legitimize_address): ...here.
11133 (aarch64_builtin_vectorization_cost): Handle polynomial
11134 TYPE_VECTOR_SUBPARTS.
11135 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11137 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11138 number of elements from the PARALLEL rather than the mode.
11139 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11140 rather than GET_MODE_BITSIZE.
11141 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11142 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11143 (aarch64_expand_vec_perm_const_1): Handle polynomial
11144 d->perm.length () and d->perm elements.
11145 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11146 Apply to_constant to d->perm elements.
11147 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11148 polynomial CONST_VECTOR_NUNITS.
11149 (aarch64_move_pointer): Take amount as a poly_int64 rather
11151 (aarch64_progress_pointer): Avoid temporary variable.
11152 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11153 the mode attribute instead of GET_MODE.
11155 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11156 Alan Hayward <alan.hayward@arm.com>
11157 David Sherwood <david.sherwood@arm.com>
11159 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11160 x exists before using it.
11161 (aarch64_add_constant_internal): Rename to...
11162 (aarch64_add_offset_1): ...this. Replace regnum with separate
11163 src and dest rtxes. Handle the case in which they're different,
11164 including when the offset is zero. Replace scratchreg with an rtx.
11165 Use 2 additions if there is no spare register into which we can
11166 move a 16-bit constant.
11167 (aarch64_add_constant): Delete.
11168 (aarch64_add_offset): Replace reg with separate src and dest
11169 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11170 Use aarch64_add_offset_1.
11171 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11172 an rtx rather than an int. Take the delta as a poly_int64
11173 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11174 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11175 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11176 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11177 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11178 and aarch64_add_sp.
11179 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11180 aarch64_add_constant.
11182 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11184 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11185 Use scalar_float_mode.
11187 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11189 * config/aarch64/aarch64-simd.md
11190 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11191 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11192 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11193 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11194 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11195 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11196 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11197 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11198 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11199 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11201 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11204 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11205 targ_options->x_arm_arch_string is non NULL.
11207 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11209 * config/aarch64/aarch64.h
11210 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11212 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11215 * expmed.c (emit_store_flag_force): Swap if const op0
11216 and change VOIDmode to mode of op0.
11218 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11220 PR rtl-optimization/83761
11221 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11222 than bytes to mode_for_size.
11224 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11226 PR middle-end/83189
11227 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11228 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11231 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11233 PR middle-end/83575
11234 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11235 when in layout mode.
11236 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11237 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11240 2018-01-10 Michael Collison <michael.collison@arm.com>
11242 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11243 * config/aarch64/aarch64-option-extension.def: Add
11244 AARCH64_OPT_EXTENSION of 'fp16fml'.
11245 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11246 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11247 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11248 * config/aarch64/constraints.md (Ui7): New constraint.
11249 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11250 (VFMLA_SEL_W): Ditto.
11253 (VFMLA16_LOW): New int iterator.
11254 (VFMLA16_HIGH): Ditto.
11255 (UNSPEC_FMLAL): New unspec.
11256 (UNSPEC_FMLSL): Ditto.
11257 (UNSPEC_FMLAL2): Ditto.
11258 (UNSPEC_FMLSL2): Ditto.
11259 (f16mac): New code attribute.
11260 * config/aarch64/aarch64-simd-builtins.def
11261 (aarch64_fmlal_lowv2sf): Ditto.
11262 (aarch64_fmlsl_lowv2sf): Ditto.
11263 (aarch64_fmlalq_lowv4sf): Ditto.
11264 (aarch64_fmlslq_lowv4sf): Ditto.
11265 (aarch64_fmlal_highv2sf): Ditto.
11266 (aarch64_fmlsl_highv2sf): Ditto.
11267 (aarch64_fmlalq_highv4sf): Ditto.
11268 (aarch64_fmlslq_highv4sf): Ditto.
11269 (aarch64_fmlal_lane_lowv2sf): Ditto.
11270 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11271 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11272 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11273 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11274 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11275 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11276 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11277 (aarch64_fmlal_lane_highv2sf): Ditto.
11278 (aarch64_fmlsl_lane_highv2sf): Ditto.
11279 (aarch64_fmlal_laneq_highv2sf): Ditto.
11280 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11281 (aarch64_fmlalq_lane_highv4sf): Ditto.
11282 (aarch64_fmlsl_lane_highv4sf): Ditto.
11283 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11284 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11285 * config/aarch64/aarch64-simd.md:
11286 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11287 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11288 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11289 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11290 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11291 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11292 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11293 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11294 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11295 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11296 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11297 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11298 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11299 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11300 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11301 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11302 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11303 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11304 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11305 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11306 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11307 (vfmlsl_low_u32): Ditto.
11308 (vfmlalq_low_u32): Ditto.
11309 (vfmlslq_low_u32): Ditto.
11310 (vfmlal_high_u32): Ditto.
11311 (vfmlsl_high_u32): Ditto.
11312 (vfmlalq_high_u32): Ditto.
11313 (vfmlslq_high_u32): Ditto.
11314 (vfmlal_lane_low_u32): Ditto.
11315 (vfmlsl_lane_low_u32): Ditto.
11316 (vfmlal_laneq_low_u32): Ditto.
11317 (vfmlsl_laneq_low_u32): Ditto.
11318 (vfmlalq_lane_low_u32): Ditto.
11319 (vfmlslq_lane_low_u32): Ditto.
11320 (vfmlalq_laneq_low_u32): Ditto.
11321 (vfmlslq_laneq_low_u32): Ditto.
11322 (vfmlal_lane_high_u32): Ditto.
11323 (vfmlsl_lane_high_u32): Ditto.
11324 (vfmlal_laneq_high_u32): Ditto.
11325 (vfmlsl_laneq_high_u32): Ditto.
11326 (vfmlalq_lane_high_u32): Ditto.
11327 (vfmlslq_lane_high_u32): Ditto.
11328 (vfmlalq_laneq_high_u32): Ditto.
11329 (vfmlslq_laneq_high_u32): Ditto.
11330 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11331 (AARCH64_FL_FOR_ARCH8_4): New.
11332 (AARCH64_ISA_F16FML): New ISA flag.
11333 (TARGET_F16FML): New feature flag for fp16fml.
11334 (doc/invoke.texi): Document new fp16fml option.
11336 2018-01-10 Michael Collison <michael.collison@arm.com>
11338 * config/aarch64/aarch64-builtins.c:
11339 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11340 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11341 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11342 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11343 (AARCH64_ISA_SHA3): New ISA flag.
11344 (TARGET_SHA3): New feature flag for sha3.
11345 * config/aarch64/iterators.md (sha512_op): New int attribute.
11346 (CRYPTO_SHA512): New int iterator.
11347 (UNSPEC_SHA512H): New unspec.
11348 (UNSPEC_SHA512H2): Ditto.
11349 (UNSPEC_SHA512SU0): Ditto.
11350 (UNSPEC_SHA512SU1): Ditto.
11351 * config/aarch64/aarch64-simd-builtins.def
11352 (aarch64_crypto_sha512hqv2di): New builtin.
11353 (aarch64_crypto_sha512h2qv2di): Ditto.
11354 (aarch64_crypto_sha512su0qv2di): Ditto.
11355 (aarch64_crypto_sha512su1qv2di): Ditto.
11356 (aarch64_eor3qv8hi): Ditto.
11357 (aarch64_rax1qv2di): Ditto.
11358 (aarch64_xarqv2di): Ditto.
11359 (aarch64_bcaxqv8hi): Ditto.
11360 * config/aarch64/aarch64-simd.md:
11361 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11362 (aarch64_crypto_sha512su0qv2di): Ditto.
11363 (aarch64_crypto_sha512su1qv2di): Ditto.
11364 (aarch64_eor3qv8hi): Ditto.
11365 (aarch64_rax1qv2di): Ditto.
11366 (aarch64_xarqv2di): Ditto.
11367 (aarch64_bcaxqv8hi): Ditto.
11368 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11369 (vsha512h2q_u64): Ditto.
11370 (vsha512su0q_u64): Ditto.
11371 (vsha512su1q_u64): Ditto.
11372 (veor3q_u16): Ditto.
11373 (vrax1q_u64): Ditto.
11374 (vxarq_u64): Ditto.
11375 (vbcaxq_u16): Ditto.
11376 * config/arm/types.md (crypto_sha512): New type attribute.
11377 (crypto_sha3): Ditto.
11378 (doc/invoke.texi): Document new sha3 option.
11380 2018-01-10 Michael Collison <michael.collison@arm.com>
11382 * config/aarch64/aarch64-builtins.c:
11383 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11384 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11385 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11386 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11387 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11388 (AARCH64_ISA_SM4): New ISA flag.
11389 (TARGET_SM4): New feature flag for sm4.
11390 * config/aarch64/aarch64-simd-builtins.def
11391 (aarch64_sm3ss1qv4si): Ditto.
11392 (aarch64_sm3tt1aq4si): Ditto.
11393 (aarch64_sm3tt1bq4si): Ditto.
11394 (aarch64_sm3tt2aq4si): Ditto.
11395 (aarch64_sm3tt2bq4si): Ditto.
11396 (aarch64_sm3partw1qv4si): Ditto.
11397 (aarch64_sm3partw2qv4si): Ditto.
11398 (aarch64_sm4eqv4si): Ditto.
11399 (aarch64_sm4ekeyqv4si): Ditto.
11400 * config/aarch64/aarch64-simd.md:
11401 (aarch64_sm3ss1qv4si): Ditto.
11402 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11403 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11404 (aarch64_sm4eqv4si): Ditto.
11405 (aarch64_sm4ekeyqv4si): Ditto.
11406 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11407 (sm3part_op): Ditto.
11408 (CRYPTO_SM3TT): Ditto.
11409 (CRYPTO_SM3PART): Ditto.
