1 2018-05-22 Martin Sebor <msebor@redhat.com>
3 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8 * config/aarch64/aarch64-ldpstp.md: Replace uses of
9 aarch64_mem_pair_operand with memory_operand and delete operand swapping
11 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12 Add check for legitimate_address.
13 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
14 (aarch64_swap_ldrstr_operands): New.
15 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21 * config/aarch64/aarch64.md: New patterns to generate stp
23 (store_pair_sw, store_pair_dw): New patterns to generate stp for
24 single words and double words.
25 (load_pair_sw, load_pair_dw): Likewise.
26 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
28 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
30 * config/aarch64/aarch64-ldpstp.md: Modify peephole
31 for different mode ldpstp and add peephole for merged zero stores.
33 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
35 (aarch64_gen_store_pair): Rename calls to match new patterns.
36 (aarch64_gen_load_pair): Rename calls to match new patterns.
37 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
38 (load_pair<DREG:mode><DREG2:mode>): ... This.
39 (store_pair<mode>): Rename to...
40 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
41 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
43 (V_INT_EQUIV): Handle SImode.
44 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
47 2018-05-22 Martin Sebor <msebor@redhat.com>
50 * calls.c (maybe_warn_nonstring_arg): Use string length to set
51 or ajust the presumed bound on an operation to avoid unnecessary
54 2018-05-22 Martin Sebor <msebor@redhat.com>
56 PR tree-optimization/85826
57 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
58 assuming that a DECL necesarily has a constant size.
60 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
63 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
65 2018-05-22 Richard Biener <rguenther@suse.de>
67 PR tree-optimization/85834
68 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
69 non-constant and non-zero memset arguments.
71 2018-05-22 Martin Liska <mliska@suse.cz>
74 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
76 2018-05-22 Richard Biener <rguenther@suse.de>
78 PR tree-optimization/85863
79 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
80 comparisons when vectype is specified.
81 (vectorizable_condition): Do not specify vectype for
82 vect_is_simple_cond when SLP vectorizing.
84 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
87 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
88 define __ibm128 as long double.
89 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
90 as a distinct type when IEEE 128-bit support is enabled.
91 (init_float128_ieee): Fix up conversions between IFmode and IEEE
92 128-bit types to use the correct functions.
93 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
94 convert between 128-bit floating point types that have different
95 modes but the same representation, instead of using gen_lowpart to
97 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
99 (IFKF_reg): New attributes to give the register constraints for
101 (extend<mode>tf2_internal): New insns to mark an explicit
102 conversion between 128-bit floating point types that have a
103 different mode but share the same representation.
105 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
107 PR tree-optimization/85814
108 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
109 a null return from get_strinfo when unsharing the next
110 strinfo in the chain.
112 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
115 * varasm.c (weak_finish): Clean up weak_decls.
117 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
119 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
120 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
121 UNSPEC_UADALP values.
122 * config/aarch64/iterators.md (ABAL): New int iterator.
125 (sur): Add mappings for the above.
126 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
128 (aarch64_<sur>abal<mode>_4): Likewise.
129 (aarch64_<sur>adalp<mode>_3): Likewise.
130 (<sur>sadv16qi): New define_expand.
132 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
134 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
135 (*movdf_internal): Ditto.
136 (*rcpsf2_sse): Ditto.
137 (*rsqrtsf2_sse): Ditto.
138 (*sqrt<mode>2_sse): Ditto.
140 2018-05-21 Tamar Christina <tamar.christina@arm.com>
142 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
144 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
145 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
146 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
147 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
149 * config/aarch64/arm_neon.h: Likewise.
150 * config/aarch64/iterators.md (VQ_I): New.
152 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
154 * config.gcc: Add arc/t-multilib-linux to tmake_file for
156 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
159 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
161 * config/nds32/constraints.md (S): New constraint.
162 * config/nds32/nds32.md (call_internal): Use constraint S.
163 (call_value_internal): Likewise.
164 (sibcall_internal): Likewise.
165 (sibcall_value_internal): Likewise.
167 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
168 Chung-Ju Wu <jasonwucj@gmail.com>
170 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
173 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
174 Chung-Ju Wu <jasonwucj@gmail.com>
176 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
177 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
178 (nds32_rtx_costs_impl): Simplify.
179 (nds32_address_cost_impl): Simplify.
180 (nds32_init_rtx_costs): New function.
181 (nds32_rtx_costs_speed_prefer): Likewise.
182 (nds32_rtx_costs_size_prefer): Likewise.
183 (nds32_address_cost_speed_prefer): Likewise.
184 (nds32_address_cost_speed_fwprop): Likewise.
185 (nds32_address_cost_size_prefer): Likewise.
186 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
187 * config/nds32/nds32.c (nds32_option_override): Use
188 nds32_init_rtx_costs function.
190 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
192 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
193 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
194 (TARGET_PIPELINE_N8): Likewise.
195 (TARGET_PIPELINE_N10): Likewise.
196 (TARGET_PIPELINE_N13): Likewise.
197 (TARGET_PIPELINE_GRAYWOLF): Likewise.
199 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
201 * config/nds32/nds32-fpu.md: Update copyright year.
203 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
205 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
207 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
209 * config/nds32/nds32.c
210 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
211 * config/nds32/nds32.opt (minline-asm-r15): New option.
213 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
215 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
217 * config/nds32/nds32.md (abssi2): New pattern.
219 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
221 * config/i386/i386.md (rex64namesuffix): New mode attribute.
222 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
223 Merge insn pattern from sse_cvtsi2ss<round_name> and
224 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
225 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
226 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
227 using SWI48 mode iterator.
228 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
229 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
230 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
231 pattern from sse_cvttss2si<round_saeonly_name>
232 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
233 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
234 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
235 using SWI48 mode iterator.
236 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
237 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
238 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
239 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
240 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
241 using SWI48 mode iterator.
242 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
243 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
244 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
245 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
246 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
248 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
249 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
250 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
251 pattern from sse_cvttsd2si<round_saeonly_name>
252 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
254 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
256 * config/nds32/nds32-md-auxiliary.c
257 (nds32_valid_smw_lwm_base_p): Refine.
258 (nds32_output_smw_single_word): Refine.
259 (nds32_output_smw_double_word): New.
260 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
262 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
264 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
265 (nds32_output_stack_pop): Refine.
266 (nds32_expand_unaligned_load): Refine.
267 (nds32_expand_unaligned_store): Refine.
269 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
270 Chung-Ju Wu <jasonwucj@gmail.com>
272 * config/nds32/constants.md: Add TP_REGNUM constant.
273 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
274 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
276 * config/nds32/nds32-doubleword.md: Consider flag_pic.
277 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
278 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
279 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
280 and PIC code generation.
281 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
283 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
285 * config/nds32/nds32.md: Support TLS and PIC.
286 * config/nds32/nds32.c: Support TLS and PIC.
287 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
288 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
291 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
293 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
296 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
297 Chung-Ju Wu <jasonwucj@gmail.com>
299 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
300 * config/nds32/nds32-md-auxiliary.c
301 (symbolic_reference_mentioned_p): New.
302 (nds32_legitimize_ict_address): New.
303 (nds32_expand_ict_move): New.
304 (nds32_indirect_call_referenced_p): New.
305 (nds32_symbol_binds_local_p): Delete.
306 (nds32_long_call_p): Modify.
307 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
308 * config/nds32/nds32-protos.h
309 (symbolic_reference_mentioned_p): Declare.
310 (nds32_legitimize_ict_address): Declare.
311 (nds32_expand_ict_move): Declare.
312 (nds32_indirect_call_referenced_p): Declare.
313 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
314 (nds32_relax_group): Use nds32_ict_const_p as condition.
315 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
316 (nds32_asm_file_start): Output ict_model directive in asm code.
317 (nds32_legitimate_address_p): Consider indirect call.
318 (nds32_print_operand): Consider indirect call.
319 (nds32_print_operand_address): Consider indirect call.
320 (nds32_insert_attributes): Handle "indirect_call" attribute.
321 (TARGET_LEGITIMATE_ADDRESS_P): Define.
322 (TARGET_LEGITIMATE_CONSTANT_P): Define.
323 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
324 (TARGET_DELEGITIMIZE_ADDRESS): Define.
325 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
326 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
327 (TARGET_ICT_MODEL_SMALL): Define.
328 (TARGET_ICT_MODEL_LARGE): Define.
329 * config/nds32/nds32.md (movsi): Consider ict model.
330 (call, call_value): Consider ict model.
331 (sibcall, sibcall_value): Consider ict model.
332 * config/nds32/nds32.opt (mict-model): New option.
333 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
336 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
337 Monk Chiang <sh.chiang04@gmail.com>
338 Jim Wilson <jimw@sifive.com>
340 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
341 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
342 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
343 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
344 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
345 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
346 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
347 compute save_libcall_adjustment properly.
348 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
349 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
350 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
351 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
352 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
353 (ABI_SPEC): Handle mabi=ilp32e.
354 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
356 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
357 <-march>: Add rv32e as an example.
359 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
362 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
363 (intra_create_variable_infos): Handle C++ constructors.
365 2018-05-18 Martin Liska <mliska@suse.cz>
367 * passes.def: Remove a redundant pass.
369 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
372 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
374 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
376 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
378 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
380 (armv2, armv2a, armv3, armv3m): Delete architectures.
381 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
382 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
383 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
385 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
386 (*mulsidi3adddi): Likewise.
387 (mulsidi3): Likewise.
388 (*mulsidi3_nov6): Likewise.
389 (umulsidi3): Likewise.
390 (umulsidi3_nov6): Likewise.
391 (umaddsidi4): Likewise.
392 (*umulsidi3adddi): Likewise.
393 (smulsi3_highpart): Likewise.
394 (*smulsi3_highpart_nov6): Likewise.
395 (umulsi3_highpart): Likewise.
396 (*umulsi3_highpart_nov6): Likewise.
397 * config/arm/arm.h (arm_arch3m): Delete.
398 * config/arm/arm.c (arm_arch3m): Delete.
399 (arm_option_override_internal): Update armv3-related comment.
400 (arm_configure_build_target): Delete use of isa_bit_mode32.
401 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
402 (arm_rtx_costs_internal): Delete check of arm_arch3m.
403 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
406 * config/arm/arm-protos.h (arm_arch3m): Delete.
407 * config/arm/arm-tables.opt: Regenerate.
408 * config/arm/arm-tune.md: Likewise.
409 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
410 deleted architectures.
412 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
414 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
415 (armv5t, armv5te): New features.
416 (ARMv5, ARMv5e): Delete fgroups.
417 (ARMv5t, ARMv5te): Adjust for above changes.
419 (armv5, armv5e): Delete arches.
420 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
422 (*call_reg_arm): Likewise.
423 (*call_value_reg_armv5): Likewise.
424 (*call_value_reg_arm): Likewise.
425 (*call_symbol): Likewise.
426 (*call_value_symbol): Likewise.
427 (*sibcall_insn): Likewise.
428 (*sibcall_value_insn): Likewise.
430 (prefetch): Likewise.
431 (define_split and define_peephole2 dependent on arm_arch5):
433 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
435 (TARGET_ARM_QBIT): Likewise.
436 (TARGET_DSP_MULTIPLY): Likewise.
437 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
438 (arm_arch5, arm_arch5e): Delete.
439 (arm_arch5t, arm_arch5te): Declare.
440 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
441 (arm_arch5t): Declare.
442 (arm_option_reconfigure_globals): Update for the above.
443 (arm_options_perform_arch_sanity_checks): Update comment, replace
444 use of arm_arch5 with arm_arch5t.
445 (use_return_insn): Likewise.
446 (arm_emit_call_insn): Likewise.
447 (output_return_instruction): Likewise.
448 (arm_final_prescan_insn): Likewise.
449 (arm_coproc_builtin_available): Likewise.
450 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
451 arm_arch5e with arm_arch5t and arm_arch5te.
452 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
453 (arm_arch5t, arm_arch5te): Declare.
454 * config/arm/arm-tables.opt: Regenerate.
455 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
456 * config/arm/t-multilib: Likewise.
457 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
458 instead of arm_arch5.
459 (*call_reg_thumb1): Likewise.
460 (*call_value_reg_thumb1_v5): Likewise.
461 (*call_value_reg_thumb1): Likewise.
462 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
464 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
466 2018-05-18 Martin Liska <mliska@suse.cz>
468 PR gcov-profile/84846
469 * doc/gcov.texi: Document -t option of gcov tool.
471 2018-05-18 Martin Liska <mliska@suse.cz>
473 PR gcov-profile/84846
474 * gcov.c (print_usage): Add new -t option.
475 (process_args): Handle the option.
476 (generate_results): Use stdout as output when requested by
479 2018-05-18 Martin Liska <mliska@suse.cz>
481 PR gcov-profile/84846
482 * coverage.c (coverage_init): Write PWD to .gcno file.
483 * doc/gcov.texi: Document how working directory is printed.
484 * gcov-dump.c (dump_gcov_file): Print PWD.
485 * gcov.c (output_intermediate_file): Likewise.
486 (read_graph_file): Read PWD string.
487 (output_lines): Print PWD.
489 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
492 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
493 for retval and return false if all args to phi are zero.
495 2018-05-18 Richard Biener <rguenther@suse.de>
497 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
499 (evrp_dom_walker::before_dom_children): Call it.
501 2018-05-18 Richard Biener <rguenther@suse.de>
503 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
504 results when processing array refs with variable index.
506 2018-05-18 Toon Moene <toon@moene.org>
508 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
509 directly after that of -floop-interchange. Indicate that both
510 options are enabled by default when specifying -O3.