11410 (UNSPEC_SM3SS1): New unspec.
11411 (UNSPEC_SM3TT1A): Ditto.
11412 (UNSPEC_SM3TT1B): Ditto.
11413 (UNSPEC_SM3TT2A): Ditto.
11414 (UNSPEC_SM3TT2B): Ditto.
11415 (UNSPEC_SM3PARTW1): Ditto.
11416 (UNSPEC_SM3PARTW2): Ditto.
11417 (UNSPEC_SM4E): Ditto.
11418 (UNSPEC_SM4EKEY): Ditto.
11419 * config/aarch64/constraints.md (Ui2): New constraint.
11420 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11421 * config/arm/types.md (crypto_sm3): New type attribute.
11422 (crypto_sm4): Ditto.
11423 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11424 (vsm3tt1aq_u32): Ditto.
11425 (vsm3tt1bq_u32): Ditto.
11426 (vsm3tt2aq_u32): Ditto.
11427 (vsm3tt2bq_u32): Ditto.
11428 (vsm3partw1q_u32): Ditto.
11429 (vsm3partw2q_u32): Ditto.
11430 (vsm4eq_u32): Ditto.
11431 (vsm4ekeyq_u32): Ditto.
11432 (doc/invoke.texi): Document new sm4 option.
11434 2018-01-10 Michael Collison <michael.collison@arm.com>
11436 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11437 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11438 (AARCH64_FL_FOR_ARCH8_4): New.
11439 (AARCH64_FL_V8_4): New flag.
11440 (doc/invoke.texi): Document new armv8.4-a option.
11442 2018-01-10 Michael Collison <michael.collison@arm.com>
11444 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11445 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11446 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11447 * config/aarch64/aarch64-option-extension.def: Add
11448 AARCH64_OPT_EXTENSION of 'sha2'.
11449 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11450 (crypto): Disable sha2 and aes if crypto disabled.
11451 (crypto): Enable aes and sha2 if enabled.
11452 (simd): Disable sha2 and aes if simd disabled.
11453 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11455 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11456 (TARGET_SHA2): New feature flag for sha2.
11457 (TARGET_AES): New feature flag for aes.
11458 * config/aarch64/aarch64-simd.md:
11459 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11460 conditional on TARGET_AES.
11461 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11462 (aarch64_crypto_sha1hsi): Make pattern conditional
11464 (aarch64_crypto_sha1hv4si): Ditto.
11465 (aarch64_be_crypto_sha1hv4si): Ditto.
11466 (aarch64_crypto_sha1su1v4si): Ditto.
11467 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11468 (aarch64_crypto_sha1su0v4si): Ditto.
11469 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11470 (aarch64_crypto_sha256su0v4si): Ditto.
11471 (aarch64_crypto_sha256su1v4si): Ditto.
11472 (doc/invoke.texi): Document new aes and sha2 options.
11474 2018-01-10 Martin Sebor <msebor@redhat.com>
11476 PR tree-optimization/83781
11477 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11480 2018-01-11 Martin Sebor <msebor@gmail.com>
11481 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11483 PR tree-optimization/83501
11484 PR tree-optimization/81703
11486 * tree-ssa-strlen.c (get_string_cst): Rename...
11487 (get_string_len): ...to this. Handle global constants.
11488 (handle_char_store): Adjust.
11490 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11491 Jim Wilson <jimw@sifive.com>
11493 * config/riscv/riscv-protos.h (riscv_output_return): New.
11494 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11495 (riscv_attribute_table, riscv_output_return),
11496 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11497 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11498 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11500 (riscv_expand_prologue): Add early return for naked function.
11501 (riscv_expand_epilogue): Likewise.
11502 (riscv_function_ok_for_sibcall): Return false for naked function.
11503 (riscv_set_current_function): New.
11504 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11505 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11506 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11507 * doc/extend.texi (RISC-V Function Attributes): New.
11509 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11511 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11512 check for 128-bit long double before checking TCmode.
11513 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11514 128-bit long doubles before checking TFmode or TCmode.
11515 (FLOAT128_IBM_P): Likewise.
11517 2018-01-10 Martin Sebor <msebor@redhat.com>
11519 PR tree-optimization/83671
11520 * builtins.c (c_strlen): Unconditionally return zero for the empty
11522 Use -Warray-bounds for warnings.
11523 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11524 for non-constant array indices with COMPONENT_REF, arrays of
11525 arrays, and pointers to arrays.
11526 (gimple_fold_builtin_strlen): Determine and set length range for
11527 non-constant character arrays.
11529 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11531 PR middle-end/81897
11532 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11535 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11537 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11539 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11542 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11543 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11544 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11545 indexed_or_indirect_operand predicate.
11546 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11547 (*vsx_le_perm_load_v8hi): Likewise.
11548 (*vsx_le_perm_load_v16qi): Likewise.
11549 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11550 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11551 (*vsx_le_perm_store_v8hi): Likewise.
11552 (*vsx_le_perm_store_v16qi): Likewise.
11553 (eight unnamed splitters): Likewise.
11555 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11557 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11558 * config/rs6000/emmintrin.h: Likewise.
11559 * config/rs6000/mmintrin.h: Likewise.
11560 * config/rs6000/xmmintrin.h: Likewise.
11562 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11565 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11567 * tree.c (tree_nop_conversion): Return true for location wrapper
11569 (maybe_wrap_with_location): New function.
11570 (selftest::check_strip_nops): New function.
11571 (selftest::test_location_wrappers): New function.
11572 (selftest::tree_c_tests): Call it.
11573 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11574 (maybe_wrap_with_location): New decl.
11575 (EXPR_LOCATION_WRAPPER_P): New macro.
11576 (location_wrapper_p): New inline function.
11577 (tree_strip_any_location_wrapper): New inline function.
11579 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11582 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11583 stack_realign_offset for the largest alignment of stack slot
11585 (ix86_find_max_used_stack_alignment): New function.
11586 (ix86_finalize_stack_frame_flags): Use it. Set
11587 max_used_stack_alignment if we don't realign stack.
11588 * config/i386/i386.h (machine_function): Add
11589 max_used_stack_alignment.
11591 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11593 * config/arm/arm.opt (-mbranch-cost): New option.
11594 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11597 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11600 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11601 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11603 2018-01-10 Richard Biener <rguenther@suse.de>
11606 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11607 early out so it also covers the case where we have a non-NULL
11610 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11612 PR tree-optimization/83753
11613 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11614 for non-strided grouped accesses if the number of elements is 1.
11616 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11619 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11620 * i386.h (TARGET_USE_GATHER): Define.
11621 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11623 2018-01-10 Martin Liska <mliska@suse.cz>
11626 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11627 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11629 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11630 CLEANUP_NO_PARTITIONING is not set.
11632 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11634 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11635 for vectors, as a partial revert of r254296.
11636 * rtl.h (const_vec_p): Delete.
11637 (const_vec_duplicate_p): Don't test for vector CONSTs.
11638 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11639 * expmed.c (make_tree): Likewise.
11642 * common.md (E, F): Use CONSTANT_P instead of checking for
11644 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11645 checking for CONST_VECTOR.
11647 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11649 PR middle-end/83575
11650 * predict.c (force_edge_cold): Handle in more sane way edges
11651 with no prediction.
11653 2018-01-09 Carl Love <cel@us.ibm.com>
11655 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11657 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11658 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11659 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11660 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11661 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11662 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11663 * config/rs6000/rs6000-protos.h: Add extern defition for
11664 rs6000_generate_float2_double_code.
11665 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11667 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11668 (float2_v2df): Add define_expand.
11670 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11673 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11674 op_mode in the force_to_mode call.
11676 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11678 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11679 instead of checking each element individually.
11680 (aarch64_evpc_uzp): Likewise.
11681 (aarch64_evpc_zip): Likewise.
11682 (aarch64_evpc_ext): Likewise.
11683 (aarch64_evpc_rev): Likewise.
11684 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11685 instead of checking each element individually. Return true without
11687 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11688 whether all selected elements come from the same input, instead of
11689 checking each element individually. Remove calls to gen_rtx_REG,
11690 start_sequence and end_sequence and instead assert that no rtl is
11693 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11695 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11696 order of HIGH and CONST checks.
11698 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11700 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11701 if the destination isn't an SSA_NAME.
11703 2018-01-09 Richard Biener <rguenther@suse.de>
11705 PR tree-optimization/83668
11706 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11708 (canonicalize_loop_form): ... here, renamed from ...
11709 (canonicalize_loop_closed_ssa_form): ... this and amended to
11710 swap successor edges for loop exit blocks to make us use
11711 the RPO order we need for initial schedule generation.
11713 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11715 PR tree-optimization/64811
11716 * match.pd: When optimizing comparisons with Inf, avoid
11717 introducing or losing exceptions from comparisons with NaN.
11719 2018-01-09 Martin Liska <mliska@suse.cz>
11722 * asan.c (shadow_mem_size): Add gcc_assert.
11724 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11726 Don't save registers in main().
11729 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11730 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11731 * config/avr/avr.c (avr_set_current_function): Don't error if
11732 naked, OS_task or OS_main are specified at the same time.
11733 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11735 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11737 * common/config/avr/avr-common.c (avr_option_optimization_table):
11738 Switch on -mmain-is-OS_task for optimizing compilations.
11740 2018-01-09 Richard Biener <rguenther@suse.de>
11742 PR tree-optimization/83572
11743 * graphite.c: Include cfganal.h.
11744 (graphite_transform_loops): Connect infinite loops to exit
11745 and remove fake edges at the end.
11747 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11749 * ipa-inline.c (edge_badness): Revert accidental checkin.
11751 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11754 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11755 symbols; not inline clones.
11757 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11760 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11761 hard registers. Formatting fixes.
11763 PR preprocessor/83722
11764 * gcc.c (try_generate_repro): Pass
11765 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11766 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11769 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11770 Kito Cheng <kito.cheng@gmail.com>
11772 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11773 (riscv_leaf_function_p): Delete.
11774 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11776 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11778 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11780 (do_ifelse): New function.
11781 (do_isel): New function.
11782 (do_sub3): New function.
11783 (do_add3): New function.
11784 (do_load_mask_compare): New function.
11785 (do_overlap_load_compare): New function.
11786 (expand_compare_loop): New function.
11787 (expand_block_compare): Call expand_compare_loop() when appropriate.
11788 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11789 option description.
11790 (-mblock-compare-inline-loop-limit): New option.
11792 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11795 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11796 Reverse order of second and third operands in first alternative.
11797 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11798 of first and second elements in UNSPEC_VPERMR vector.
11799 (altivec_expand_vec_perm_le): Likewise.
11801 2018-01-08 Jeff Law <law@redhat.com>
11803 PR rtl-optimizatin/81308
11804 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11805 (process_switch): If group_case_labels makes a change, then set
11807 (pass_convert_switch::execute): If a switch is converted, then
11808 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11810 PR rtl-optimization/81308
11811 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11814 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11816 PR target/83663 - Revert r255946
11817 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11818 generation for cases where splatting a value is not useful.
11819 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11820 across a vec_duplicate and a paradoxical subreg forming a vector
11821 mode to a vec_concat.
11823 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11825 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11826 -march=armv8.3-a variants.
11827 * config/arm/t-multilib: Likewise.
11828 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11830 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11832 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11834 (cceq_ior_compare_complement): Give it a name so I can use it, and
11835 change boolean_or_operator predicate to boolean_operator so it can
11836 be used to generate a crand.
11837 (eqne): New code iterator.
11838 (bd/bd_neg): New code_attrs.
11839 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11840 a single define_insn.
11841 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11842 decrement (bdnzt/bdnzf/bdzt/bdzf).
11843 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11844 with the new names of the branch decrement patterns, and added the
11845 names of the branch decrement conditional patterns.
11847 2018-01-08 Richard Biener <rguenther@suse.de>
11849 PR tree-optimization/83563
11850 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11853 2018-01-08 Richard Biener <rguenther@suse.de>
11855 PR middle-end/83713
11856 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11858 2018-01-08 Richard Biener <rguenther@suse.de>
11860 PR tree-optimization/83685
11861 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11862 references to abnormals.
11864 2018-01-08 Richard Biener <rguenther@suse.de>
11867 * dwarf2out.c (output_indirect_strings): Handle empty
11868 skeleton_debug_str_hash.
11869 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11871 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11873 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11874 (emit_store_direct): Likewise.
11875 (arc_trampoline_adjust_address): Likewise.
11876 (arc_asm_trampoline_template): New function.
11877 (arc_initialize_trampoline): Use asm_trampoline_template.
11878 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11879 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11880 * config/arc/arc.md (flush_icache): Delete pattern.
11882 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11884 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11885 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11888 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11891 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11892 by not USED_FOR_TARGET.
11893 (make_pass_resolve_sw_modes): Likewise.
11895 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11897 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11900 2018-01-08 Richard Biener <rguenther@suse.de>
11902 PR middle-end/83580
11903 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11905 2018-01-08 Richard Biener <rguenther@suse.de>
11907 PR middle-end/83517
11908 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11910 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11912 PR middle-end/81897
11913 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11914 basic blocks with a small number of successors.
11915 (convert_control_dep_chain_into_preds): Improve handling of
11917 (dump_predicates): Split apart into...
11918 (dump_pred_chain): ...here...
11919 (dump_pred_info): ...and here.
11920 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11921 (can_chain_union_be_invalidated_p): Improve check for invalidation
11923 (uninit_uses_cannot_happen): Avoid unnecessary if
11924 convert_control_dep_chain_into_preds yielded nothing.
11926 2018-01-06 Martin Sebor <msebor@redhat.com>
11928 PR tree-optimization/83640
11929 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11930 subtracting negative offset from size.
11931 (builtin_access::overlap): Adjust offset bounds of the access to fall
11932 within the size of the object if possible.
11934 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11936 PR rtl-optimization/83699
11937 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11938 extract_bit_field_as_subreg to cases in which the extracted
11939 value is also a vector.
11941 * lra-constraints.c (process_alt_operands): Test for the equivalence
11942 substitutions when detecting a possible reload cycle.
11944 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11947 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11948 by default if flag_selective_schedling{,2}. Formatting fixes.
11950 PR rtl-optimization/83682
11951 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11952 if it has non-VECTOR_MODE element mode.
11953 (vec_duplicate_p): Likewise.
11955 PR middle-end/83694
11956 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11957 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11959 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11962 * config/i386/i386-builtin.def
11963 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11964 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11965 Require also OPTION_MASK_ISA_AVX512F in addition to
11966 OPTION_MASK_ISA_GFNI.
11967 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11968 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11969 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11970 to OPTION_MASK_ISA_GFNI.
11971 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11972 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11973 OPTION_MASK_ISA_AVX512BW.
11974 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11975 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11976 addition to OPTION_MASK_ISA_GFNI.
11977 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11978 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11979 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11980 to OPTION_MASK_ISA_GFNI.
11981 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11982 a requirement for all ISAs rather than any of them with a few
11984 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11986 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11987 bitmasks to be enabled with 3 exceptions, instead of requiring any
11988 enabled ISA with lots of exceptions.
11989 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11990 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11991 Change avx512bw in isa attribute to avx512f.
11992 * config/i386/sgxintrin.h: Add license boilerplate.
11993 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11994 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11995 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11996 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11998 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11999 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12000 temporarily sse2 rather than sse if not enabled already.
12003 * config/i386/sse.md (VI248_VLBW): Rename to ...
12004 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12005 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12006 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12007 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12008 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12009 mode iterator instead of VI248_VLBW.
12011 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12013 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12014 (record_modified): Skip clobbers; add debug output.
12015 (param_change_prob): Use sreal frequencies.
12017 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12019 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12020 punt for user-aligned variables.
12022 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12024 * tree-chrec.c (chrec_contains_symbols): Return true for
12027 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12030 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12031 of (x|y) == x for BICS pattern.
12033 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12035 PR tree-optimization/83605
12036 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12037 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12040 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12042 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12043 * config/epiphany/rtems.h: New file.
12045 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12046 Uros Bizjak <ubizjak@gmail.com>
12049 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12050 QIreg_operand instead of register_operand predicate.
12051 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12052 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12053 comments instead of -fmitigate[-_]rop.
12055 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12058 * cgraphunit.c (symbol_table::compile): Switch to text_section
12059 before calling assembly_start debug hook.
12060 * run-rtl-passes.c (run_rtl_passes): Likewise.
12063 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12065 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12066 range_int_cst_p rather than !symbolic_range_p before calling
12067 extract_range_from_multiplicative_op_1.
12069 2018-01-04 Jeff Law <law@redhat.com>
12071 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12072 redundant test in assertion.
12074 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12076 * doc/rtl.texi: Document machine_mode wrapper classes.
12078 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12080 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12081 using tree_to_uhwi.
12083 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12085 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12086 the VEC_PERM_EXPR fold to fail.
12088 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12091 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12092 to switched_sections.
12094 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12097 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12098 test for d.testing.
12100 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12103 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12104 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12106 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12109 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12110 is BLKmode and bitpos not zero or mode change is needed.
12112 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12115 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12118 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12121 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12122 instead of MULT rtx. Update all corresponding splitters.
12123 (*saddl_se): Ditto.
12124 (*ssub<modesuffix>): Ditto.
12125 (*ssubl_se): Ditto.
12126 (*cmp_sadd_di): Update split patterns.
12127 (*cmp_sadd_si): Ditto.
12128 (*cmp_sadd_sidi): Ditto.
12129 (*cmp_ssub_di): Ditto.
12130 (*cmp_ssub_si): Ditto.
12131 (*cmp_ssub_sidi): Ditto.
12132 * config/alpha/predicates.md (const23_operand): New predicate.
12133 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12134 Look for ASHIFT, not MULT inner operand.
12135 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12137 2018-01-04 Martin Liska <mliska@suse.cz>
12139 PR gcov-profile/83669
12140 * gcov.c (output_intermediate_file): Add version to intermediate
12142 * doc/gcov.texi: Document new field 'version' in intermediate
12143 file format. Fix location of '-k' option of gcov command.
12145 2018-01-04 Martin Liska <mliska@suse.cz>
12148 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12150 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12152 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12154 2018-01-03 Martin Sebor <msebor@redhat.com>
12156 PR tree-optimization/83655
12157 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12158 checking calls with invalid arguments.
12160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12162 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12163 (vectorizable_mask_load_store): Delete.
12164 (vectorizable_call): Return false for masked loads and stores.
12165 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12166 instead of gimple_assign_rhs1.
12167 (vectorizable_load): Handle IFN_MASK_LOAD.
12168 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12172 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12174 (vectorizable_mask_load_store): ...here.
12175 (vectorizable_load): ...and here.
12177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12179 * tree-vect-stmts.c (vect_build_all_ones_mask)
12180 (vect_build_zero_merge_argument): New functions, split out from...
12181 (vectorizable_load): ...here.
12183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12185 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12187 (vectorizable_mask_load_store): ...here.
12188 (vectorizable_store): ...and here.
12190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12192 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12194 (vectorizable_mask_load_store): ...here.
12196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12198 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12199 (vect_model_store_cost): Take a vec_load_store_type instead of a
12201 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12202 (vect_model_store_cost): Take a vec_load_store_type instead of a
12204 (vectorizable_mask_load_store): Update accordingly.
12205 (vectorizable_store): Likewise.
12206 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12210 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12211 IFN_MASK_LOAD calls here rather than...