512 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
514 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
515 iterator. Delete separate integer-mode vec_set<mode> expander.
516 (aarch64_simd_vec_setv2di): Delete.
517 (vec_setv2di): Delete.
518 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
519 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
520 the "w, r" alternative.
522 2018-05-18 Martin Liska <mliska@suse.cz>
524 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
525 * tree-pass.h (make_pass_lower_switch_O0): New function.
526 * tree-switch-conversion.c (node_has_low_bound): Remove.
527 (node_has_high_bound): Likewise.
528 (node_is_bounded): Likewise.
529 (class pass_lower_switch): Make it a template type and create
531 (pass_lower_switch::execute): Add template argument.
532 (make_pass_lower_switch): New function.
533 (make_pass_lower_switch_O0): New function.
534 (do_jump_if_equal): Remove.
535 (emit_case_nodes): Simplify to just handle all 3 cases and leave
536 all the hard work to tree optimization passes.
538 2018-05-18 Martin Liska <mliska@suse.cz>
540 * dbgcnt.c (limit_low): Renamed from limit.
541 (limit_high): New variable.
542 (dbg_cnt_is_enabled): Check for upper limit.
543 (dbg_cnt): Adjust dumping.
544 (dbg_cnt_set_limit_by_index): Add new argument for high
546 (dbg_cnt_set_limit_by_name): Likewise.
547 (dbg_cnt_process_single_pair): Parse new format.
548 (dbg_cnt_process_opt): Use strtok.
549 (dbg_cnt_list_all_counters): Remove 'value' and add
551 * doc/invoke.texi: Document changes.
553 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
555 * doc/sourcebuild.texi (scalar_all_fma): Document.
556 * tree.def (FMA_EXPR): Delete.
557 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
558 * internal-fn.c (ternary_direct): New macro.
559 (expand_ternary_optab_fn): Likewise.
560 (direct_ternary_optab_supported_p): Likewise.
561 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
562 * builtins.c (fold_builtin_fma): Delete.
563 (fold_builtin_3): Don't call it.
564 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
565 * expr.c (expand_expr_real_2): Likewise.
566 * fold-const.c (operand_equal_p): Likewise.
567 (fold_ternary_loc): Likewise.
568 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
569 * gimple.c (DEFTREECODE): Likewise.
570 * gimplify.c (gimplify_expr): Likewise.
571 * optabs-tree.c (optab_for_tree_code): Likewise.
572 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
573 * tree-eh.c (operation_could_trap_p): Likewise.
574 (stmt_could_throw_1_p): Likewise.
575 * tree-inline.c (estimate_operator_cost): Likewise.
576 * tree-pretty-print.c (dump_generic_node): Likewise.
577 (op_code_prio): Likewise.
578 * tree-ssa-loop-im.c (stmt_cost): Likewise.
579 * tree-ssa-operands.c (get_expr_operands): Likewise.
580 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
581 * fold-const-call.h (fold_fma): Delete.
582 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
583 CFN_FNMA and CFN_FNMS.
585 * genmatch.c (combined_fn): New enum.
586 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
587 (commutative_op): New function.
588 (commutate): Use it. Handle more than 2 operands.
589 (dt_operand::gen_gimple_expr): Use commutative_op.
590 (parser::parse_expr): Allow :c to be used with non-binary
591 operators if the commutative operand is known.
592 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
593 CFN_FMS, CFN_FNMA and CFN_FNMS.
594 (backprop::process_assign_use): Remove FMA_EXPR handling.
595 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
596 (gen_hsa_fma): New function.
597 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
598 IFN_FNMA and IFN_FNMS.
599 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
600 * gimple-fold.h (follow_all_ssa_edges): Declare.
601 * gimple-fold.c (follow_all_ssa_edges): New function.
602 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
603 gimple_build interface and use follow_all_ssa_edges to fold the result.
604 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
605 instead of checking for optabs directly.
606 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
607 rather than FMA_EXPRs.
608 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
609 call to IFN_FMA instead of an FMA_EXPR.
611 2018-05-17 Jim Wilson <jimw@sifive.com>
613 * expr.c (do_tablejump): When converting index to Pmode, if we have a
614 sign extended promoted subreg, and the range does not have the sign bit
615 set, then do a sign extend.
617 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
618 test, check for sign extended subreg and/or constant operands, and
619 do a sign extend in that case.
621 2018-05-17 Steve Ellcey <sellcey@cavium.com>
623 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
624 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
626 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
627 Change logics_shift_reg to logics_shift_imm.
628 (thunderx2t99_fp_loadpair_basic): Delete.
629 (thunderx2t99_fp_storepair_basic): Delete.
630 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
631 (thunderx2t99_asimd_polynomial): Delete.
632 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
633 and neon_fp_mul_d_scalar_q.
634 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
635 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
636 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
637 (thunderx2t99_asimd_lut): Add missing tbl types.
638 (thunderx2t99_asimd_ext): Delete.
639 (thunderx2t99_asimd_load1_1_mult): Delete.
640 (thunderx2t99_asimd_load1_2_mult): Delete.
641 (thunderx2t99_asimd_load1_ldp): New.
642 (thunderx2t99_asimd_load1): New.
643 (thunderx2t99_asimd_load2): Add missing *load2* types.
644 (thunderx2t99_asimd_load3): New.
645 (thunderx2t99_asimd_load4): New.
646 (thunderx2t99_asimd_store1_1_mult): Delete.
647 (thunderx2t99_asimd_store1_2_mult): Delete.
648 (thunderx2t99_asimd_store2_mult): Delete.
649 (thunderx2t99_asimd_store2_onelane): Delete.
650 (thunderx2t99_asimd_store_stp): New.
651 (thunderx2t99_asimd_store1): New.
652 (thunderx2t99_asimd_store2): New.
653 (thunderx2t99_asimd_store3): New.
654 (thunderx2t99_asimd_store4): New.
656 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
658 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
659 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
661 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
662 Segher Boessenkool <segher@kernel.crashing.org>
665 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
668 2018-05-17 Richard Biener <rguenther@suse.de>
670 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
671 for pruning loop and prune defs feeding only already visited PHIs.
673 2018-05-17 Richard Biener <rguenther@suse.de>
675 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
677 2018-05-17 Bin Cheng <bin.cheng@arm.com>
678 Richard Biener <rguenther@suse.de>
680 PR tree-optimization/85793
681 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
682 for VMAT_ELEMENTWISE.
684 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
686 * internal-fn.h (lookup_internal_fn): Declare
687 * internal-fn.c (lookup_internal_fn): New function.
688 * gimple.c (gimple_build_call_from_tree): Handle calls to
690 * gimple-pretty-print.c (dump_gimple_call): Print "." before
691 internal function names.
692 * tree-pretty-print.c (dump_generic_node): Likewise.
693 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
695 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
697 * gimple-fold.h (gimple_build): Make the function forms take
698 combined_fn rather than built_in_function.
699 (gimple_simplify): Likewise.
700 * gimple-match-head.c (gimple_simplify): Likewise.
701 * gimple-fold.c (gimple_build): Likewise.
702 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
703 rather than gimple_build_call_internal.
704 (get_initial_defs_for_reduction): Likewise.
705 (vect_create_epilog_for_reduction): Likewise.
706 (vectorizable_live_operation): Likewise.
708 2018-05-17 Martin Liska <mliska@suse.cz>
710 * gimple-ssa-sprintf.c (format_directive): Do not use
711 space in between 'G_' and '('.
713 2018-05-17 Jakub Jelinek <jakub@redhat.com>
716 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
717 even if the mask is not all ones.
720 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
722 (ix86_gimple_fold_builtin): Likewise.
725 * config/i386/i386.c: Include tree-vector-builder.h.
726 (ix86_vector_shift_count): New function.
727 (ix86_fold_builtin): Fold shift builtins by scalar count.
728 (ix86_gimple_fold_builtin): Likewise.
730 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
731 _mm512_setzero): New intrinsics.
733 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
734 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
736 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
737 code generation for cases where splatting a value is not useful.
738 * simplify-rtx.c (simplify_ternary_operation): Simplify
739 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
740 mode to a vec_concat.
742 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
744 * config.gcc: Support "goldmont-plus".
745 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
747 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
748 PROCESSOR_GOLDMONT_PLUS.
749 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
750 (processor_target_table): Add "goldmont-plus".
751 (PTA_GOLDMONT_PLUS): Define.
752 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
753 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
754 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
755 (fold_builtin_cpu): Add "goldmont-plus".
756 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
757 (ix86_option_override_internal): Add "goldmont-plus".
758 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
759 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
760 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
761 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
763 2018-05-17 Richard Biener <rguenther@suse.de>
765 PR tree-optimization/85757
766 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
767 remove defs that only feed that PHI from further processing.
769 2018-05-16 Jim Wilson <jimw@sifive.com>
771 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
773 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
775 2018-05-16 Mark Wielaard <mark@klomp.org>
777 * dwarf2out.c (count_index_strings): New function.
778 (output_indirect_strings): Call count_index_strings and generate
779 header for dwarf_version >= 5.
781 2018-05-16 Mark Wielaard <mark@klomp.org>
783 * dwarf2out.c (dwarf_FORM): New function.
784 (set_indirect_string): Use dwarf_FORM.
785 (reset_indirect_string): Likewise.
786 (size_of_die): Likewise.
787 (value_format): Likewise.
788 (output_die): Likewise.
789 (add_skeleton_AT_string): Likewise.
790 (output_macinfo_op): Likewise.
791 (index_string): Likewise.
792 (output_index_string_offset): Likewise.
793 (output_index_string): Likewise.
794 (count_index_strings): Likewise.
796 2018-05-16 Carl Love <cel@us.ibm.com>
798 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
799 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
801 2018-05-16 Martin Jambor <mjambor@suse.cz>
803 * ipa-prop.c (ipa_free_all_edge_args): Remove.
804 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
806 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
808 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
809 (fnma<mode>4): Likewise.
810 (fms<mode>4): Likewise.
811 (fnms<mode>4): Likewise.
812 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
813 (aarch64_fnma<mode>4): Likewise.
814 (aarch64_fms<mode>4): Likewise.
815 (aarch64_fnms<mode>4): Likewise.
816 (aarch64_fnmadd<mode>4): Likewise.
818 2018-05-16 Jason Merrill <jason@redhat.com>
820 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
822 2018-05-16 Richard Biener <rguenther@suse.de>
824 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
825 (dump_stmt_cost): Declare.
826 (add_stmt_cost): Dump cost we add.
827 (add_stmt_costs): New function.
828 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
830 (vect_analyze_stmt): Adjust prototype.
831 (vectorizable_condition): Likewise.
832 (vectorizable_live_operation): Likewise.
833 (vectorizable_reduction): Likewise.
834 (vectorizable_induction): Likewise.
835 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
836 cost vector to pass to vectorizable_ and record afterwards.
837 (vect_model_reduction_cost): Take cost vector argument and adjust.
838 (vect_model_induction_cost): Likewise.
839 (vectorizable_reduction): Likewise.
840 (vectorizable_induction): Likewise.
841 (vectorizable_live_operation): Likewise.
842 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
843 SLP_TREE_NUMBER_OF_VEC_STMTS.
844 (vect_analyze_slp_cost_1): Remove.
845 (vect_analyze_slp_cost): Likewise.
846 (vect_slp_analyze_node_operations): Take visited args and
847 a target cost vector. Avoid processing already visited stmt sets.
848 (vect_slp_analyze_operations): Use a local cost vector to gather
849 costs and register those of non-discarded instances.
850 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
851 (vect_schedule_slp_instance): Remove copying of
852 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
854 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
855 adding cost. Record cost entry location.
856 (vect_prologue_cost_for_slp_op): Function to compute cost of
857 a constant or invariant generated for SLP vect in the prologue,
858 split out from vect_analyze_slp_cost_1.
859 (vect_model_simple_cost): Make static. Adjust for SLP costing.
860 (vect_model_promotion_demotion_cost): Likewise.
861 (vect_model_store_cost): Likewise, make static.
862 (vect_model_load_cost): Likewise.
863 (vectorizable_bswap): Add cost vector arg and adjust.
864 (vectorizable_call): Likewise.
865 (vectorizable_simd_clone_call): Likewise.
866 (vectorizable_conversion): Likewise.
867 (vectorizable_assignment): Likewise.
868 (vectorizable_shift): Likewise.
869 (vectorizable_operation): Likewise.
870 (vectorizable_store): Likewise.
871 (vectorizable_load): Likewise.
872 (vectorizable_condition): Likewise.
873 (vectorizable_comparison): Likewise.
874 (can_vectorize_live_stmts): Likewise.
875 (vect_analyze_stmt): Likewise.
876 (vect_transform_stmt): Adjust calls to vectorizable_*.
877 * tree-vectorizer.c: Include gimple-pretty-print.h.
878 (dump_stmt_cost): New function.
880 2018-05-16 Richard Biener <rguenther@suse.de>
882 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
883 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
884 * tree-ssa-dse.c: Include tree-ssa-loop.h.
885 (check_name): New callback.
886 (dse_classify_store): Track cycles via a visited bitmap of PHI
887 defs and simplify handling of in-loop and across loop dead stores
888 and properly fail for loop-variant refs. Handle byte-tracking with
889 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
892 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
894 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
895 (vect_get_mask_type_for_stmt): Likewise.
896 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
898 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
899 to determine the statement's vector type and the vector type that
900 should be used for calculating nunits. Deal with cases in which
901 the type has to be deferred.