12212 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12215 Alan Hayward <alan.hayward@arm.com>
12216 David Sherwood <david.sherwood@arm.com>
12218 * expmed.c (extract_bit_field_1): For vector extracts,
12219 fall back to extract_bit_field_as_subreg if vec_extract
12222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12223 Alan Hayward <alan.hayward@arm.com>
12224 David Sherwood <david.sherwood@arm.com>
12226 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12227 they are variable or constant sized.
12228 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12229 slots for constant-sized data.
12231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12232 Alan Hayward <alan.hayward@arm.com>
12233 David Sherwood <david.sherwood@arm.com>
12235 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12236 handling COND_EXPRs with boolean comparisons, try to find a better
12237 basis for the mask type than the boolean itself.
12239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12241 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12242 is calculated and how it can be overridden.
12243 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12244 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12246 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12250 Alan Hayward <alan.hayward@arm.com>
12251 David Sherwood <david.sherwood@arm.com>
12253 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12254 Remove the mode argument.
12255 (aarch64_simd_valid_immediate): Remove the mode and inverse
12257 * config/aarch64/iterators.md (bitsize): New iterator.
12258 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12259 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12260 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12261 aarch64_simd_valid_immediate.
12262 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12263 (aarch64_reg_or_bic_imm): Likewise.
12264 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12265 with an insn_type enum and msl with a modifier_type enum.
12266 Replace element_width with a scalar_mode. Change the shift
12267 to unsigned int. Add constructors for scalar_float_mode and
12268 scalar_int_mode elements.
12269 (aarch64_vect_float_const_representable_p): Delete.
12270 (aarch64_can_const_movi_rtx_p)
12271 (aarch64_simd_scalar_immediate_valid_for_move)
12272 (aarch64_simd_make_constant): Update call to
12273 aarch64_simd_valid_immediate.
12274 (aarch64_advsimd_valid_immediate_hs): New function.
12275 (aarch64_advsimd_valid_immediate): Likewise.
12276 (aarch64_simd_valid_immediate): Remove mode and inverse
12277 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12278 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12279 and aarch64_float_const_representable_p on the result.
12280 (aarch64_output_simd_mov_immediate): Remove mode argument.
12281 Update call to aarch64_simd_valid_immediate and use of
12282 simd_immediate_info.
12283 (aarch64_output_scalar_simd_mov_immediate): Update call
12286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12287 Alan Hayward <alan.hayward@arm.com>
12288 David Sherwood <david.sherwood@arm.com>
12290 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12291 (mode_nunits): Likewise CONST_MODE_NUNITS.
12292 * machmode.def (ADJUST_NUNITS): Document.
12293 * genmodes.c (mode_data::need_nunits_adj): New field.
12294 (blank_mode): Update accordingly.
12295 (adj_nunits): New variable.
12296 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12298 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12299 listed in adj_nunits.
12300 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12301 listed in adj_nunits. Don't emit case statements for such modes.
12302 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12303 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12304 nothing if adj_nunits is nonnull.
12305 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12306 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12307 (emit_mode_fbit): Update use of print_maybe_const_decl.
12308 (emit_move_size): Likewise. Treat the array as non-const
12310 (emit_mode_adjustments): Handle adj_nunits.
12312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12314 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12315 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12316 (VECTOR_MODES): Use it.
12317 (make_vector_modes): Take the prefix as an argument.
12319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12320 Alan Hayward <alan.hayward@arm.com>
12321 David Sherwood <david.sherwood@arm.com>
12323 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12324 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12325 for MODE_VECTOR_BOOL.
12326 * machmode.def (VECTOR_BOOL_MODE): Document.
12327 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12328 (make_vector_bool_mode): New function.
12329 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12331 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12332 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12334 * stor-layout.c (int_mode_for_mode): Likewise.
12335 * tree.c (build_vector_type_for_mode): Likewise.
12336 * varasm.c (output_constant_pool_2): Likewise.
12337 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12338 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12339 for MODE_VECTOR_BOOL.
12340 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12341 of mode class checks.
12342 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12343 instead of a list of mode class checks.
12344 (expand_vector_scalar_condition): Likewise.
12345 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12348 Alan Hayward <alan.hayward@arm.com>
12349 David Sherwood <david.sherwood@arm.com>
12351 * machmode.h (mode_size): Change from unsigned short to
12353 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12354 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12355 or if measurement_type is not polynomial.
12356 (fixed_size_mode::includes_p): Check for constant-sized modes.
12357 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12358 return a poly_uint16 rather than an unsigned short.
12359 (emit_mode_size): Change the type of mode_size from unsigned short
12360 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12361 (emit_mode_adjustments): Cope with polynomial vector sizes.
12362 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12364 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12366 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12367 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12368 * caller-save.c (setup_save_areas): Likewise.
12369 (replace_reg_with_saved_mem): Likewise.
12370 * calls.c (emit_library_call_value_1): Likewise.
12371 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12372 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12373 (gen_lowpart_for_combine): Likewise.
12374 * convert.c (convert_to_integer_1): Likewise.
12375 * cse.c (equiv_constant, cse_insn): Likewise.
12376 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12377 (cselib_subst_to_values): Likewise.
12378 * dce.c (word_dce_process_block): Likewise.
12379 * df-problems.c (df_word_lr_mark_ref): Likewise.
12380 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12381 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12382 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12383 (rtl_for_decl_location): Likewise.
12384 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12385 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12386 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12387 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12388 (expand_expr_real_1): Likewise.
12389 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12390 (pad_below): Likewise.
12391 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12392 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12393 * ira.c (get_subreg_tracking_sizes): Likewise.
12394 * ira-build.c (ira_create_allocno_objects): Likewise.
12395 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12396 (ira_sort_regnos_for_alter_reg): Likewise.
12397 * ira-costs.c (record_operand_costs): Likewise.
12398 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12399 (resolve_simple_move): Likewise.
12400 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12401 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12402 (lra_constraints): Likewise.
12403 (CONST_POOL_OK_P): Reject variable-sized modes.
12404 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12405 (add_pseudo_to_slot, lra_spill): Likewise.
12406 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12407 * optabs-query.c (get_best_extraction_insn): Likewise.
12408 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12409 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12410 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12411 * recog.c (offsettable_address_addr_space_p): Likewise.
12412 * regcprop.c (maybe_mode_change): Likewise.
12413 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12414 * regrename.c (build_def_use): Likewise.
12415 * regstat.c (dump_reg_info): Likewise.
12416 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12417 (find_reloads, find_reloads_subreg_address): Likewise.
12418 * reload1.c (eliminate_regs_1): Likewise.
12419 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12420 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12421 (simplify_binary_operation_1, simplify_subreg): Likewise.
12422 * targhooks.c (default_function_arg_padding): Likewise.
12423 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12424 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12425 (verify_gimple_assign_ternary): Likewise.
12426 * tree-inline.c (estimate_move_cost): Likewise.
12427 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12428 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12429 (get_address_cost_ainc): Likewise.
12430 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12431 (vect_supportable_dr_alignment): Likewise.
12432 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12433 (vectorizable_reduction): Likewise.
12434 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12435 (vectorizable_operation, vectorizable_load): Likewise.
12436 * tree.c (build_same_sized_truth_vector_type): Likewise.
12437 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12438 * var-tracking.c (emit_note_insn_var_location): Likewise.
12439 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12440 (ADDR_VEC_ALIGN): Likewise.
12442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12443 Alan Hayward <alan.hayward@arm.com>
12444 David Sherwood <david.sherwood@arm.com>
12446 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12448 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12449 or if measurement_type is polynomial.
12450 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12451 * combine.c (make_extraction): Likewise.
12452 * dse.c (find_shift_sequence): Likewise.
12453 * dwarf2out.c (mem_loc_descriptor): Likewise.
12454 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12455 (extract_bit_field, extract_low_bits): Likewise.
12456 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12457 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12458 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12459 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12460 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12461 * reload.c (find_reloads): Likewise.
12462 * reload1.c (alter_reg): Likewise.
12463 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12464 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12465 * tree-if-conv.c (predicate_mem_writes): Likewise.
12466 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12467 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12468 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12469 * valtrack.c (dead_debug_insert_temp): Likewise.
12470 * varasm.c (mergeable_constant_section): Likewise.
12471 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12474 Alan Hayward <alan.hayward@arm.com>
12475 David Sherwood <david.sherwood@arm.com>
12477 * expr.c (expand_assignment): Cope with polynomial mode sizes
12478 when assigning to a CONCAT.
12480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12481 Alan Hayward <alan.hayward@arm.com>
12482 David Sherwood <david.sherwood@arm.com>
12484 * machmode.h (mode_precision): Change from unsigned short to
12486 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12488 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12489 or if measurement_type is not polynomial.
12490 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12491 in which the mode is already known to be a scalar_int_mode.
12492 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12493 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12495 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12496 for GET_MODE_PRECISION.
12497 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12498 for GET_MODE_PRECISION.
12499 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12501 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12502 (expand_field_assignment, make_extraction): Likewise.
12503 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12504 (get_last_value): Likewise.
12505 * convert.c (convert_to_integer_1): Likewise.
12506 * cse.c (cse_insn): Likewise.
12507 * expr.c (expand_expr_real_1): Likewise.
12508 * lra-constraints.c (simplify_operand_subreg): Likewise.
12509 * optabs-query.c (can_atomic_load_p): Likewise.
12510 * optabs.c (expand_atomic_load): Likewise.
12511 (expand_atomic_store): Likewise.
12512 * ree.c (combine_reaching_defs): Likewise.
12513 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12514 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12515 * tree.h (type_has_mode_precision_p): Likewise.
12516 * ubsan.c (instrument_si_overflow): 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 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12523 polynomial numbers of units.
12524 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12525 (valid_vector_subparts_p): New function.
12526 (build_vector_type): Remove temporary shim and take the number
12527 of units as a poly_uint64 rather than an int.