902 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
903 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
904 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
905 (vect_determine_vf_for_stmt): New functions, split out from...
906 (vect_determine_vectorization_factor): ...here.
907 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
908 (vect_get_mask_type_for_stmt): New functions, split out from
909 vect_determine_vectorization_factor.
911 2018-05-16 Richard Biener <rguenther@suse.de>
913 * tree-cfg.c (verify_gimple_assign_ternary): Properly
914 verify the [VEC_]COND_EXPR embedded comparison.
916 2018-05-15 Martin Sebor <msebor@redhat.com>
918 PR tree-optimization/85753
919 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
920 RECORD_TYPE in addition to ARRAY_TYPE.
922 2018-05-15 Martin Sebor <msebor@redhat.com>
925 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
927 2018-05-15 Richard Biener <rguenther@suse.de>
929 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
930 add by_clobber_p one. Change algorithm to collect all defs
931 representing uses we need to walk and try reducing them to
932 a single one before failing.
933 (dse_dom_walker::dse_optimize_stmt): Adjust.
935 2018-05-13 Mark Wielaard <mark@klomp.org>
937 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
938 (size_of_loc_descr): Likewise.
939 (output_loc_operands): Likewise.
940 (output_loc_operands_raw): Likewise.
941 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
942 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
943 (hash_loc_operands): Likewise.
944 (compare_loc_operands): Likewise.
946 2018-05-14 Mark Wielaard <mark@klomp.org>
948 * dwarf2out.c (count_index_addrs): New function.
949 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
951 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
953 PR tree-optimization/83648
954 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
955 return value as malloc candidate.
957 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
960 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
961 as true in call to suggest_attribute.
963 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
965 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
968 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
970 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
972 (aarch64_crypto_aesd_fused): Likewise.
974 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
976 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
977 (movsi_aarch64): Likewise.
978 (load_pairsi): Likewise.
979 (load_pairdi): Likewise.
980 (store_pairsi): Likewise.
981 (store_pairdi): Likewise.
982 (load_pairsf): Likewise.
983 (load_pairdf): Likewise.
984 (store_pairsf): Likewise.
985 (store_pairdf): Likewise.
986 (zero_extend): Likewise.
987 (trunc): Swap alternatives.
988 (fcvt_target): Add '?' to prefer w over r.
990 2018-05-14 Jakub Jelinek <jakub@redhat.com>
993 * config/i386/i386.md: Disallow non-commutative arithmetics in
994 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
995 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
996 in the peephole2 before it.
998 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1000 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1001 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1002 (ix86_handle_option): Handle -mcldemote.
1003 * config.gcc: New header.
1004 * config/i386/cldemoteintrin.h: New file.
1005 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1006 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1008 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1009 OPTION_MASK_ISA_CLDEMOTE.
1010 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1011 (ix86_valid_target_attribute_inner_p): Ditto.
1012 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1013 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1014 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1015 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1016 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1018 * config/i386/i386.opt: Add -mcldemote.
1019 * config/i386/x86intrin.h: New header.
1020 * doc/invoke.texi: Add -mcldemote.
1022 2018-05-14 Richard Biener <rguenther@suse.de>
1024 * doc/match-and-simplify.texi: Adjust :s documentation.
1026 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1028 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1029 intended memcpy size.
1030 (REORDER_45): Likewise.
1032 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1034 * sort.cc: New file.
1035 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1036 * vec.c (qsort_chk): Use gcc_qsort.
1037 * Makefile.in (OBJS-libcommon): Add sort.o.
1038 (build/sort.o): New target. Use it...
1039 (BUILD_RTL): ... here, and...
1040 (build/gencfn-macros): ... here, and...
1041 (build/genmatch): ... here.
1043 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1044 Chung-Ju Wu <jasonwucj@gmail.com>
1046 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1047 * config/nds32/nds32-graywolf.md: New file.
1048 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1049 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1051 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1052 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1053 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1054 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1055 * config/nds32/pipelines.md: Include n15 settings.
1057 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1058 Chung-Ju Wu <jasonwucj@gmail.com>
1060 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1061 * config/nds32/nds32-n13.md: New file.
1062 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1063 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1065 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1066 * config/nds32/nds32.md (pipeline_model): Add n13.
1067 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1068 * config/nds32/pipelines.md: Include n13 settings.
1070 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1071 Chung-Ju Wu <jasonwucj@gmail.com>
1073 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1074 * config/nds32/nds32-n10.md: New file.
1075 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1076 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1078 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1079 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1080 * config/nds32/nds32.md (pipeline_model): Add n10.
1081 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1082 * config/nds32/pipelines.md: Include n10 settings.
1084 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1085 Kito Cheng <kito.cheng@gmail.com>
1086 Chung-Ju Wu <jasonwucj@gmail.com>
1088 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1089 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1090 Add enum values for DSP extension instructions.
1091 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1093 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1094 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1096 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1097 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1098 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1099 * config/nds32/nds32-intrinsic.md: Likewise.
1100 * config/nds32/nds32_intrinsic.h: Likewise.
1101 * config/nds32/nds32-md-auxiliary.c: Likewise.
1102 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1103 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1104 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1105 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1106 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1107 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1108 TYPE_DMAC in switch statement.
1109 * config/nds32/nds32.c: New checking and implementation for DSP
1110 extension instructions.
1111 * config/nds32/nds32.h: Likewise.
1112 * config/nds32/nds32.md: Likewise.
1113 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1114 * config/nds32/predicates.md: Implement new predicates for DSP
1117 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1119 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1120 Reformat alternatives and attributes so it is easier to identify
1121 which constraints/attributes go with which instruction.
1122 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1123 (mov<mode>_softfloat32, FMOVE64): Likewise.
1124 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1125 (mov<mode>_softfloat64, FMOVE64): Likewise.
1127 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1129 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1131 (Basic PowerPC Built-in Functions): The new name of the
1132 subsection previously known as "PowerPC Built-in Functions".
1133 (Basic PowerPC Built-in Functions Available on all Configurations):
1135 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1136 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1137 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1138 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1140 2018-05-11 Martin Jambor <mjambor@suse.cz>
1143 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1146 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1149 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1151 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1153 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1154 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1155 (ix86_handle_option): Handle -mwaitpkg.
1156 * config.gcc: New header.
1157 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1158 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1159 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1161 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1162 OPTION_MASK_ISA_WAITPKG.
1163 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1164 (ix86_option_override_internal): Add PTA_WAITPKG.
1165 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1166 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1167 IX86_BUILTIN_TPAUSE.
1168 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1169 __builtin_ia32_umwait and __builtin_ia32_tpause.
1170 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1171 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1172 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1173 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1174 UNSPECV_TPAUSE): New.
1175 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1176 * config/i386/i386.opt: Add -mwaitpkg.
1177 * config/i386/waitpkgintrin.h: New file.
1178 * config/i386/x86intrin.h: New header.
1179 * doc/invoke.texi: Add -mwaitpkg.
1181 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1184 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1186 (cortex-m0): Use armv6s-m isa.
1187 (cortex-m0plus): Likewise.
1188 (cortex-m1): Likewise.
1189 (cortex-m0.small-multiply): Likewise.
1190 (cortex-m0plus.small-multiply): Likewise.
1191 (cortex-m1.small-multiply): Likewise.
1193 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1194 Jakub Jelinek <jakub@redhat.com>
1196 PR tree-optimization/85692
1197 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1198 source permute as well.
1200 2018-05-11 Martin Liska <mliska@suse.cz>
1203 * doc/extend.texi: Document LLVM style format for no_sanitize
1206 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1208 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1209 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1210 (mode_supports_vsx_dform_quad): Likewise.
1211 (mode_supports_vmx_dform): Move these functions to be next to the
1212 other mode_supports functions.
1213 (mode_supports_dq_form): Likewise.
1214 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1215 mode_supports_dq_form.
1216 (reg_offset_addressing_ok_p): Likewise.
1217 (offsettable_ok_by_alignment): Likewise.
1218 (rs6000_legitimate_offset_address_p): Likewise.
1219 (legitimate_lo_sum_address_p): Likewise.
1220 (rs6000_legitimize_address): Likewise.
1221 (rs6000_legitimize_reload_address): Likewise.
1222 (rs6000_secondary_reload_inner): Likewise.
1223 (rs6000_preferred_reload_class): Likewise.
1224 (rs6000_output_move_128bit): Likewise.
1226 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1228 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1229 Generate SImode target register for null target.
1230 <case IX86_BUILTIN_XGETBV>: Ditto.
1231 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1232 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1234 2018-05-10 Carl Love <cel@us.ibm.com>
1236 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1237 dcbtt and dcbtstt if operands[2] is 0.
1239 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1242 * config/i386/sse.md (usadv64qi): New expander.
1244 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1246 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1247 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1248 -maltivec=be support.
1249 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1250 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1251 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1252 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1253 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1254 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1255 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1256 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1257 altivec_vsumsws): Adjust.
1258 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1259 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1260 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1262 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1263 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1264 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1265 (altivec_lve<VI_char>x): Delete expand.
1266 (*altivec_lve<VI_char>x_internal): Rename to...
1267 (altivec_lve<VI_char>x): ... this.
1268 (altivec_lvxl_<mode>): Delete expand.
1269 (*altivec_lvxl_<mode>_internal): Rename to ...
1270 (altivec_lvxl_<mode>): ... this.
1271 (altivec_stvxl_<mode>): Delete expand.
1272 (*altivec_stvxl_<mode>_internal): Rename to ...
1273 (altivec_stvxl_<mode>): ... this.
1274 (altivec_stve<VI_char>x): Delete expand.
1275 (*altivec_stve<VI_char>x_internal): Rename to ...
1276 (altivec_stve<VI_char>x): ... this.
1277 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1278 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1279 reduc_plus_scal_<mode>): Adjust.
1280 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1282 (rs6000_cpu_cpp_builtins): Adjust.
1283 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1284 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1285 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1286 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1287 -maltivec=be support.
1288 (rs6000_split_vec_extract_var): Adjust.
1289 (rs6000_split_v4si_init): Adjust.
1290 (swap_selector_for_mode): Delete.
1291 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1292 altivec_expand_stvex_be): Delete.
1293 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1294 -maltivec=be support.
1295 (rs6000_gimple_fold_builtin): Ditto.
1296 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1298 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1299 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1300 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1301 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1302 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1303 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1304 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1305 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1306 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1307 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1308 anonymous split): Adjust.
1309 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1310 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1312 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1314 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1315 when --with-gxx-include-dir is also specified.
1316 * configure: Regenerate.
1318 2018-05-09 Jim Wilson <jimw@sifive.com>
1321 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1322 * config/riscv/t-withmultilib: New.
1323 * config/riscv/withmultilib.h: New.
1324 * doc/install.texi: Document RISC-V --with-multilib-list support.
1326 2018-05-09 Richard Biener <rguenther@suse.de>
1328 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1330 (vect_bb_vectorization_profitable_p): Adjust. Compute
1331 actual scalar cost using the cost vector and the add_stmt_cost
1334 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1336 PR rtl-optimization/85645
1337 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1338 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1340 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1342 PR rtl-optimization/85645
1343 * shrink-wrap.c (spread_components): Return a boolean saying if
1344 anything was changed.
1345 (try_shrink_wrapping_separate): Iterate spread_components until
1346 nothing changes anymore.
1348 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1350 PR rtl-optimization/85645
1351 * regrename.c (build_def_use): Also kill the chains that include the
1352 destination of a REG_CFA_REGISTER note.
1354 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1356 PR rtl-optimization/85645
1357 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1358 insn that has a REG_CFA_REGISTER note.
1360 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1362 * cfgexpand.c (expand_clobber): New function.
1363 (expand_gimple_stmt_1): Use it.
1364 * tree-vect-stmts.c (vect_clobber_variable): New function,
1366 (vectorizable_simd_clone_call): ...here.
1367 (vectorizable_store): Emit a clobber either side of an
1368 IFN_STORE_LANES sequence.
1369 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1371 2018-05-09 Tom de Vries <tom@codesourcery.com>
1374 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1375 (define_insn "trap_if_false"): Add exit after trap.
1377 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1379 PR rtl-optimization/85638
1380 * bb-reorder.c: Include common/common-target.h.
1381 (create_forwarder_block): New function extracted from...
1382 (fix_up_crossing_landing_pad): ...here. Rename into...
1383 (dw2_fix_up_crossing_landing_pad): ...this.
1384 (sjlj_fix_up_crossing_landing_pad): New function.
1385 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1386 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1387 from both partitions and exit the loop after one iteration.
1389 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1392 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1394 (Basic PowerPC Built-in Functions): The new name of the
1395 subsection previously known as "PowerPC Built-in Functions".
1396 (Basic PowerPC Built-in Functions Available on all Configurations):
1398 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1400 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1402 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1404 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1407 2018-05-08 Jim Wilson <jimw@sifive.com>
1409 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1410 (LD_EMUL_SUFFIX): New.
1411 (LINK_SPEC): Use it.
1413 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1415 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1417 (Basic PowerPC Built-in Functions): The new name of the
1418 subsection previously known as "PowerPC Built-in Functions".
1419 (Basic PowerPC Built-in Functions Available on all Configurations):
1421 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1423 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1425 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1427 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1430 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1433 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1434 after cmpelim optimization.
1436 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1438 * config.gcc: Support "goldmont".
1439 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1440 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1442 * config/i386/i386.c (m_GOLDMONT): Define.
1443 (processor_target_table): Add "goldmont".
1444 (PTA_GOLDMONT): Define.