12528 (build_opaque_vector_type): Take the number of units as a
12529 poly_uint64 rather than an int.
12530 * tree.c (build_vector_from_ctor): Handle polynomial
12531 TYPE_VECTOR_SUBPARTS.
12532 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12533 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12534 (build_vector_from_val): If the number of units is variable,
12535 use build_vec_duplicate_cst for constant operands and
12536 VEC_DUPLICATE_EXPR otherwise.
12537 (make_vector_type): Remove temporary is_constant ().
12538 (build_vector_type, build_opaque_vector_type): Take the number of
12539 units as a poly_uint64 rather than an int.
12540 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12542 * cfgexpand.c (expand_debug_expr): Likewise.
12543 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12544 (store_constructor, expand_expr_real_1): Likewise.
12545 (const_scalar_mask_from_tree): Likewise.
12546 * fold-const-call.c (fold_const_reduction): Likewise.
12547 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12548 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12549 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12550 (fold_relational_const): Likewise.
12551 (native_interpret_vector): Likewise. Change the size from an
12552 int to an unsigned int.
12553 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12554 TYPE_VECTOR_SUBPARTS.
12555 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12556 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12557 duplicating a non-constant operand into a variable-length vector.
12558 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12559 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12560 * ipa-icf.c (sem_variable::equals): Likewise.
12561 * match.pd: Likewise.
12562 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12563 * print-tree.c (print_node): Likewise.
12564 * stor-layout.c (layout_type): Likewise.
12565 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12566 * tree-cfg.c (verify_gimple_comparison): Likewise.
12567 (verify_gimple_assign_binary): Likewise.
12568 (verify_gimple_assign_ternary): Likewise.
12569 (verify_gimple_assign_single): Likewise.
12570 * tree-pretty-print.c (dump_generic_node): Likewise.
12571 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12572 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12573 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12574 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12575 (vect_shift_permute_load_chain): Likewise.
12576 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12577 (expand_vector_condition, optimize_vector_constructor): Likewise.
12578 (lower_vec_perm, get_compute_type): Likewise.
12579 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12580 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12581 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12582 (vect_recog_mask_conversion_pattern): Likewise.
12583 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12584 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12585 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12586 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12587 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12588 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12589 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12590 (supportable_widening_operation): Likewise.
12591 (supportable_narrowing_operation): Likewise.
12592 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12594 * varasm.c (output_constant): Likewise.
12596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12597 Alan Hayward <alan.hayward@arm.com>
12598 David Sherwood <david.sherwood@arm.com>
12600 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12601 so that both the length == 3 and length != 3 cases set up their
12602 own permute vectors. Add comments explaining why we know the
12603 number of elements is constant.
12604 (vect_permute_load_chain): Likewise.
12606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12607 Alan Hayward <alan.hayward@arm.com>
12608 David Sherwood <david.sherwood@arm.com>
12610 * machmode.h (mode_nunits): Change from unsigned char to
12612 (ONLY_FIXED_SIZE_MODES): New macro.
12613 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12614 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12615 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12617 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12618 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12619 or if measurement_type is not polynomial.
12620 * genmodes.c (ZERO_COEFFS): New macro.
12621 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12623 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12624 Use ZERO_COEFFS when emitting initializers.
12625 * data-streamer.h (bp_pack_poly_value): New function.
12626 (bp_unpack_poly_value): Likewise.
12627 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12628 for GET_MODE_NUNITS.
12629 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12630 for GET_MODE_NUNITS.
12631 * tree.c (make_vector_type): Remove temporary shim and make
12632 the real function take the number of units as a poly_uint64
12633 rather than an int.
12634 (build_vector_type_for_mode): Handle polynomial nunits.
12635 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12636 * emit-rtl.c (const_vec_series_p_1): Likewise.
12637 (gen_rtx_CONST_VECTOR): Likewise.
12638 * fold-const.c (test_vec_duplicate_folding): Likewise.
12639 * genrecog.c (validate_pattern): Likewise.
12640 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12641 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12642 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12643 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12644 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12645 * rtlanal.c (subreg_get_info): Likewise.
12646 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12647 (vect_grouped_load_supported): Likewise.
12648 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12649 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12650 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12651 (simplify_const_unary_operation, simplify_binary_operation_1)
12652 (simplify_const_binary_operation, simplify_ternary_operation)
12653 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12654 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12655 instead of CONST_VECTOR_NUNITS.
12656 * varasm.c (output_constant_pool_2): Likewise.
12657 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12658 explicit-encoded elements in the XVEC for variable-length vectors.
12660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12662 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12665 Alan Hayward <alan.hayward@arm.com>
12666 David Sherwood <david.sherwood@arm.com>
12668 * coretypes.h (fixed_size_mode): Declare.
12669 (fixed_size_mode_pod): New typedef.
12670 * builtins.h (target_builtins::x_apply_args_mode)
12671 (target_builtins::x_apply_result_mode): Change type to
12672 fixed_size_mode_pod.
12673 * builtins.c (apply_args_size, apply_result_size, result_vector)
12674 (expand_builtin_apply_args_1, expand_builtin_apply)
12675 (expand_builtin_return): Update accordingly.
12677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12679 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12680 * cselib.c (cselib_hash_rtx): Likewise.
12681 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12682 CONST_VECTOR encoding.
12684 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12685 Jeff Law <law@redhat.com>
12688 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12689 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12690 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12691 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12694 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12695 explicitly probe *sp in a noreturn function if there were any callee
12696 register saves or frame pointer is needed.
12698 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12701 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12702 BLKmode for ternary, binary or unary expressions.
12705 * var-tracking.c (delete_vta_debug_insn): New inline function.
12706 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12707 insns from get_insns () to NULL instead of each bb separately.
12708 Use delete_vta_debug_insn. No longer static.
12709 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12710 delete_vta_debug_insns callers.
12711 * rtl.h (delete_vta_debug_insns): Declare.
12712 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12713 instead of variable_tracking_main.
12715 2018-01-03 Martin Sebor <msebor@redhat.com>
12717 PR tree-optimization/83603
12718 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12719 arguments past the endof the argument list in functions declared
12720 without a prototype.
12721 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12722 Avoid checking when arguments are null.
12724 2018-01-03 Martin Sebor <msebor@redhat.com>
12727 * doc/extend.texi (attribute const): Fix a typo.
12728 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12729 issuing -Wsuggest-attribute for void functions.
12731 2018-01-03 Martin Sebor <msebor@redhat.com>
12733 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12734 offset_int::from instead of wide_int::to_shwi.
12735 (maybe_diag_overlap): Remove assertion.
12736 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12737 * gimple-ssa-sprintf.c (format_directive): Same.
12738 (parse_directive): Same.
12739 (sprintf_dom_walker::compute_format_length): Same.
12740 (try_substitute_return_value): Same.
12742 2018-01-03 Jeff Law <law@redhat.com>
12744 PR middle-end/83654
12745 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12746 non-constant residual for zero at runtime and avoid probing in
12747 that case. Reorganize code for trailing problem to mirror handling
12750 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12752 PR tree-optimization/83501
12753 * tree-ssa-strlen.c (get_string_cst): New.
12754 (handle_char_store): Call get_string_cst.
12756 2018-01-03 Martin Liska <mliska@suse.cz>
12758 PR tree-optimization/83593
12759 * tree-ssa-strlen.c: Include tree-cfg.h.
12760 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12761 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12762 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12764 (strlen_dom_walker::before_dom_children): Call
12765 gimple_purge_dead_eh_edges. Dump tranformation with details
12767 (strlen_dom_walker::before_dom_children): Update call by adding
12768 new argument cleanup_eh.
12769 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12771 2018-01-03 Martin Liska <mliska@suse.cz>
12774 * cif-code.def (VARIADIC_THUNK): New enum value.
12775 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12778 2018-01-03 Jan Beulich <jbeulich@suse.com>
12780 * sse.md (mov<mode>_internal): Tighten condition for when to use
12781 vmovdqu<ssescalarsize> for TI and OI modes.
12783 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12785 Update copyright years.
12787 2018-01-03 Martin Liska <mliska@suse.cz>
12790 * ipa-visibility.c (function_and_variable_visibility): Skip
12791 functions with noipa attribure.
12793 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12795 * gcc.c (process_command): Update copyright notice dates.
12796 * gcov-dump.c (print_version): Ditto.
12797 * gcov.c (print_version): Ditto.
12798 * gcov-tool.c (print_version): Ditto.
12799 * gengtype.c (create_file): Ditto.
12800 * doc/cpp.texi: Bump @copying's copyright year.
12801 * doc/cppinternals.texi: Ditto.
12802 * doc/gcc.texi: Ditto.
12803 * doc/gccint.texi: Ditto.
12804 * doc/gcov.texi: Ditto.
12805 * doc/install.texi: Ditto.
12806 * doc/invoke.texi: Ditto.
12808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12810 * vector-builder.h (vector_builder::m_full_nelts): Change from
12811 unsigned int to poly_uint64.
12812 (vector_builder::full_nelts): Update prototype accordingly.
12813 (vector_builder::new_vector): Likewise.
12814 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12815 (vector_builder::operator ==): Likewise.
12816 (vector_builder::finalize): Likewise.
12817 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12818 Take the number of elements as a poly_uint64 rather than an
12820 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12821 from unsigned int to poly_uint64.
12822 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12823 (vec_perm_indices::new_vector): Likewise.
12824 (vec_perm_indices::length): Likewise.
12825 (vec_perm_indices::nelts_per_input): Likewise.
12826 (vec_perm_indices::input_nelts): Likewise.
12827 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12828 number of elements per input as a poly_uint64 rather than an
12829 unsigned int. Use the original encoding for variable-length
12830 vectors, rather than clamping each individual element.
12831 For the second and subsequent elements in each pattern,
12832 clamp the step and base before clamping their sum.