1445 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1446 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1447 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1448 (fold_builtin_cpu): Add "goldmont".
1449 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1450 (ix86_option_override_internal): Add "goldmont".
1451 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1452 (processor_type): Add PROCESSOR_GOLDMONT.
1453 * config/i386/i386.md: Add CPU "glm".
1454 * config/i386/glm.md: New file.
1455 * config/i386/x86-tune.def: Add m_GOLDMONT.
1456 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1458 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1461 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1463 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1464 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1465 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1468 * config/i386/i386.c (ix86_fold_builtin): Handle
1469 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1472 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1473 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1475 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1478 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1479 (check_arch): Likewise.
1480 (check_fpu): Return the result rather than printing it.
1481 (end arch): Fix operator precedence.
1482 (end cpu): Likewise.
1483 (END): Print the result from check_fpu.
1485 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1486 Alan Hayward <alan.hayward@arm.com>
1487 David Sherwood <david.sherwood@arm.com>
1489 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1490 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1491 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1492 (*fcmuo<mode>_and): New patterns.
1494 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1496 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1497 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1498 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1499 (cmp_op, sve_imm_con): New code attributes.
1500 (SVE_COND_INT_CMP, imm_con): Delete.
1501 (cmp_op): Remove above unspecs from int attribute.
1502 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1504 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1505 comparison-specific unspecs.
1506 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1507 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1508 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1509 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1510 (*vec_fcm<cmp_op><mode>): Rename to...
1511 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1512 (*vec_fcmuo<mode>): Rename to...
1513 (*fcmuo<mode>): ...this and adjust likewise.
1514 (*pred_fcm<cmp_op><mode>): New pattern.
1515 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1516 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1518 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1520 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1521 (aarch64_emit_sve_predicated_cond): New function.
1522 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1523 (aarch64_emit_unspec_cond_or): Replace with...
1524 (aarch64_emit_sve_or_conds): ...this new function. Use
1525 aarch64_emit_sve_ptrue_op for the individual comparisons and
1526 aarch64_emit_binop to OR them together.
1527 (aarch64_emit_inverted_unspec_cond): Replace with...
1528 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1529 aarch64_emit_sve_ptrue_op for the comparison and
1530 aarch64_emit_unop to invert the result.
1531 (aarch64_expand_sve_vec_cmp_float): Update after the above
1532 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1534 2018-05-07 Nathan Sidwell <nathan@acm.org>
1536 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1537 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1538 (Backwards Compatibility): Likewise.
1540 2018-05-07 Luis Machado <luis.machado@linaro.org>
1544 2018-05-07 Luis Machado <luis.machado@linaro.org>
1546 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1547 <prefetch_dynamic_strides>: New const bool field.
1548 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1549 prefetch_dynamic_strides.
1550 (exynosm1_prefetch_tune): Likewise.
1551 (thunderxt88_prefetch_tune): Likewise.
1552 (thunderx_prefetch_tune): Likewise.
1553 (thunderx2t99_prefetch_tune): Likewise.
1554 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1555 (aarch64_override_options_internal): Update to set
1556 PARAM_PREFETCH_DYNAMIC_STRIDES.
1557 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1558 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1559 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1560 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1561 prefetch-dynamic-strides setting.
1563 2018-05-07 Luis Machado <luis.machado@linaro.org>
1565 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1566 <minimum_stride>: New const int field.
1567 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1568 minimum_stride field.
1569 (exynosm1_prefetch_tune): Likewise.
1570 (thunderxt88_prefetch_tune): Likewise.
1571 (thunderx_prefetch_tune): Likewise.
1572 (thunderx2t99_prefetch_tune): Likewise.
1573 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1574 (aarch64_override_options_internal): Update to set
1575 PARAM_PREFETCH_MINIMUM_STRIDE.
1576 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1577 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1578 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1579 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1580 stride is constant and is below the minimum stride threshold.
1582 2018-05-07 Luis Machado <luis.machado@linaro.org>
1584 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1587 2018-05-07 Luis Machado <luis.machado@linaro.org>
1589 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1590 <prefetch_dynamic_strides>: New const bool field.
1591 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1592 prefetch_dynamic_strides.
1593 (exynosm1_prefetch_tune): Likewise.
1594 (thunderxt88_prefetch_tune): Likewise.
1595 (thunderx_prefetch_tune): Likewise.
1596 (thunderx2t99_prefetch_tune): Likewise.
1597 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1598 (aarch64_override_options_internal): Update to set
1599 PARAM_PREFETCH_DYNAMIC_STRIDES.
1600 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1601 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1602 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1603 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1604 prefetch-dynamic-strides setting.
1606 2018-05-07 Luis Machado <luis.machado@linaro.org>
1608 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1609 <minimum_stride>: New const int field.
1610 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1611 minimum_stride field.
1612 (exynosm1_prefetch_tune): Likewise.
1613 (thunderxt88_prefetch_tune): Likewise.
1614 (thunderx_prefetch_tune): Likewise.
1615 (thunderx2t99_prefetch_tune): Likewise.
1616 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1617 (aarch64_override_options_internal): Update to set
1618 PARAM_PREFETCH_MINIMUM_STRIDE.
1619 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1620 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1621 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1622 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1623 stride is constant and is below the minimum stride threshold.
1625 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1628 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1629 the type is addressable. Don't force op into register if it has
1632 2018-05-05 Roland McGrath <mcgrathr@google.com>
1635 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1636 any section for which we don't know a specific type it should have,
1637 regardless of name. Previously this was done only for the exact
1638 names ".init_array", ".fini_array", and ".preinit_array".
1639 (default_elf_asm_named_section): Add comment about
1640 relationship with default_section_type_flags and SECTION_NOTYPE.
1641 (get_section): Don't consider it a type conflict if one side has
1642 SECTION_NOTYPE and the other doesn't, as long as neither has the
1643 SECTION_BSS et al used in the default_section_type_flags logic.
1645 2018-05-05 Tom de Vries <tom@codesourcery.com>
1648 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1649 (workaround_barsyncs): New function.
1650 (nvptx_reorg): Use workaround_barsyncs.
1651 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1652 (define_expand "nvptx_membar_cta"): New define_expand.
1653 (define_insn "*nvptx_membar_cta"): New insn.
1655 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1657 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1658 To improve optimization opportunities.
1659 * builtin-types.def: The new needed builtin types for the above.
1661 2018-05-04 Richard Biener <rguenther@suse.de>
1663 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1664 * gimple-ssa-store-merging.c
1665 (imm_store_chain_info::output_merged_store): Remove redundant create,
1666 release split_store vector contents on failure.
1667 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1668 scalar stmt vector on cache hit.
1670 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1672 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1674 * config.gcc (powerpc-xilinx-eabi*): Remove.
1675 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1677 (fusion_addis_mem_combo_load): Ditto.
1678 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1680 (rs6000_cpu_cpp_builtins): Ditto.
1681 * config/rs6000/rs6000-linux.c
1682 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1683 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1684 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1686 (rs6000_setup_reg_addr_masks): Ditto.
1687 (rs6000_init_hard_regno_mode_ok): Ditto.
1688 (rs6000_option_override_internal): Ditto.
1689 (legitimate_lo_sum_address_p): Ditto.
1690 (rs6000_legitimize_address): Ditto.
1691 (rs6000_legitimize_reload_address): Ditto.
1692 (rs6000_legitimate_address_p): Ditto.
1693 (abi_v4_pass_in_fpr): Ditto.
1694 (setup_incoming_varargs): Ditto.
1695 (rs6000_gimplify_va_arg): Ditto.
1696 (rs6000_split_multireg_move): Ditto.
1697 (rs6000_savres_strategy): Ditto.
1698 (rs6000_emit_prologue_components): Ditto.
1699 (rs6000_emit_epilogue_components): Ditto.
1700 (rs6000_emit_prologue): Ditto.
1701 (rs6000_emit_epilogue): Ditto.
1702 (rs6000_elf_file_end): Ditto.
1703 (rs6000_function_value): Ditto.
1704 (rs6000_libcall_value): Ditto.
1705 * config/rs6000/rs6000.h: Ditto.
1706 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1707 (TARGET_MINMAX): ... this. New.
1708 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1709 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1710 (*movsi_internal1_single): Delete.
1711 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1712 mfpu=, mxilinx-fpu): Delete.
1713 * config/rs6000/singlefp.h: Delete.
1714 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1715 * config/rs6000/t-rs6000: Ditto.
1716 * config/rs6000/t-xilinx: Delete.
1717 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1718 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1719 (VStype_simple): Delete.
1720 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1721 * config/rs6000/xfpu.h: Delete.
1722 * config/rs6000/xfpu.md: Delete.
1723 * config/rs6000/xilinx.h: Delete.
1724 * config/rs6000/xilinx.opt: Delete.
1725 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1726 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1728 2018-05-04 Tom de Vries <tom@codesourcery.com>
1731 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1734 2018-05-04 Richard Biener <rguenther@suse.de>
1737 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1738 (expand_complex_div_wide): Likewise.
1739 (expand_complex_operations_1): Likewise.
1740 (expand_complex_libcall): Preserve EH info of the original stmt.
1741 (tree_lower_complex): Handle removed blocks.
1742 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1743 on complex multiplication and division libcall builtins.
1745 2018-05-04 Richard Biener <rguenther@suse.de>
1748 * fold-const.c (negate_expr_p): Restrict negation of operand
1749 zero of a division to when we know that can happen without
1751 (fold_negate_expr_1): Likewise.
1753 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1756 * real.h (real_nextafter): Declare.
1757 * real.c (real_nextafter): New function.
1758 * fold-const-call.c (fold_const_nextafter): New function.
1759 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1760 CASE_CFN_NEXTTOWARD.
1761 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1762 even when arg1_mode is different from arg0_mode.
1764 2018-05-03 Nathan Sidwell <nathan@acm.org>
1766 * doc/extend.texi (Deprecated Features): Remove
1768 (Backwards Compatibility): Likewise.
1769 * doc/invoke.texi (C++ Language Options): Likewise.
1770 (C++ Dialect Options): Likewise.
1772 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1775 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1776 _mm512_mask_mullox_epi64): New intrinsics.
1778 2018-05-03 Tom de Vries <tom@codesourcery.com>
1781 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1782 dump files): Add offload-tree.
1784 2018-05-03 Richard Biener <rguenther@suse.de>
1786 PR tree-optimization/85615
1787 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1788 to loops not nested in BBs loop father to avoid creating multi-entry
1791 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1793 PR tree-optimization/70291
1794 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1795 arguments. Change return type to tree. Emit libcall as a new
1796 statement rather than replacing existing one when inplace_p is true.
1797 (expand_complex_multiplication_components): New function.
1798 (expand_complex_multiplication): Expand floating-point complex
1799 multiplication using the above.
1800 (expand_complex_division): Rename inner_type parameter to type.
1801 Update expand_complex_libcall call-site.
1802 (expand_complex_operations_1): Update expand_complex_multiplication
1803 and expand_complex_division call-sites.
1805 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1808 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1809 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1810 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1811 the highest significant bit of the shift count mask is clear. In
1812 check whether and[sq]i3 is needed verify that all significant bits
1813 of the shift count other than the highest are set.
1815 2018-05-02 Tom de Vries <tom@codesourcery.com>
1818 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1819 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1820 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1821 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1822 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1823 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1824 __builtin_goacc_parlevel_size.
1826 2018-05-02 Richard Biener <rguenther@suse.de>
1828 PR tree-optimization/85597
1829 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1830 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1833 2018-05-02 Tom de Vries <tom@codesourcery.com>
1836 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1837 dump files): Add ltrans-tree.
1839 2018-05-02 Tom de Vries <tom@codesourcery.com>
1842 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1843 dump files): Add wpa-ipa.
1845 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1847 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1848 powerpc*-*-linux*paired* target.
1849 * config/rs6000/750cl.h: Delete.
1850 * config/rs6000/paired.h: Delete.
1851 * config/rs6000/paired.md: Delete.
1852 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1854 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1855 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1856 comment. Remove paired float support.
1857 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1858 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1860 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1861 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1863 * config/rs6000/rs6000.c: Remove paired float support.
1864 (paired_expand_vector_init, paired_expand_vector_move,
1865 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1866 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1867 paired_expand_builtin, paired_expand_predicate_builtin,
1868 paired_init_builtins): Delete.
1869 * config/rs6000/rs6000.h: Remove paired float support.
1870 * config/rs6000/rs6000.md: Remove paired float support.
1871 (move_from_CR_ov_bit): Delete.
1872 * config/rs6000/rs6000.opt (mpaired): Delete.
1873 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1874 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1876 2018-05-02 Richard Biener <rguenther@suse.de>
1879 * gimplify.c (gimplify_save_expr): When in SSA form allow
1880 SAVE_EXPRs to compute to SSA vars.
1882 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1885 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1886 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1887 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1888 clobber operands[2], instead use a new pseudo. Formatting fixes.
1890 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1892 PR tree-optimization/85586
1893 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1894 exit early for statements in the same group if the accesses are
1897 2018-05-02 Tom de Vries <tom@codesourcery.com>
1900 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1903 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1905 PR tree-optimization/85143
1906 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1908 2018-05-01 Tom de Vries <tom@codesourcery.com>
1911 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1912 not found" error message.
1914 2018-05-01 Tom de Vries <tom@codesourcery.com>
1917 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1918 * vec.c (test_ordered_remove_if): New function.
1919 (vec_c_tests): Call test_ordered_remove_if.
1920 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1921 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1922 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1923 VEC_ORDERED_REMOVE_IF.
1925 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1927 PR tree-optimization/82665
1928 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1929 pointer subtraction where arguments come from a memchr call.
1931 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1933 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1934 --push-state --as-needed and --pop-state instead of --as-needed and
1935 --no-as-needed if ld supports it.
1936 * configure: Regenerated.
1939 * doc/install.texi2html: Replace _002d with - and _002a with * in
1940 generated html files using sed.
1942 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1945 * gcc-rich-location.c (blank_line_before_p): New function.
1946 (use_new_line): New function.
1947 (gcc_rich_location::add_fixit_insert_formatted): New function.
1948 * gcc-rich-location.h
1949 (gcc_rich_location::add_fixit_insert_formatted): New function.
1951 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1953 * selftest.c (assert_streq): Rename "expected" and "actual" to
1954 "val1" and "val2". Extend NULL-handling to cover both inputs
1955 symmetrically, while still requiring both to be non-NULL for a pass.
1956 * selftest.h (assert_streq): Rename "expected" and "actual" to
1958 (ASSERT_EQ): Likewise.
1959 (ASSERT_EQ_AT): Likewise.
1960 (ASSERT_KNOWN_EQ): Likewise.
1961 (ASSERT_KNOWN_EQ_AT): Likewise.
1962 (ASSERT_NE): Likewise.
1963 (ASSERT_MAYBE_NE): Likewise.
1964 (ASSERT_MAYBE_NE_AT): Likewise.
1965 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1966 the assertion to pass.
1967 (ASSERT_STREQ_AT): Likewise.
1969 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1971 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1972 interaction with -pie.
1974 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1976 * selftest.h: Fix alphabetization of per-source-file selftest
1979 2018-04-30 Jason Merrill <jason@redhat.com>
1981 PR c++/61982 - dead stores to destroyed objects.
1982 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1985 2018-04-30 Jason Merrill <jason@redhat.com>
1987 * tree.c (build_clobber): New.
1988 * tree.h: Declare it.
1989 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1991 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1993 * diagnostic-show-locus.c (layout::layout): Update for
1994 location_get_source_line returning a char_span.
1995 (struct char_span): Move to input.h.
1996 (struct correction): Update for fields in char_span becoming
1998 (struct source_line): Update for location_get_source_line
1999 returning a char_span.
2000 (layout::print_line): Likewise.
2001 * edit-context.c (edited_file::print_content): Likewise.
2002 (edited_file::print_diff_hunk): Likewise.
2003 (edited_file::print_run_of_changed_lines): Likewise.
2004 (edited_file::get_num_lines): Likewise.
2005 (edited_line::edited_line): Likewise.
2006 * final.c (asm_show_source): Likewise.
2007 * input.c (location_get_source_line): Convert return type
2008 from const char * to char_span, losing the final "line_len"
2010 (dump_location_info): Update for the above.
2011 (get_substring_ranges_for_loc): Likewise. Use a char_span
2012 when handling the literal within the line.
2013 (test_reading_source_line): Update for location_get_source_line
2014 returning a char_span.
2015 * input.h (class char_span): Move here from
2016 diagnostic-show-locus.c, converting from a struct to a class.
2017 Make data members private.
2018 (char_span::operator bool): New.
2019 (char_span::length): New.
2020 (char_span::get_buffer): New.
2021 (char_span::operator[]): New.
2022 (char_span::subspan): Make const.
2023 (char_span::xstrdup): New.
2024 (location_get_source_line): Convert return type from const char *
2025 to char_span, losing the final "line_size" param.
2027 2018-04-30 Jan Hubicka <jh@suse.cz>
2029 * lto-wrapper.c (ltrans_priorities): New static var.
2030 (cmp_priority): New.
2031 (run_gcc): Read priorities and if doing parallel build order
2032 the Makefile by them.
2034 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2036 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2038 2018-04-30 Richard Biener <rguenther@suse.de>
2040 * tree-cfg.c (verify_address): Remove base argument, add
2041 flag whether to check TREE_ADDRESSABLE and do that.
2042 (verify_expr): Remove.
2043 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2044 (verify_gimple_assign_single): Likewise.
2045 (verify_gimple_switch): Likewise.
2046 (verify_expr_location_1): Dereference tp once. Add (disabled)
2047 piece from verify_expr.
2048 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2050 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2052 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2054 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2056 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2057 (small_data_pattern): Likewise.
2058 (arc_rewrite_small_data): Likewise.
2059 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2060 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2061 (get_symbol_alignment): New function.
2062 (legitimate_small_data_address_p): Likewise.
2063 (legitimate_scaled_address): Update, call
2064 legitimate_small_data_address_p.
2065 (output_sdata): New static variable.
2066 (arc_print_operand): Update how we handle small data operands.
2067 (arc_print_operand_address): Likewise.
2068 (arc_legitimate_address_p): Update, use
2069 legitimate_small_data_address_p.
2070 (arc_rewrite_small_data_p): Remove.
2071 (arc_rewrite_small_data_1): Likewise.
2072 (arc_rewrite_small_data): Likewise.
2073 (small_data_pattern): Likewise.
2074 (compact_sda_memory_operand): Update to use
2075 legitimate_small_data_address_p and get_symbol_alignment.
2076 (prepare_move_operands): Don't rewite sdata pattern.
2077 (prepare_extend_operands): Remove.
2078 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2080 (zero_extendqisi2): Likewise.
2081 (zero_extendhisi2): Likewise.
2082 (extendqihi2): Likewise.
2083 (extendqisi2): Likewise.
2084 (extendhisi2): Likewise.
2088 * config/arc/constraints.md (Usd): Change it to memory constraint.
2090 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2092 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2093 as source of std instructions.
2094 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2095 allow 6-bit constants as source for store instructions.
2096 (movdi_insn): Update instruction pattern to allow 6-bit constants
2097 as source for store instructions.
2099 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2101 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2103 2018-04-30 Nathan Sidwell <nathan@acm.org>
2104 Sandra Loosemore <sandra@codesourcery.com>
2106 * dumpfile.c (dump_open): Allow '-' for stdout.
2107 * doc/invoke.texi (Developer Options): Document dump filename
2108 determination early. Document stdin/stdout selection.
2110 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2112 Microblaze Target: PIC data text relative
2114 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2115 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2117 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2118 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2119 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2120 New addressing mode for data-text relative position indepenedent code.
2121 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2122 'ADDRESS_SYMBOLIC_TXT_REL'.
2123 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2124 (microblaze_legitimate_pic_operand): Exclude function calls from
2125 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2126 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2128 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2129 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2130 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2131 for 'address + offset'.
2132 (microblaze_expand_prologue): Add new function prologue call for
2134 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2135 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2136 table in case of TARGET_PIC_DATA_TEXT_REL.
2137 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2138 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2139 Add new macros 'UNSPEC_TEXT',
2140 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2141 + exclude function calls from 'UNSPEC_PLT' in case of data text
2143 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2144 new target hook for generating address diff vector tables in case of
2146 * doc/tm.texi : Regenerate.
2147 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2148 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2149 of addr diff vector generation.
2150 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2151 target hook definition.
2152 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2153 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2154 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2155 option for data text relative.
2157 2018-04-30 Richard Biener <rguenther@suse.de>
2159 * tree-chrec.h (evolution_function_is_constant_p): Remove
2161 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2163 2018-04-30 Richard Biener <rguenther@suse.de>
2166 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2168 2018-04-30 Richard Biener <rguenther@suse.de>
2170 PR tree-optimization/28364
2171 PR tree-optimization/85275
2172 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2173 copying first exit test.
2175 2018-04-28 Mark Wielaard <mark@klomp.org>
2177 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2179 (dwarf_AT): Handle DW_AT_addr_base.
2180 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2182 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2185 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2186 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2187 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2188 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2190 2018-04-28 Richard Biener <rguenther@suse.de>
2192 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2193 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2194 to reflect use. Only add interesting stmts.
2196 2018-04-27 Martin Jambor <mjambor@suse.cz>
2199 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2200 the jump function allows for passing through aggregate values.
2202 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2204 * input.h (in_system_header_at): Convert from macro to inline
2206 (from_macro_expansion_at): Likewise.
2207 (from_macro_definition_at): Likewise.
2209 2018-04-27 Jeff Law <law@redhat.com>
2211 * config.gcc: Mark tile* targets as deprecated/obsolete.
2213 2018-04-27 Richard Biener <rguenther@suse.de>
2215 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2218 2018-04-27 Richard Biener <rguenther@suse.de>
2220 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2222 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2224 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2225 with Yd constraint. Set "preferred_for_speed" attribute from
2226 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2228 (*movdi_internal): Ditto.
2229 (movti_interunit splitters): Remove
2230 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2231 (movdi_interunit splitters): Ditto.
2232 * config/i386/constraints.md (Ye): Remove.
2233 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2235 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2238 * config/aarch64/constraints.md (Usg): Limit to 31.
2241 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2243 PR tree-optimization/85529
2244 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2245 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2246 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2247 zero extension or masking of the MSB bit.
2248 (optimize_range_tests): Add FIRST_BB argument, pass it through
2249 to optimize_range_tests_var_bound.
2250 (maybe_optimize_range_tests, reassociate_bb): Adjust
2251 optimize_range_tests callers.
2253 2018-04-26 Richard Biener <rguenther@suse.de>
2254 Jakub Jelinek <jakub@redhat.com>
2256 * cgraph.h (symbol_table): Just declare debug method here.
2257 * symtab.c (symbol_table::debug): Define.
2259 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2261 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2263 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2265 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2266 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2267 (*movdi_internal): Substitute Yi and Yj constraint with x
2268 and Ym and Yn constraint with y constraint. Update "isa"
2269 attribute and set "preferred_for_speed" attribute from
2270 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2271 (*movsi_internal): Ditto.
2272 (*movdf_internal): Ditto.
2273 (*movsf_internal): Ditto.
2274 (*zero_extendsidi2): Ditto.
2275 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2276 (sse2_loadld): Ditto.
2277 (*vec_extract<ssevecmodelower>_0): Ditto.
2278 (*vec_extractv4si_0_zext_sse4): Ditto.
2279 (vec_concatv2di): Ditto.
2280 (*vec_dup<mode>): Ditto.
2281 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2282 * config/i386/constraints.md (Yi): Remove.
2287 2018-04-26 Nathan Sidwell <nathan@acm.org>
2289 * dumpfile.c (dump_open): New.
2290 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2291 (dump_finish): Detect stdio/stderr by value not name.
2293 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2295 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2297 2018-04-26 Tom de Vries <tom@codesourcery.com>
2300 * config/nvptx/nvptx.c (verify_neutering_jumps)
2301 (verify_neutering_labels): New function
2302 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2304 2018-04-26 Tom de Vries <tom@codesourcery.com>
2307 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2308 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2311 2018-04-26 Richard Biener <rguenther@suse.de>
2312 Tom de Vries <tom@codesourcery.com>
2315 * lto-streamer-out.c (output_function): Fixup loops if required to match
2316 discovery done in the reader.
2318 2018-04-26 Richard Biener <rguenther@suse.de>
2320 PR tree-optimization/85116
2321 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2322 have a loop exit from the single latch predecessor. Remove
2323 case of header with just condition.
2324 (ch_base::copy_headers): Exclude infinite loops from any
2326 (pass_ch::execute): Record exits.
2328 2018-04-26 Richard Biener <rguenther@suse.de>
2330 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2331 prologue cost vector and pass it to vect_get_load_cost.
2332 (vect_get_peeling_costs_all_drs): Likewise.
2333 (vect_peeling_hash_get_lowest_cost): Likewise.
2334 (vect_enhance_data_refs_alignment): Likewise.
2336 2018-04-26 Richard Biener <rguenther@suse.de>
2339 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2340 checking of integer<->pointer conversions.
2341 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2342 sign-/zero-extending pointer types.
2343 (expand_omp_for_static_chunk): Likewise.
2345 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2346 Jean Lee <xiaoyur347@gmail.com>
2348 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2349 (TARGET_ASAN_SHADOW_OFFSET): Define.
2350 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2351 true for -fsanitize=address.
2353 2018-04-25 Mark Wielaard <mark@klomp.org>
2355 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2358 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2360 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2361 than "alu", remove explicit "memory" and "imm_disp" attributes.
2362 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2365 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2366 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2367 gen_lowpart_no_emit.
2369 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2372 * config/i386/i386.c (ix86_expand_builtin): Change memory
2373 operand to XI, extend p0 to Pmode.
2374 * config/i386/i386.md: Change unspec volatile and operand
2375 1 mode to XI, change operand 0 mode to P.
2377 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2379 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2380 GET_MODE_MASK before any checking.
2381 (nds32_can_use_bset_p): Likewise.
2382 (nds32_can_use_btgl_p): Likewise.
2384 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2386 * config/nds32/nds32-doubleword.md: New define_split pattern for
2387 illegal register number.
2389 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2391 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2393 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2395 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2397 2018-04-25 Richard Biener <rguenther@suse.de>
2399 * lto-streamer.h (LTO_major_version): Bump to 8.
2401 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2403 * BASE-VER: Set to 9.0.0.
2405 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2407 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2408 in __abskf2 and __powikf2.
2410 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2413 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2414 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2415 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2416 Use the above on operand 2. Reindent.
2417 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2419 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2422 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2424 * config/i386/i386.opt (mcet): Removed.
2425 * doc/install.texi: Remove -mcet documentation.
2426 * doc/invoke.texi: Likewise.
2428 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2431 * doc/install.texi: Remove -mcet from bootstrap-cet.
2433 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2436 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2437 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2441 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2442 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2443 containing a CONST_VECTOR.