12833 (vec_perm_indices::series_p): Handle polynomial element counts.
12834 (vec_perm_indices::all_in_range_p): Likewise.
12835 (vec_perm_indices_to_tree): Likewise.
12836 (vec_perm_indices_to_rtx): Likewise.
12837 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12838 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12839 (tree_vector_builder::new_binary_operation): Handle polynomial
12840 element counts. Return false if we need to know the number
12841 of elements at compile time.
12842 * fold-const.c (fold_vec_perm): Punt if the number of elements
12843 isn't known at compile time.
12845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12847 * vec-perm-indices.h (vec_perm_builder): Change element type
12848 from HOST_WIDE_INT to poly_int64.
12849 (vec_perm_indices::element_type): Update accordingly.
12850 (vec_perm_indices::clamp): Handle polynomial element_types.
12851 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12852 (vec_perm_indices::all_in_range_p): Likewise.
12853 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12855 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12856 polynomial vec_perm_indices element types.
12857 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12858 * fold-const.c (fold_vec_perm): Likewise.
12859 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12860 * tree-vect-generic.c (lower_vec_perm): Likewise.
12861 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12862 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12863 element type to HOST_WIDE_INT.
12865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12866 Alan Hayward <alan.hayward@arm.com>
12867 David Sherwood <david.sherwood@arm.com>
12869 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12870 rather than an int. Use plus_constant.
12871 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12872 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12875 Alan Hayward <alan.hayward@arm.com>
12876 David Sherwood <david.sherwood@arm.com>
12878 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12879 a HOST_WIDE_INT to a poly_int64.
12881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12882 Alan Hayward <alan.hayward@arm.com>
12883 David Sherwood <david.sherwood@arm.com>
12885 * calls.c (load_register_parameters): Cope with polynomial
12886 mode sizes. Require a constant size for BLKmode parameters
12887 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12888 forces a parameter to be padded at the lsb end in order to
12889 fill a complete number of words, require the parameter size
12890 to be ordered wrt UNITS_PER_WORD.
12892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12893 Alan Hayward <alan.hayward@arm.com>
12894 David Sherwood <david.sherwood@arm.com>
12896 * reload1.c (spill_stack_slot_width): Change element type
12897 from unsigned int to poly_uint64_pod.
12898 (alter_reg): Treat mode sizes as polynomial.
12900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12901 Alan Hayward <alan.hayward@arm.com>
12902 David Sherwood <david.sherwood@arm.com>
12904 * reload.c (complex_word_subreg_p): New function.
12905 (reload_inner_reg_of_subreg, push_reload): Use it.
12907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12908 Alan Hayward <alan.hayward@arm.com>
12909 David Sherwood <david.sherwood@arm.com>
12911 * lra-constraints.c (process_alt_operands): Reject matched
12912 operands whose sizes aren't ordered.
12913 (match_reload): Refer to this check here.
12915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12916 Alan Hayward <alan.hayward@arm.com>
12917 David Sherwood <david.sherwood@arm.com>
12919 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12920 that the mode size is in the set {1, 2, 4, 8, 16}.
12922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12923 Alan Hayward <alan.hayward@arm.com>
12924 David Sherwood <david.sherwood@arm.com>
12926 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12927 Use plus_constant instead of gen_rtx_PLUS.
12929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12930 Alan Hayward <alan.hayward@arm.com>
12931 David Sherwood <david.sherwood@arm.com>
12933 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12934 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12935 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12936 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12937 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12938 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12939 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12940 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12941 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12942 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12944 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12945 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12946 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12947 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12948 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12949 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12950 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12951 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12952 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12953 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12955 * expr.c (emit_move_resolve_push): Treat the input and result
12956 of PUSH_ROUNDING as a poly_int64.
12957 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12958 (emit_push_insn): Likewise.
12959 * lra-eliminations.c (mark_not_eliminable): Likewise.
12960 * recog.c (push_operand): Likewise.
12961 * reload1.c (elimination_effects): Likewise.
12962 * rtlanal.c (nonzero_bits1): Likewise.
12963 * calls.c (store_one_arg): Likewise. Require the padding to be
12964 known at compile time.
12966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12967 Alan Hayward <alan.hayward@arm.com>
12968 David Sherwood <david.sherwood@arm.com>
12970 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12971 Use plus_constant instead of gen_rtx_PLUS.
12973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12974 Alan Hayward <alan.hayward@arm.com>
12975 David Sherwood <david.sherwood@arm.com>
12977 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12978 rather than an int.
12980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12981 Alan Hayward <alan.hayward@arm.com>
12982 David Sherwood <david.sherwood@arm.com>
12984 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12985 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12986 via stack temporaries. Treat the mode size as polynomial too.
12988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12989 Alan Hayward <alan.hayward@arm.com>
12990 David Sherwood <david.sherwood@arm.com>
12992 * expr.c (expand_expr_real_2): When handling conversions involving
12993 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12994 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12995 as a poly_uint64 too.
12997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12998 Alan Hayward <alan.hayward@arm.com>
12999 David Sherwood <david.sherwood@arm.com>
13001 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13004 Alan Hayward <alan.hayward@arm.com>
13005 David Sherwood <david.sherwood@arm.com>
13007 * combine.c (can_change_dest_mode): Handle polynomial
13008 REGMODE_NATURAL_SIZE.
13009 * expmed.c (store_bit_field_1): Likewise.
13010 * expr.c (store_constructor): Likewise.
13011 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13012 and polynomial REGMODE_NATURAL_SIZE.
13013 (gen_lowpart_common): Likewise.
13014 * reginfo.c (record_subregs_of_mode): Likewise.
13015 * rtlanal.c (read_modify_subreg_p): Likewise.
13017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13018 Alan Hayward <alan.hayward@arm.com>
13019 David Sherwood <david.sherwood@arm.com>
13021 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13022 numbers of elements.
13024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13025 Alan Hayward <alan.hayward@arm.com>
13026 David Sherwood <david.sherwood@arm.com>
13028 * match.pd: Cope with polynomial numbers of vector elements.
13030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13031 Alan Hayward <alan.hayward@arm.com>
13032 David Sherwood <david.sherwood@arm.com>
13034 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13035 in a POINTER_PLUS_EXPR.
13037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13038 Alan Hayward <alan.hayward@arm.com>
13039 David Sherwood <david.sherwood@arm.com>
13041 * omp-simd-clone.c (simd_clone_subparts): New function.
13042 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13043 (ipa_simd_modify_function_body): Likewise.
13045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13046 Alan Hayward <alan.hayward@arm.com>
13047 David Sherwood <david.sherwood@arm.com>
13049 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13050 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13051 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13052 (expand_vector_condition, vector_element): Likewise.
13053 (subparts_gt): New function.
13054 (get_compute_type): Use subparts_gt.
13055 (count_type_subparts): Delete.
13056 (expand_vector_operations_1): Use subparts_gt instead of
13057 count_type_subparts.
13059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13060 Alan Hayward <alan.hayward@arm.com>
13061 David Sherwood <david.sherwood@arm.com>
13063 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13064 (vect_compile_time_alias): ...this new function. Do the calculation
13065 on poly_ints rather than trees.
13066 (vect_prune_runtime_alias_test_list): Update call accordingly.
13068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13069 Alan Hayward <alan.hayward@arm.com>
13070 David Sherwood <david.sherwood@arm.com>
13072 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13074 (vect_schedule_slp_instance): Likewise.
13076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13077 Alan Hayward <alan.hayward@arm.com>
13078 David Sherwood <david.sherwood@arm.com>
13080 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13081 constant and extern definitions for variable-length vectors.
13082 (vect_get_constant_vectors): Note that the number of units
13083 is known to be constant.
13085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13086 Alan Hayward <alan.hayward@arm.com>
13087 David Sherwood <david.sherwood@arm.com>
13089 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13090 of units as polynomial. Choose between WIDE and NARROW based
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 * tree-vect-stmts.c (simd_clone_subparts): New function.
13098 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13105 vectors as polynomial. Use build_index_vector for
13106 IFN_GOMP_SIMD_LANE.
13108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13109 Alan Hayward <alan.hayward@arm.com>
13110 David Sherwood <david.sherwood@arm.com>
13112 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13113 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13114 for variable-length vectors.
13115 (vectorizable_mask_load_store): Treat the number of units as
13116 polynomial, asserting that it is constant if the condition has
13117 already been enforced.
13118 (vectorizable_store, vectorizable_load): Likewise.
13120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13121 Alan Hayward <alan.hayward@arm.com>
13122 David Sherwood <david.sherwood@arm.com>
13124 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13125 of units as polynomial. Punt if we can't tell at compile time
13126 which vector contains the final result.
13128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13129 Alan Hayward <alan.hayward@arm.com>
13130 David Sherwood <david.sherwood@arm.com>
13132 * tree-vect-loop.c (vectorizable_induction): Treat the number
13133 of units as polynomial. Punt on SLP inductions. Use an integer
13134 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13135 cast of such a series for variable-length floating-point
13138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13139 Alan Hayward <alan.hayward@arm.com>
13140 David Sherwood <david.sherwood@arm.com>
13142 * tree.h (build_index_vector): Declare.
13143 * tree.c (build_index_vector): New function.
13144 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13145 of units as polynomial, forcibly converting it to a constant if
13146 vectorizable_reduction has already enforced the condition.
13147 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13148 to create a {1,2,3,...} vector.
13149 (vectorizable_reduction): Treat the number of units as polynomial.
13150 Choose vectype_in based on the largest scalar element size rather
13151 than the smallest number of units. Enforce the restrictions
13154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13155 Alan Hayward <alan.hayward@arm.com>
13156 David Sherwood <david.sherwood@arm.com>
13158 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13159 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13166 * target.def (autovectorize_vector_sizes): Return the vector sizes
13167 by pointer, using vector_sizes rather than a bitmask.