2445 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2447 * doc/install.texi: Update newlib dependency for nvptx.
2449 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2452 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2453 instead of INTVAL when shifting x left.
2455 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2457 PR tree-optimization/85478
2458 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2459 vect_grouped_store_supported for single element vectors.
2461 2018-04-24 Richard Biener <rguenther@suse.de>
2464 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2465 load cost increase to the case of non-constant step.
2467 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2470 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2471 destination if any_malformed_asm.
2473 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2476 * expr.c (store_field): In the bitfield case, if the value comes from
2477 a function call and is returned in registers by means of a PARALLEL,
2478 do not change the mode of the temporary unless BLKmode and VOIDmode.
2480 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2482 PR rtl-optimization/85423
2483 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2484 dependencies to debug insns when the previous insn is non-debug.
2486 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2488 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2489 enums into a single definition.
2490 (fls): Fix predicates and printing.
2493 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2495 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2496 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2497 and short u6 immediate.
2498 (check_if_valid_sleep_operand): Remove.
2499 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2501 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2503 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2504 flag_always_save_lp condition.
2505 * config/nds32/nds32.opt (malways-save-lp): New option.
2507 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2509 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2510 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2511 * config/nds32/nds32.h
2512 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2513 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2515 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2517 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2518 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2520 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2521 Chung-Ju Wu <jasonwucj@gmail.com>
2523 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2525 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2526 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2528 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2530 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2532 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2534 * config/nds32/nds32-protos.h (nds32_data_alignment,
2535 nds32_local_alignment): Declare.
2536 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2537 nds32_local_alignment): New functions.
2538 (TARGET_CONSTANT_ALIGNMENT): Define.
2539 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2541 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2543 * config/nds32/nds32.c
2544 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2545 (TARGET_MODES_TIEABLE_P): Likewise.
2547 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2549 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2552 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2553 Chung-Ju Wu <jasonwucj@gmail.com>
2555 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2556 for unaligned access.
2557 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2559 * config/nds32/nds32-intrinsic.md: Likewise.
2560 * config/nds32/nds32_intrinsic.h: Likewise.
2561 * config/nds32/nds32.h (nds32_builtins): Likewise.
2562 * config/nds32/nds32.opt (munaligned-access): New option.
2563 * config/nds32/nds32.c (nds32_asm_file_start): Display
2564 flag_unaligned_access status.
2566 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2568 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2569 -mno-relax is present.
2570 * config/riscv/linux.h (LINK_SPEC): Ditto.
2572 2018-04-20 Martin Sebor <msebor@redhat.com>
2575 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2577 (gimple_fold_builtin_stxcpy_chk): Same.
2578 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2580 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2583 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2584 __powikf2 when long double is IEEE 128-bit.
2586 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2588 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2589 step to make sure stack always aligned.
2591 2018-04-20 Carl Love <cel@us.ibm.com>
2594 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2595 size check for arg0.
2597 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2598 Tom de Vries <tom@codesourcery.com>
2601 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2602 Emit insns for calls too.
2603 (nvptx_find_par): Always look for worker-level predecessor insn.
2604 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2606 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2607 (nvptx_process_pars): Propagate frames for calls.
2609 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2612 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2614 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2615 (ix86_handle_option): Don't handle OPT_mibt.
2616 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2618 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2620 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2621 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2622 (ix86_target_macros): Define __CET__ with flag_cf_protection
2623 for -fcf-protection.
2624 * config/i386/i386.c (isa2_opts): Remove -mibt.
2625 * config/i386/i386.h (TARGET_IBT): Removed.
2626 (TARGET_IBT_P): Likewise.
2627 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2628 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2629 * config/i386/i386.opt (mcet): Update help message.
2632 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2633 -mcet as an alias for -mshstk.
2635 2018-04-20 Richard Biener <rguenther@suse.de>
2638 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2639 complexity by forcing a single use of the multiply operand.
2641 2018-04-20 Martin Jambor <mjambor@suse.cz>
2644 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2645 recursion dependency to only apply to non-clones.
2647 2018-04-20 Martin Jambor <mjambor@suse.cz>
2650 * ipa-cp.c (create_specialized_node): Check that clones of
2651 self-recursive edges exist during IPA-CP.
2653 2018-04-19 Toon Moene <toon@moene.org>
2655 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2658 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2660 PR tree-optimization/85467
2661 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2662 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2663 VECTOR_CST element to type.
2665 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2668 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2669 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2670 (builtin_longjmp): Likewise.
2671 (save_stack_nonlocal): New pattern.
2672 (restore_stack_nonlocal): Likewise.
2674 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2677 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2678 Replace ASM_OUTPUT_LABEL with fprintf.
2680 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2683 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2684 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2685 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2686 define __IBT__ and __SHSTK__ for -fcf-protection.
2687 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2689 (ix86_trampoline_init): Likewise.
2690 (x86_output_mi_thunk): Likewise.
2691 (ix86_notrack_prefixed_insn_p): Likewise.
2692 (ix86_option_override_internal): Don't disallow -fcf-protection.
2693 * config/i386/i386.md (rdssp<mode>): Also enable for
2695 (incssp<mode>): Likewise.
2696 (nop_endbr): Likewise.
2697 * config/i386/i386.opt (mcet): Change help message to built-in
2701 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2702 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2703 enable CET built-in functions.
2705 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2707 * common/config/i386/i386-common.c
2708 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2709 OPTION_MASK_ISA_MOVDIRI_UNSET,
2710 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2711 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2712 * config.gcc (movdirintrin.h): New header.
2713 * config/i386/cpuid.h (bit_MOVDIRI,
2714 bit_MOVDIR64B): New bits.
2715 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2717 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2718 (VOID, PVOID, PCVOID)): New function types.
2719 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2720 __builtin_ia32_directstoreu_u64,
2721 __builtin_ia32_movdir64b): New builtins.
2722 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2723 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2725 (ix86_valid_target_attribute_inner_p): Ditto.
2726 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2727 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2728 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2729 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2730 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2731 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2732 (movdiri<mode>, movdir64b_<mode>): New.
2733 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2734 * config/i386/immintrin.h: Include movdirintrin.h.
2735 * config/i386/movdirintrin.h: New file.
2736 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2738 2018-04-19 Richard Biener <rguenther@suse.de>
2741 * cfg.c (clear_bb_flags): When loop state says we have
2742 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2744 2018-04-19 Richard Biener <rguenther@suse.de>
2746 PR tree-optimization/84737
2747 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2748 copying restrict info.
2749 (vect_setup_realignment): Use it.
2750 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2751 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2752 the first DR to all generated stores.
2753 (vectorizable_load): Likewise for loads.
2755 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2757 PR tree-optimization/85446
2758 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2759 the integral and pointer types to have the same precision.
2761 * doc/install.texi: Document --disable-cet being the default and
2764 2018-04-18 Martin Liska <mliska@suse.cz>
2766 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2769 2018-04-18 Martin Liska <mliska@suse.cz>
2772 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2775 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2776 arguments if they are comparable.
2778 2018-04-18 Martin Liska <mliska@suse.cz>
2781 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2784 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2787 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2790 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2791 ENDBR after calling __morestack.
2793 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2796 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2797 by using gcc_base_ver to generate a gcc_driver_version, and use
2798 it when generating GCC_DRIVER_NAME.
2799 * configure: Regenerate.
2801 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2804 * config.gcc: Obsolete powerpc*-*-*spe*.
2806 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2809 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2810 (stabstr_D): Change type of unum from unsigned int to
2811 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2814 2018-04-17 Jim Wilson <jimw@sifive.com>
2817 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2818 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2819 Set arg_pointer_offset after using pretend_args_size.
2821 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2823 PR rtl-optimization/85431
2824 * dse.c (record_store): Ignore zero width stores.
2827 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2828 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2829 __builtin_stack_restore rather than after it.
2830 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2831 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2832 argument instead of virtual_dynamic_stack_rtx.
2834 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2836 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2838 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2839 Add note to error message to explain internal mapping of overloaded
2840 built-in function name to non-overloaded built-in function name.
2841 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2844 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2847 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2848 where the inputs overlap with the output.
2850 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2852 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2853 (=v, v) alternative and explicit "memory" attribute.
2854 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2855 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2857 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2858 "sselog1" type instead of "sselog".
2859 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2860 "sselog". Remove explicit "memory" attribute.
2861 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2862 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2864 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2865 "sselog1" type instead of "sselog". Remove explicit "memory"
2867 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2868 use "sselog1" type instead of "sselog". Remove explicit "memory"
2870 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2871 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2873 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2874 "sselog1" type instead of "sselog". Remove explicit "memory"
2876 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2877 use "sselog1" type instead of "sselog". Remove explicit "memory"
2881 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2884 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2887 2018-04-17 Martin Jambor <mjambor@suse.cz>
2890 * ipa-cp.c (create_specialized_node): Call
2891 expand_all_artificial_thunks if necessary.
2893 2018-04-17 Martin Liska <mliska@suse.cz>
2896 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2897 in message, remote space in between '_G' and '('.
2899 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2902 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2903 avx512f_vmcmp<mode>3<round_saeonly_name>,
2904 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2905 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2906 avx512f_rndscale<mode><round_saeonly_name>,
2907 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2908 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2909 Use %<iptr>2 instead of %2 for -masm=intel.
2910 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2911 avx512f_vcvttss2usi<round_saeonly_name>,
2912 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2914 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2915 avx512f_vcvttsd2usi<round_saeonly_name>,
2916 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2917 Use %q1 instead of %1 for -masm=intel.
2918 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2919 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2920 of %3 for -masm=intel.
2921 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2923 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2925 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2927 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2928 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2930 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2932 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2933 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2934 %g1 and one with %0 and %1.
2935 (avx512er_vmrcp28<mode><round_saeonly_name>,
2936 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2938 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2939 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2940 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2941 of %0 and %{%4%} for -masm=intel.
2942 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2943 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2944 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2945 order of %0 and %{%5%}%{z%} for -masm=intel.
2947 2018-04-17 Jan Hubicka <jh@suse.cz>
2950 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2952 2018-04-17 Martin Liska <mliska@suse.cz>
2955 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2956 for target_clone error message. Make default implementation
2957 clone to be a local declaration.
2958 (separate_attrs): Add new argument and check for an empty
2960 (expand_target_clones): Handle it.
2961 (ipa_target_clone): Make redirection just for target_clones
2964 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2965 Tom de Vries <tom@codesourcery.com>
2968 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2969 tiled basic blocks without omp continue statements.
2971 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2974 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2975 vec_extract expression as having side effects to make sure it gets
2978 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2981 * config/i386/i386.c (get_builtin_code_for_version): Check
2984 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2987 * gcc/config.gcc: Support "skylake".
2988 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2990 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2991 (processor_target_table): Add "skylake".
2992 (ix86_option_override_internal): Add "skylake".
2993 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2994 PROCESSOR_CANNONLAKE.
2995 (get_builtin_code_for_version): Fix priority for
2996 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2997 PROCESSOR_SKYLAKE-AVX512.
2998 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2999 (processor_type): Add PROCESSOR_SKYLAKE.
3001 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3002 Jason Merrill <jason@redhat.com>
3005 * convert.c (convert_to_integer_1): Use direct recursion for
3006 enumeral types and types with a precision less than the number
3007 of bits in their mode.
3009 2018-04-16 Julia Koval <julia.koval@intel.com>
3012 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3013 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3015 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3018 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3019 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3020 and -mno-direct-move.
3022 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3025 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3026 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3027 Ensure negative shifts result in {0}.
3029 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3031 PR rtl-optimization/79916
3032 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3033 regs (if any) to define how to gnerate SD moves when LRA is in
3036 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3038 PR rtl-optimization/85393
3039 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3040 * except.c (expand_dw2_landing_pad_for_region): Make static.
3041 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3042 a label and unconditional jump to old_bb, rather than
3043 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3046 PR rtl-optimization/85376
3047 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3048 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3049 instead of a specific value.
3051 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3052 Bin Cheng <bin.cheng@arm.com>
3054 PR tree-optimization/82965
3055 PR tree-optimization/83991
3056 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3057 by_profile_only parameter.
3058 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3059 information if the loop was predicted to iterate too many times.
3060 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3062 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3065 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3068 2018-04-13 Martin Liska <mliska@suse.cz>
3069 Jakub Jelinek <jakub@redhat.com>
3072 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3073 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3074 * builtins.c (expand_builtin_memory_copy_args): Use
3075 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3076 handle dest_addr == pc_rtx.
3078 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3081 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3082 asked to not generate direct moves.
3083 (fix_trunc<mode>si2_stfiwx): Similar.
3084 (fix_trunc<mode>si2_internal): Similar.
3086 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3089 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3090 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3091 lookup if dest in some wider mode is known to be const0_rtx and
3092 if so, record permanent equivalence for it to be ZERO_EXTEND of
3093 the narrower mode destination.
3095 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3097 * lto-streamer-out.c (output_function): Revert 259346.
3098 * omp-expand.c (expand_oacc_for): Likewise.
3100 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3102 PR rtl-optimization/85354
3103 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3104 * sel-sched.c (sel_global_init): ... here.
3106 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3109 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3110 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3111 mode for PE-COFF targets.
3112 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3113 (i386_pe_asm_lto_end): Likewise.
3114 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3115 (TARGET_ASM_LTO_END): Likewise.
3116 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3117 (i386_pe_asm_lto_start): New function.
3118 (i386_pe_asm_lto_end): Likewise.