13168 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13169 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13170 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13172 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13173 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13174 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13175 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13176 * omp-general.c (omp_max_vf): Likewise.
13177 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13178 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13179 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13180 * tree-vect-slp.c (vect_slp_bb): Likewise.
13181 * doc/tm.texi: Regenerate.
13182 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13184 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13185 the vector size as a poly_uint64 rather than an unsigned int.
13186 (current_vector_size): Change from an unsigned int to a poly_uint64.
13187 (get_vectype_for_scalar_type): Update accordingly.
13188 * tree.h (build_truth_vector_type): Take the size and number of
13189 units as a poly_uint64 rather than an unsigned int.
13190 (build_vector_type): Add a temporary overload that takes
13191 the number of units as a poly_uint64 rather than an unsigned int.
13192 * tree.c (make_vector_type): Likewise.
13193 (build_truth_vector_type): Take the number of units as a poly_uint64
13194 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
13201 as poly_uint64s rather than unsigned ints.
13202 * targhooks.h (default_get_mask_mode): Update accordingly.
13203 * targhooks.c (default_get_mask_mode): Likewise.
13204 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13205 * doc/tm.texi: Regenerate.
13207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13208 Alan Hayward <alan.hayward@arm.com>
13209 David Sherwood <david.sherwood@arm.com>
13211 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13212 * omp-general.c (omp_max_vf): Likewise.
13213 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13214 (expand_omp_simd): Handle polynomial safelen.
13215 * omp-low.c (omplow_simd_context): Add a default constructor.
13216 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13217 (lower_rec_simd_input_clauses): Update accordingly.
13218 (lower_rec_input_clauses): Likewise.
13220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13221 Alan Hayward <alan.hayward@arm.com>
13222 David Sherwood <david.sherwood@arm.com>
13224 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13225 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13226 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13227 (vect_analyze_slp_cost): Likewise.
13228 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13229 (vect_model_load_cost): Likewise.
13231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13232 Alan Hayward <alan.hayward@arm.com>
13233 David Sherwood <david.sherwood@arm.com>
13235 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13236 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13237 from an unsigned int * to a poly_uint64_pod *.
13238 (calculate_unrolling_factor): New function.
13239 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13242 Alan Hayward <alan.hayward@arm.com>
13243 David Sherwood <david.sherwood@arm.com>
13245 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13246 from an unsigned int to a poly_uint64.
13247 (_loop_vec_info::slp_unrolling_factor): Likewise.
13248 (_loop_vec_info::vectorization_factor): Change from an int
13250 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13251 (vect_get_num_vectors): New function.
13252 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13253 (vect_get_num_copies): Use vect_get_num_vectors.
13254 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13255 to an unsigned int *.
13256 (vect_analyze_data_refs): Change min_vf from an int * to a
13258 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13259 than an unsigned HOST_WIDE_INT.
13260 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13261 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13262 to an unsigned int *.
13263 (vect_analyze_data_ref_dependences): Likewise.
13264 (vect_compute_data_ref_alignment): Handle polynomial vf.
13265 (vect_enhance_data_refs_alignment): Likewise.
13266 (vect_prune_runtime_alias_test_list): Likewise.
13267 (vect_shift_permute_load_chain): Likewise.
13268 (vect_supportable_dr_alignment): Likewise.
13269 (dependence_distance_ge_vf): Take the vectorization factor as a
13270 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13271 (vect_analyze_data_refs): Change min_vf from an int * to a
13273 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13274 vfm1 as a poly_uint64 rather than an int. Make the same change
13275 for the returned bound_scalar.
13276 (vect_gen_vector_loop_niters): Handle polynomial vf.
13277 (vect_do_peeling): Likewise. Update call to
13278 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13279 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13281 * tree-vect-loop.c (vect_determine_vectorization_factor)
13282 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13283 (vect_get_known_peeling_cost): Likewise.
13284 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13285 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13286 (vect_transform_loop): Likewise. Use the lowest possible VF when
13287 updating the upper bounds of the loop.
13288 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13289 rather than an int.
13290 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13291 polynomial unroll factors.
13292 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13293 (vect_make_slp_decision): Likewise.
13294 (vect_supported_load_permutation_p): Likewise, and polynomial
13296 (vect_analyze_slp_cost): Handle polynomial vf.
13297 (vect_slp_analyze_node_operations): Likewise.
13298 (vect_slp_analyze_bb_1): Likewise.
13299 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13300 than an unsigned HOST_WIDE_INT.
13301 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13302 (vectorizable_load): Handle polynomial vf.
13303 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13305 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
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 * match.pd: Handle bit operations involving three constants
13312 and try to fold one pair.
13314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13316 * tree-vect-loop-manip.c: Include gimple-fold.h.
13317 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13318 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13319 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13320 Add a path that uses a step of VF instead of 1, but disable it
13322 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13323 and niters_no_overflow parameters. Update calls to
13324 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13325 Create a new SSA name if the latter choses to use a ste other
13326 than zero, and return it via niters_vector_mult_vf_var.
13327 * tree-vect-loop.c (vect_transform_loop): Update calls to
13328 vect_do_peeling, vect_gen_vector_loop_niters and
13329 slpeel_make_loop_iterate_ntimes.
13330 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13331 (vect_gen_vector_loop_niters): Update declarations after above changes.
13333 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13335 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13336 128-bit round to integer instructions.
13337 (ceil<mode>2): Likewise.
13338 (btrunc<mode>2): Likewise.
13339 (round<mode>2): Likewise.
13341 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13343 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13344 unaligned VSX load/store on P8/P9.
13345 (expand_block_clear): Allow the use of unaligned VSX
13346 load/store on P8/P9.
13348 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13350 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13352 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13353 swap associated with both a load and a store.
13355 2018-01-02 Andrew Waterman <andrew@sifive.com>
13357 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13358 * config/riscv/riscv.md (clear_cache): Use it.
13360 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13362 * web.c: Remove out-of-date comment.
13364 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13366 * expr.c (fixup_args_size_notes): Check that any existing
13367 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13368 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13369 (emit_single_push_insn): ...here.
13371 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13373 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13374 (const_vector_encoded_nelts): New function.
13375 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13376 (const_vector_int_elt, const_vector_elt): Declare.
13377 * emit-rtl.c (const_vector_int_elt_1): New function.
13378 (const_vector_elt): Likewise.
13379 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13380 of CONST_VECTOR_ELT.
13382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13384 * expr.c: Include rtx-vector-builder.h.
13385 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13386 directly on the tree encoding.
13387 (const_vector_from_tree): Likewise.
13388 * optabs.c: Include rtx-vector-builder.h.
13389 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13390 sequence of "u" values.
13391 * vec-perm-indices.c: Include rtx-vector-builder.h.
13392 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13393 directly on the vec_perm_indices encoding.
13395 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13397 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13398 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13399 * rtx-vector-builder.h: New file.
13400 * rtx-vector-builder.c: Likewise.
13401 * rtl.h (rtx_def::u2): Add a const_vector field.
13402 (CONST_VECTOR_NPATTERNS): New macro.
13403 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13404 (CONST_VECTOR_DUPLICATE_P): Likewise.
13405 (CONST_VECTOR_STEPPED_P): Likewise.
13406 (CONST_VECTOR_ENCODED_ELT): Likewise.
13407 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13408 (unwrap_const_vec_duplicate): Likewise.
13409 (const_vec_series_p): Check for a non-duplicated vector encoding.
13410 Say that the function only returns true for integer vectors.
13411 * emit-rtl.c: Include rtx-vector-builder.h.
13412 (gen_const_vec_duplicate_1): Delete.
13413 (gen_const_vector): Call gen_const_vec_duplicate instead of
13414 gen_const_vec_duplicate_1.
13415 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13416 (gen_const_vec_duplicate): Use rtx_vector_builder.
13417 (gen_const_vec_series): Likewise.
13418 (gen_rtx_CONST_VECTOR): Likewise.
13419 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13420 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13421 Build a new vector rather than modifying a CONST_VECTOR in-place.
13422 (handle_special_swappables): Update call accordingly.
13423 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13424 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13425 Build a new vector rather than modifying a CONST_VECTOR in-place.
13426 (handle_special_swappables): Update call accordingly.
13428 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13430 * simplify-rtx.c (simplify_const_binary_operation): Use
13431 CONST_VECTOR_ELT instead of XVECEXP.
13433 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13435 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13436 the selector elements to be different from the data elements
13437 if the selector is a VECTOR_CST.
13438 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13439 ssizetype for the selector.
13441 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13443 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13444 before testing each element individually.
13445 * tree-vect-generic.c (lower_vec_perm): Likewise.
13447 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13449 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13450 * selftest-run-tests.c (selftest::run_tests): Call it.
13451 * vector-builder.h (vector_builder::operator ==): New function.
13452 (vector_builder::operator !=): Likewise.
13453 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13454 (vec_perm_indices::all_from_input_p): New function.
13455 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13456 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13457 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13458 instead of reading the VECTOR_CST directly. Detect whether both
13459 vector inputs are the same before constructing the vec_perm_indices,
13460 and update the number of inputs argument accordingly. Use the
13461 utility functions added above. Only construct sel2 if we need to.
13463 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13465 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13466 the broadcast of the low byte.
13467 (expand_mult_highpart): Use an explicit encoding for the permutes.
13468 * optabs-query.c (can_mult_highpart_p): Likewise.
13469 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13470 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13471 (vectorizable_bswap): Likewise.
13472 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13473 explicit encoding for the power-of-2 permutes.
13474 (vect_permute_store_chain): Likewise.
13475 (vect_grouped_load_supported): Likewise.
13476 (vect_permute_load_chain): Likewise.
13478 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13480 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13481 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13482 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13483 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13484 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13485 (vect_gen_perm_mask_any): Likewise.
13487 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13489 * int-vector-builder.h: New file.