3120 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3121 Richard Biener <rguenther@suse.de>
3124 * lto-streamer-out.c (output_function): Fix CFG loop state before
3126 * omp-expand.c (expand_oacc_for): Handle calls to internal
3127 functions like regular functions.
3129 2018-04-12 Richard Biener <rguenther@suse.de>
3132 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3133 for the early LTO debug to properly generate references to it
3134 during DIE emission. Do not re-use that for the skeleton for
3136 (dwarf2out_early_finish): Likewise.
3138 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3141 * config/i386/sse.md
3142 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3143 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3144 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3145 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3146 and output is a reg, avoid creating invalid lowpart subreg, but
3147 instead split into a 512-bit move. Don't split if not AVX512VL,
3148 input is xmm16+ reg and output is a mem.
3149 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3150 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3151 xmm16+ reg and output is a mem.
3153 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3155 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3156 also for flag_dwarf2_cfi_asm.
3158 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3160 PR rtl-optimization/85342
3161 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3162 a bool scalar var inside of the loop instead. Don't try to update
3163 recog_data.operand after failed apply_change_group.
3165 2018-04-12 Tom de Vries <tom@codesourcery.com>
3168 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3169 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3170 array with flexible array member as array without given dimension.
3171 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3172 argument for undefined param to true.
3174 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3177 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3178 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3179 from PowerPC section.
3180 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3181 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3182 help text that is too long.
3183 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3184 help text that is too long.
3185 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3186 help text that is too long.
3188 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3190 * config/alpha/alpha.md (stack_probe_internal): Rename
3191 from "probe_stack". Update all callers.
3193 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3195 PR rtl-optimization/84566
3196 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3197 sched_macro_fuse_insns.
3199 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3202 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3203 (compute_block_dependences): ... from here.
3205 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3207 PR tree-optimization/85331
3208 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3209 from int to HOST_WIDE_INT.
3211 2018-04-11 Martin Jambor <mjambor@suse.cz>
3214 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3215 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3216 not the same as the source val.
3217 (cgraph_edge_brings_value_p): New parameter.
3218 (gather_edges_for_value): Pass destination value to
3219 cgraph_edge_brings_value_p.
3220 (perhaps_add_new_callers): Likewise.
3221 (get_info_about_necessary_edges): Likewise and exclude values brought
3222 only by self-recursive edges.
3223 (create_specialized_node): Redirect only clones of self-calling edges.
3224 (+self_recursive_pass_through_p): New function.
3225 (find_more_scalar_values_for_callers_subset): Use it.
3226 (find_aggregate_values_for_callers_subset): Likewise.
3227 (known_aggs_to_agg_replacement_list): Removed.
3228 (decide_whether_version_node): Re-calculate known constants for all
3229 remaining context clones.
3231 2018-04-11 Richard Biener <rguenther@suse.de>
3234 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3235 from early DWARF output.
3236 (dwarf2out_early_finish): Output line info unconditionally into
3237 early DWARF and add reference to it.
3239 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3242 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3243 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3244 other than V2DFmode using iptr mode attribute.
3245 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3247 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3249 PR rtl-optimization/84659
3250 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3252 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3255 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3257 (output_loc_list): Pass address of a dummy size variable even in the
3258 locview handling loop.
3259 (index_location_lists): Add comment on why skip_loc_list_entry can't
3262 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3265 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3268 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3271 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3272 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3273 and -mstring-compare-inline-limit.
3275 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3278 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3279 for stack clash protection in a register whenever we need it to be in
3282 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3284 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3285 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3287 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3290 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3292 (mlong-double-): Ditto.
3293 * config/rs6000/sysv4.opt (msdata=): Ditto.
3294 (mtls-size=): Ditto.
3296 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3298 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3299 erroneous entries for
3300 "vector int vec_ldl (int, long int *)", and
3301 "vector unsigned int vec_ldl (int, unsigned long int *)".
3302 Add comments and entries for
3303 "vector bool char vec_ldl (int, bool char *)",
3304 "vector bool short vec_ldl (int, bool short *)",
3305 "vector bool int vec_ldl (int, bool int *)",
3306 "vector bool long long vec_ldl (int, bool long long *)",
3307 "vector pixel vec_ldl (int, pixel *)",
3308 "vector long long vec_ldl (int, long long *)",
3309 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3310 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3311 type tree bool_long_long_type_node and correct definition of
3312 bool_V2DI_type_node to make reference to this new type tree.
3313 (rs6000_mangle_type): Replace erroneous reference to
3314 bool_long_type_node with bool_long_long_type_node.
3315 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3316 comments to emphasize sign distinctions for char and int types and
3317 replace RS6000_BTI_bool_long constant with
3318 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3319 use of RS6000_BTI_pixel.
3320 (bool_long_type_node): Remove this macro definition.
3321 (bool_long_long_type_node): New macro definition
3323 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3325 PR rtl-optimization/85300
3326 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3327 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3328 simplify_unary_operation fails.
3330 2018-04-10 Martin Liska <mliska@suse.cz>
3332 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3333 cgraph_edge and ipa_ref.
3335 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3339 * config/i386/sse.md
3340 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3341 computation of the VEC_MERGE selector from mask.
3342 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3343 Fix decoding of the VEC_MERGE selector into mask.
3345 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3347 PR tree-optimization/85286
3348 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3350 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3352 * final.c (final_1): Set insn_last_address as well as
3353 insn_current_address.
3355 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3358 * explow.c (emit_stack_probe): Call validize_mem on memory location
3359 before passing it to gen_probe_stack. Create address operand and
3360 legitimize it for the probe_stack_address case.
3362 2018-04-09 Jan Hubicka <jh@suse.cz>
3365 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3366 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3367 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3369 (free_lang_data): Rebuild type inheritance graph.
3371 2018-04-09 Martin Sebor <msebor@redhat.com>
3373 * invoke.texi (-finline-small-functions): Mention other optimization
3375 (-findirect-inlining, -fpartial-inlining): Same.
3376 (-finline-functions-called-once): Same.
3377 (-freorder-blocks-and-partition): Same.
3379 2018-04-09 Jan Hubicka <jh@suse.cz>
3382 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3383 jumps; choose last target that matches the criteria (i.e.
3384 no partition changes for non-crossing jumps).
3385 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3386 support for redirecting crossing jumps to non-crossing.
3388 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3390 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3391 also for naked functions.
3393 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3395 * config/arc/arc.md (add_shift): New pattern.
3396 (add_shift2): Likewise.
3397 (sub_shift): Likewise.
3398 (sub_shift_cmp0_noout): Likewise.
3399 (compare_si_ashiftsi): Likewise.
3400 (xbfu_cmp0_noout): New combine pattern.
3401 (xbfu_cmp0"): Likewise.
3402 (movsi_set_cc_insn): Place the predicable variant first.
3403 (commutative_binary_cmp0_noout): Remove clobber.
3404 (commutative_binary_cmp0): New pattern.
3405 (noncommutative_binary_cmp0): Likewise.
3406 (noncommutative_binary_cmp0_noout): Likewise.
3407 (noncommutative_binary_comparison_result_used): Removed.
3408 (rsub_cmp0): New pattern.
3409 (rsub_cmp0_noout): Likewise.
3410 (extzvsi): Changed, keep only meaningful variants.
3411 (SQH, SEZ): New iterators.
3412 (SQH_postfix): New mode attribute.
3413 (SEZ_prefix): New code attribute.
3414 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3415 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3416 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3418 (noncommutative_operator): Check the availability of barrel
3421 2018-04-09 Richard Biener <rguenther@suse.de>
3423 PR tree-optimization/85284
3424 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3425 Only use the niter constraining form of simple_iv when the exit
3428 2018-04-09 Tom de Vries <tom@codesourcery.com>
3431 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3432 (define_expand "*memory_barrier"): New define_expand.
3433 (define_insn "memory_barrier"): New insn.
3435 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3437 PR rtl-optimization/80463
3438 PR rtl-optimization/83972
3439 PR rtl-optimization/83480
3441 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3442 correct producer for the insn.
3443 (tidy_control_flow): Fixup seqnos in case of debug insns.
3445 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3447 PR rtl-optimization/83913
3449 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3450 different sched-times when merging exprs.
3452 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3454 PR rtl-optimization/83962
3456 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3457 tidy_fallthru_edge and tidy_control_flow.
3459 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3461 PR rtl-optimization/83530
3463 * sel-sched.c (force_next_insn): New global variable.
3464 (remove_insn_for_debug): When force_next_insn is true, also leave only
3465 next insn in the ready list.
3466 (sel_sched_region): When the region wasn't scheduled, make another pass
3467 over it with force_next_insn set to 1.
3469 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3471 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3473 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3474 for interrupt control.
3475 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3476 functions for interrupt control.
3477 * config/nds32/nds32-intrinsic.md: Likewise.
3478 * config/nds32/nds32_intrinsic.h: Likewise.
3479 * config/nds32/nds32.h (nds32_builtins): Likewise.
3481 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3483 * config/nds32/nds32.c (nds32_init_machine_status,
3484 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3485 strict_aligned_p field.
3486 (nds32_expand_to_rtl_hook): New function.
3487 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3488 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3490 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3491 Chung-Ju Wu <jasonwucj@gmail.com>
3493 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3494 * config/nds32/nds32-n7.md: New file.
3495 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3496 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3498 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3499 * config/nds32/nds32.md (pipeline_model): Add n7.
3500 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3501 * config/nds32/pipelines.md: Include n7 settings.
3503 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3504 Chung-Ju Wu <jasonwucj@gmail.com>
3506 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3507 * config/nds32/nds32-e8.md: New file.
3508 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3509 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3511 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3512 * config/nds32/nds32.md (pipeline_model): Add e8.
3513 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3514 * config/nds32/pipelines.md: Include e8 settings.
3516 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3517 Chung-Ju Wu <jasonwucj@gmail.com>
3519 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3520 * config/nds32/nds32-n8.md: New file.
3521 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3522 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3524 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3525 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3526 * config/nds32/nds32.md (pipeline_model): Add n8.
3527 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3528 * config/nds32/pipelines.md: Include n8 settings.
3530 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3531 Chung-Ju Wu <jasonwucj@gmail.com>
3533 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3534 * config/nds32/nds32-n9-2r1w.md: New file.
3535 * config/nds32/nds32-n9-3r2w.md: New file.
3536 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3537 nds32_register_ports): New or modify for cpu n9.
3538 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3540 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3541 * config/nds32/nds32-utils.c: New file.
3542 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3543 TARGET_MUL_SLOW): Define.
3544 * config/nds32/nds32.md (pipeline_model): New attribute.
3545 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3546 New options that support cpu n9.
3547 * config/nds32/pipelines.md: Include n9 settings.
3548 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3550 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3552 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3553 information if necessary.
3554 (output_cond_branch_compare_zero): Likewise.
3555 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3556 (nds32_target_alignment): Refine for alignment.
3557 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3558 (FUNCTION_BOUNDARY): Modify.
3559 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3561 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3563 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3565 * config/nds32/constants.md (unspec_volatile_element): Add values for
3566 TLB operation and data prefetch.
3567 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3568 functions for TLB operation and data prefetch.
3569 * config/nds32/nds32-intrinsic.md: Likewise.
3570 * config/nds32/nds32_intrinsic.h: Likewise.
3571 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3572 (nds32_print_operand): Likewise.
3573 * config/nds32/nds32.h (nds32_builtins): Likewise.
3575 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3576 Andrew Pinski <pinsika@gcc.gnu.org>
3579 * match.pd: Use constant_boolean_node of correct type instead of
3580 boolean_true_node or boolean_false_node for simplifying
3581 pointer comparisons to zero.
3583 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3585 PR tree-optimization/80021
3586 * tree.c (verify_type_variant): Make error call in verify_variant_match
3587 translatable and remove final full stop.
3589 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3591 * config/nds32/constants.md (unspec_volatile_element): Add
3592 UNSPEC_VOLATILE_EH_RETURN.
3593 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3594 nds32_output_stack_pop): Support dwarf exception handling process.
3595 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3596 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3597 exception handling process.
3598 (nds32_compute_stack_frame): Likewise.
3599 (nds32_return_addr_rtx): Likewise.
3600 (nds32_initial_elimination_offset): Likewise.
3601 (nds32_expand_prologue): Likewise.
3602 (nds32_expand_epilogue): Likewise.
3603 (nds32_dynamic_chain_address): New function.
3604 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3606 (DYNAMIC_CHAIN_ADDRESS): Define.
3607 (EH_RETURN_DATA_REGNO): Define.
3608 (EH_RETURN_STACKADJ_RTX): Define.
3609 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3610 patterns for dwarf exception handling.
3612 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3614 * config/nds32/nds32.h: Clean up obsolete macros.
3616 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3618 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3619 Add enum values for particular instructions.
3620 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3621 particular intrinsic functions.
3622 * config/nds32/nds32-intrinsic.md: Likewise.
3623 * config/nds32/nds32_intrinsic.h: Likewise.
3624 * config/nds32/nds32.h (nds32_builtins): Likewise.
3625 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3626 (btst, ave): New patterns for particular instructions.
3628 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3630 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3631 Add enum values for atomic load/store and memory sync.
3632 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3634 * config/nds32/nds32-intrinsic.md: Likewise.
3635 * config/nds32/nds32_intrinsic.h: Likewise.
3636 * config/nds32/nds32.h (nds32_builtins): Likewise.
3638 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3640 PR tree-optimization/85257
3641 * fold-const.c (native_encode_vector): If not all elts could fit
3642 and off is -1, return 0 rather than offset.
3643 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3644 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3645 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3646 adjust buffer in native_interpret_expr call.