13490 * vec-perm-indices.h: Include int-vector-builder.h.
13491 (vec_perm_indices): Redefine as an int_vector_builder.
13492 (auto_vec_perm_indices): Delete.
13493 (vec_perm_builder): Redefine as a stand-alone class.
13494 (vec_perm_indices::vec_perm_indices): New function.
13495 (vec_perm_indices::clamp): Likewise.
13496 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13497 (vec_perm_indices::new_vector): New function.
13498 (vec_perm_indices::new_expanded_vector): Update for new
13499 vec_perm_indices class.
13500 (vec_perm_indices::rotate_inputs): New function.
13501 (vec_perm_indices::all_in_range_p): Operate directly on the
13502 encoded form, without computing elided elements.
13503 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13504 encoding. Update for new vec_perm_indices class.
13505 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13506 the given vec_perm_builder.
13507 (expand_vec_perm_var): Update vec_perm_builder constructor.
13508 (expand_mult_highpart): Use vec_perm_builder instead of
13509 auto_vec_perm_indices.
13510 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13511 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13512 or double series encoding as appropriate.
13513 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13514 vec_perm_indices instead of auto_vec_perm_indices.
13515 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13516 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13517 (vect_permute_store_chain): Likewise.
13518 (vect_grouped_load_supported): Likewise.
13519 (vect_permute_load_chain): Likewise.
13520 (vect_shift_permute_load_chain): Likewise.
13521 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13522 (vect_transform_slp_perm_load): Likewise.
13523 (vect_schedule_slp_instance): Likewise.
13524 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13525 (vectorizable_mask_load_store): Likewise.
13526 (vectorizable_bswap): Likewise.
13527 (vectorizable_store): Likewise.
13528 (vectorizable_load): Likewise.
13529 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13530 vec_perm_indices instead of auto_vec_perm_indices. Use
13531 tree_to_vec_perm_builder to read the vector from a tree.
13532 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13533 vec_perm_builder instead of a vec_perm_indices.
13534 (have_whole_vector_shift): Use vec_perm_builder and
13535 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13536 truncation to calc_vec_perm_mask_for_shift.
13537 (vect_create_epilog_for_reduction): Likewise.
13538 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13539 from auto_vec_perm_indices to vec_perm_indices.
13540 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13541 instead of changing individual elements.
13542 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13543 the vector in d.perm.
13544 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13545 from auto_vec_perm_indices to vec_perm_indices.
13546 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13547 instead of changing individual elements.
13548 (arm_vectorize_vec_perm_const): Use new_vector to install
13549 the vector in d.perm.
13550 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13551 Update vec_perm_builder constructor.
13552 (rs6000_expand_interleave): Likewise.
13553 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13554 (rs6000_expand_interleave): Likewise.
13556 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13558 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13559 to qimode could truncate the indices.
13560 * optabs.c (expand_vec_perm_var): Likewise.
13562 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13564 * Makefile.in (OBJS): Add vec-perm-indices.o.
13565 * vec-perm-indices.h: New file.
13566 * vec-perm-indices.c: Likewise.
13567 * target.h (vec_perm_indices): Replace with a forward class
13569 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13570 * optabs.h: Include vec-perm-indices.h.
13571 (expand_vec_perm): Delete.
13572 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13573 (expand_vec_perm_const): Declare.
13574 * target.def (vec_perm_const_ok): Replace with...
13575 (vec_perm_const): ...this new hook.
13576 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13577 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13578 * doc/tm.texi: Regenerate.
13579 * optabs.def (vec_perm_const): Delete.
13580 * doc/md.texi (vec_perm_const): Likewise.
13581 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13582 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13583 expand_vec_perm for constant permutation vectors. Assert that
13584 the mode of variable permutation vectors is the integer equivalent
13585 of the mode that is being permuted.
13586 * optabs-query.h (selector_fits_mode_p): Declare.
13587 * optabs-query.c: Include vec-perm-indices.h.
13588 (selector_fits_mode_p): New function.
13589 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13590 is defined, instead of checking whether the vec_perm_const_optab
13591 exists. Use targetm.vectorize.vec_perm_const instead of
13592 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13593 fit in the vector mode before using a variable permute.
13594 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13595 vec_perm_indices instead of an rtx.
13596 (expand_vec_perm): Replace with...
13597 (expand_vec_perm_const): ...this new function. Take the selector
13598 as a vec_perm_indices rather than an rtx. Also take the mode of
13599 the selector. Update call to shift_amt_for_vec_perm_mask.
13600 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13601 Use vec_perm_indices::new_expanded_vector to expand the original
13602 selector into bytes. Check whether the indices fit in the vector
13603 mode before using a variable permute.
13604 (expand_vec_perm_var): Make global.
13605 (expand_mult_highpart): Use expand_vec_perm_const.
13606 * fold-const.c: Includes vec-perm-indices.h.
13607 * tree-ssa-forwprop.c: Likewise.
13608 * tree-vect-data-refs.c: Likewise.
13609 * tree-vect-generic.c: Likewise.
13610 * tree-vect-loop.c: Likewise.
13611 * tree-vect-slp.c: Likewise.
13612 * tree-vect-stmts.c: Likewise.
13613 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13615 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13616 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13617 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13618 (aarch64_vectorize_vec_perm_const): ...this new function.
13619 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13620 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13621 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13622 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13623 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13624 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13625 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13627 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13628 check for NEON modes.
13629 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13630 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13631 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13632 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13634 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13635 the old VEC_PERM_CONST conditions.
13636 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13637 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13638 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13639 (ia64_vectorize_vec_perm_const_ok): Merge into...
13640 (ia64_vectorize_vec_perm_const): ...this new function.
13641 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13642 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13643 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13644 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13645 * config/mips/mips.c (mips_expand_vec_perm_const)
13646 (mips_vectorize_vec_perm_const_ok): Merge into...
13647 (mips_vectorize_vec_perm_const): ...this new function.
13648 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13649 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13650 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13651 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13652 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13653 (rs6000_expand_vec_perm_const): Delete.
13654 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13656 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13657 (altivec_expand_vec_perm_const_le): Take each operand individually.
13658 Operate on constant selectors rather than rtxes.
13659 (altivec_expand_vec_perm_const): Likewise. Update call to
13660 altivec_expand_vec_perm_const_le.
13661 (rs6000_expand_vec_perm_const): Delete.
13662 (rs6000_vectorize_vec_perm_const_ok): Delete.
13663 (rs6000_vectorize_vec_perm_const): New function.
13664 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13665 an element count and rtx array.
13666 (rs6000_expand_extract_even): Update call accordingly.
13667 (rs6000_expand_interleave): Likewise.
13668 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13669 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13670 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13671 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13672 (rs6000_expand_vec_perm_const): Delete.
13673 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13674 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13675 (altivec_expand_vec_perm_const_le): Take each operand individually.
13676 Operate on constant selectors rather than rtxes.
13677 (altivec_expand_vec_perm_const): Likewise. Update call to
13678 altivec_expand_vec_perm_const_le.
13679 (rs6000_expand_vec_perm_const): Delete.
13680 (rs6000_vectorize_vec_perm_const_ok): Delete.
13681 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13682 reference to the SPE evmerge intructions.
13683 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13684 an element count and rtx array.
13685 (rs6000_expand_extract_even): Update call accordingly.
13686 (rs6000_expand_interleave): Likewise.
13687 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13688 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13690 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13692 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13694 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13695 vector mode and that that mode matches the mode of the data
13697 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13698 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13699 directly using expand_vec_perm_1 when forcing selectors into
13701 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13703 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13705 * optabs-query.h (can_vec_perm_p): Delete.
13706 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13707 * optabs-query.c (can_vec_perm_p): Split into...
13708 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13709 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13710 particular selector is valid.
13711 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13712 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13713 (vect_grouped_load_supported): Likewise.
13714 (vect_shift_permute_load_chain): Likewise.
13715 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13716 (vect_transform_slp_perm_load): Likewise.
13717 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13718 (vectorizable_bswap): Likewise.
13719 (vect_gen_perm_mask_checked): Likewise.
13720 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13721 implementations of variable permutation vectors into account
13722 when deciding which selector to use.
13723 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13724 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13725 with a false third argument.
13726 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13727 to test whether the constant selector is valid and can_vec_perm_var_p
13728 to test whether a variable selector is valid.
13730 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13732 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13733 * optabs-query.c (can_vec_perm_p): Likewise.
13734 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13735 instead of vec_perm_indices.
13736 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13737 (vect_gen_perm_mask_checked): Likewise,
13738 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13739 (vect_gen_perm_mask_checked): Likewise,
13741 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13743 * optabs-query.h (qimode_for_vec_perm): Declare.
13744 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13745 (qimode_for_vec_perm): ...this new function.
13746 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13748 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13750 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13751 does not have a conditional at the top.
13753 2018-01-02 Richard Biener <rguenther@suse.de>
13755 * ipa-inline.c (big_speedup_p): Fix expression.
13757 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13760 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13763 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13767 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13768 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13769 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13770 cond_taken_branch_cost 3->4.
13772 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13774 PR tree-optimization/83581
13775 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13776 TODO_cleanup_cfg if any changes have been made.
13778 PR middle-end/83608
13779 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13780 convert_modes if target mode has the right side, but different mode
13783 PR middle-end/83609
13784 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13785 last argument when extracting from CONCAT. If either from_real or
13786 from_imag is NULL, use expansion through memory. If result is not
13787 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13788 the parts directly to inner mode, if even that fails, use expansion
13791 PR middle-end/83623
13792 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13793 check for bswap in mode rather than HImode and use that in expand_unop
13796 Copyright (C) 2018 Free Software Foundation, Inc.
13798 Copying and distribution of this file, with or without modification,
13799 are permitted in any medium without royalty provided the copyright
13800 notice and this notice are preserved.