3648 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3650 * config/nds32/constants.md (unspec_volatile_element): Add cache
3651 control enum values.
3652 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3653 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3654 * config/nds32/nds32.c (nds32_cctl_names): New.
3655 (nds32_print_operand): Handle cache control register names.
3656 * config/nds32/nds32.h (nds32_builtins): New enum values.
3657 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3659 * config/nds32/nds32.md (type): Add mmu.
3660 * config/nds32/pipelines.md (simple_insn): Add mmu.
3662 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3664 * config/nds32/nds32.md (type): Remove call.
3665 * config/nds32/pipelines.md (simple_insn): Likewise.
3667 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3669 * config/nds32/constants.md (unspec_volatile_element): Add
3670 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3671 UNSPEC_VOLATILE_FMFCFG.
3672 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3673 description for fmfcfg and fmfcsr.
3674 (bdesc_1arg): Add fmtcsr.
3675 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3676 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3677 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3678 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3679 unspec_fmfcfg): New patterns.
3680 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3681 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3682 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3683 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3684 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3685 __nds32__fmfcfg): Define.
3687 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3689 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3690 intrinsic register names.
3691 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3692 intrinsic register enum values and macros.
3694 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3696 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3697 for load/store addressing form.
3698 (nds32_print_operand_address): Likewise.
3700 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3703 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3704 based on LABEL_REF. Remove useless assertion.
3705 (pic_address_needs_scratch): Fix formatting.
3706 (sparc_legitimize_pic_address): Minor tweaks.
3707 (sparc_delegitimize_address): Adjust assertion accordingly.
3708 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3709 into symbolic_operand.
3710 (movsi_high_pic_label_ref): Likewise.
3711 (movsi_lo_sum_pic_label_ref): Likewise.
3712 (movdi_pic_label_ref): Likewise.
3713 (movdi_high_pic_label_ref): Likewise.
3714 (movdi_lo_sum_pic_label_ref): Likewise.
3716 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3718 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3719 custom LIB_SPEC setup.
3721 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3722 Kito Cheng <kito.cheng@gmail.com>
3724 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3725 * config/riscv/freebsd.h: New.
3727 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3729 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3730 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3733 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3734 Kito Cheng <kito.cheng@gmail.com>
3736 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3737 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3738 * config/nds32/nds32-protos.h (nds32_output_call,
3739 nds32_output_return): Declare.
3740 * config/nds32/nds32.md: Refine all the call and return patterns.
3742 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3745 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3746 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3748 PR rtl-optimization/84872
3749 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3750 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3753 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3755 * expr.c (copy_blkmode_to_reg): Revert 254862.
3756 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3758 2018-04-06 Richard Biener <rguenther@suse.de>
3761 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3762 after seeing a component reference with an adjacent field. Treat
3763 refs to arrays at struct end of external decls similar to
3764 refs to unconstrained commons.
3766 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3769 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3770 look through SAVE_EXPRs with non-side-effects argument. Adjust
3772 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3775 2018-04-06 Richard Biener <rguenther@suse.de>
3778 * alias.c (find_base_term): New wrapper around find_base_term
3779 unwinding CSELIB_VAL_PTR changes.
3780 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3783 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3785 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3787 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3788 constant definitions.
3789 ("nop"): lr 0,0 -> nopr r0
3790 ("nop_lr0", "nop_lr1"): New insn definitions.
3792 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3794 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3795 NDS32_V3PUSH_AVAILABLE_P macro.
3797 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3798 Chung-Ju Wu <jasonwucj@gmail.com>
3800 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3801 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3802 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3803 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3804 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3805 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3806 * config/nds32/constraints.md: New constraints and checking for hard
3807 float configuration.
3808 * config/nds32/iterators.md: New mode iterator and attribute for hard
3809 float configuration.
3810 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3812 * config/nds32/nds32-fpu.md: New file.
3813 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3814 deal with hard float code generation.
3815 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3817 (abi_type, float_reg_number): New enum type.
3818 * config/nds32/nds32-predicates.c: New predicates for hard float.
3819 * config/nds32/nds32-protos.h: Declare functions for hard float.
3820 * config/nds32/nds32.c: Implementation for hard float configuration.
3821 * config/nds32/nds32.h: Definitions for hard float configuration.
3822 * config/nds32/nds32.md: Include hard float machine description and
3823 modify patterns for hard float configuration.
3824 * config/nds32/nds32.opt: New options for hard float configuration.
3825 * config/nds32/predicates.md: New predicates for hard float
3828 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3830 * common/config/nds32/nds32-common.c
3831 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3833 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3836 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3837 CONSTRUCTOR_ELT (ctor, ...)->value.
3839 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3842 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3844 2018-04-05 Tom de Vries <tom@codesourcery.com>
3847 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3850 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3851 Kito Cheng <kito.cheng@gmail.com>
3853 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3854 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3855 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3856 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3858 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3859 Kito Cheng <kito.cheng@gmail.com>
3861 * config/nds32/constraints.md (Ufe): New memory constraint.
3862 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3863 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3864 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3866 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3867 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3869 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3871 * config/nds32/nds32.md: Use optimize_size in the condition for
3872 alu-shift instructions.
3874 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3876 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3878 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3880 * config/nds32/nds32.md (negsi2): Refine pattern.
3882 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3883 Chung-Ju Wu <jasonwucj@gmail.com>
3885 * config/nds32/iterators.md (shift_rotate): New code iterator.
3886 (shift): New code attribute.
3887 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3888 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3889 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3890 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3891 bit-wise operations.
3892 (andsi3, *andsi3): Ditto.
3893 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3894 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3895 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3896 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3897 nds32_ior_operand, nds32_xor_operand): New predicates.
3899 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3901 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3902 (addsi3, subsi3): ... this.
3904 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3906 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3908 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3910 * config/nds32/nds32.md: Adjust indention.
3912 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3914 * config/nds32/nds32.md (feature): New attribute.
3916 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3918 * config/nds32/nds32.md (subtype): New attribute.
3920 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3923 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3924 expansion to perform a bitwise AND of the argument followed by a
3925 boolean negation of the result.
3927 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3929 PR rtl-optimization/84878
3930 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3931 the basic block. Assert the use reference is not artificial and that
3932 it has an associated insn.
3934 2018-04-04 Michael Matz <matz@suse.de>
3936 * builtins.c (compute_objsize): Pass correct operand
3937 to array_at_struct_end_p.
3939 2018-04-04 Richard Biener <rguenther@suse.de>
3942 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3943 from contexts for DINFO_LEVEL_TERSE and below.
3945 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3947 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3948 resiter_operand condition.
3949 * config/nds32/nds32.md (*move<mode>): Ditto.
3951 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3952 Monk Chiang <sh.chiang04@gmail.com>
3954 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3956 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3958 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3960 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3961 Kito Cheng <kito.cheng@gmail.com>
3963 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3964 nds32_cond_code_str, output_cond_branch,
3965 output_cond_branch_compare_zero, nds32_expand_cbranch,
3966 nds32_expand_cstore, nds32_expand_movcc,
3967 nds32_output_cbranchsi4_equality_zero,
3968 nds32_output_cbranchsi4_equality_reg,
3969 nds32_output_cbranchsi4_equality_reg_or_const_int,
3970 nds32_output_cbranchsi4_greater_less_zero: New functions.
3971 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3972 nds32_expand_cstore, nds32_expand_movcc,
3973 nds32_output_cbranchsi4_equality_zero,
3974 nds32_output_cbranchsi4_equality_reg,
3975 nds32_output_cbranchsi4_equality_reg_or_const_int,
3976 nds32_output_cbranchsi4_greater_less_zero): Declare.
3977 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3978 nds32_rimm11s_operand): New predicates.
3979 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3980 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3983 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3985 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3986 * config/nds32/nds32.md: Ditto.
3987 * config/nds32/pipelines.md: Ditto.
3989 2018-04-04 Richard Biener <rguenther@suse.de>
3991 PR tree-optimization/85168
3992 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3993 propagating abnormals.
3995 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3997 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3999 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4000 Kito Cheng <kito.cheng@gmail.com>
4002 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4003 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4004 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4005 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4006 * config/nds32/nds32.md (sibcall_internal): New.
4007 (sibcall_register): Remove.
4008 (sibcall_immediate): Remove.
4009 (sibcall_value_internal): New.
4010 (sibcall_value_register): Remove.
4011 (sibcall_value_immediate): Remove.
4012 * config/nds32/predicates.md (nds32_general_register_operand): New.
4013 (nds32_call_address_operand): New.
4015 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4017 PR rtl-optimization/85167
4018 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4019 bb_defs if *split_p, instead preinitialize it to NULL.
4021 PR tree-optimization/85156
4022 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4023 evaluating the argument multiple times.
4025 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4027 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4029 (_mm_cvtpd_ps): Likewise.
4030 (_mm_cvttpd_epi32): Likewise.
4031 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4032 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4033 vector, pixel, and bool following altivec.h include.
4035 2018-04-03 Martin Sebor <msebor@redhat.com>
4037 * doc/extend.texi (Common Function Attributes): Clarify.
4038 (const attribute): Likewise.
4039 (pure attribute): Likewise.
4041 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4044 * config/i386/i386.c (ix86_expand_vector_set): Use
4045 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4047 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4049 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4050 instructions when changing rounding bits to preserve precision bits
4051 in the x87 control word.
4053 2018-04-03 Martin Liska <mliska@suse.cz>
4055 PR tree-optimization/82491
4056 * rtl.h (strip_offset_and_add): Replace += suboffset with
4057 poly_uint64 () + suboffset.
4059 2018-03-29 Martin Liska <mliska@suse.cz>
4060 Martin Jambor <mjambor@suse.cz>
4063 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4064 param_type is not an integral or pointer type.
4066 2018-04-03 Richard Biener <rguenther@suse.de>
4068 * sese.h (recompute_all_dominators): Remove.
4070 2018-04-02 Martin Sebor <msebor@redhat.com>
4072 * doc/invoke.texi (-Wrestrict): Fix typos.
4074 2018-04-02 Jim Wilson <jimw@sifive.com>
4076 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4077 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4078 (<optab>di3, <optab>si3_extend): Likewise.
4079 (<optab>si3_mask, <optab>si3_mask_1): New.
4080 (<optab>di3_mask, <optab>di3_mask_1): New.
4081 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4082 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4083 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4085 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4087 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4090 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4092 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4093 (nds32_canonicalize_comparison): New function.
4095 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4096 Kito Cheng <kito.cheng@gmail.com>
4097 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4099 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4100 * config/nds32/constants.md (unspec_volatile_element): Add
4101 UNSPEC_VOLATILE_RELAX_GROUP.
4102 * config/nds32/nds32-relax-opt.c: New file.
4103 * config/nds32/nds32-predicates.c
4104 (nds32_symbol_load_store_p): New function.
4105 * config/nds32/nds32-protos.h
4106 (nds32_symbol_load_store_p): Declare function.
4107 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4108 * config/nds32/nds32.c
4109 (nds32_register_pass): New function to register pass.
4110 (nds32_register_passes): New function to register passes.
4111 * config/nds32/nds32.md (relax_group): New pattern.
4112 * config/nds32/nds32.opt (mrelax-hint): New option.
4113 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4115 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4117 * config/nds32/t-nds32: Modify files dependency.
4119 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4121 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4122 (PROFILE_HOOK): Define its implementation.
4124 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4126 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4127 type and 32-bit size.
4129 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4132 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4133 (V_128_256): New mode iterator.
4134 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4135 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4136 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4138 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4139 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4141 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4144 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4145 NaN inputs correctly.
4147 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4150 * config/rs6000/vsx.md (??r): New mode attribute.
4151 (*vsx_mov<mode>_64bit): Use it.
4152 (*vsx_mov<mode>_32bit): Likewise.
4154 2018-03-30 Martin Sebor <msebor@redhat.com>
4156 PR tree-optimization/84818
4157 * builtins.c (check_access): Use warning_n.
4159 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4162 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4164 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4167 2018-03-30 Julia Koval <julia.koval@intel.com>
4170 * x86-tune.def (movx, partial_reg_dependency): Enable for
4173 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4176 * lra-constraints.c (process_alt_operands): Move setting
4177 this_alternative_matches below.
4179 2018-03-29 Martin Liska <mliska@suse.cz>
4182 * doc/invoke.texi: Document how LTO works with debug info.
4183 Describe auto-load support of binutils. Mention 'x86-64'
4184 as valid option value of -march option.
4186 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4188 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4191 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4192 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4193 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4196 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4199 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4200 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4201 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4202 for RS6000_BTM_POWERPC64.
4203 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4204 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4205 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4210 2018-03-28 Carl Love <cel@us.ibm.com>
4213 2017-09-27 Carl Love <cel@us.ibm.com>
4215 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4216 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4217 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4220 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4222 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4223 instead of __vector bool.
4224 (_mm_max_pu8): Likewise.
4225 (_mm_min_pi16): Likewise.
4227 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4230 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4231 (DIVWEUO): Likewise.
4233 (DIVDEUO): Likewise.
4234 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4235 DIVWEUO and DIVDEUO.
4236 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4237 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4238 (div_extend): Likewise.
4239 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4241 (__builtin_divweuo): Likewise.
4242 (__builtin_divdeo): Likewise.
4243 (__builtin_divdeuo): Likewise.
4245 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4248 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4249 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4251 PR tree-optimization/82004
4252 * gimple-match-head.c (optimize_pow_to_exp): New function.
4253 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4254 Don't fold to exp if optimize_pow_to_exp is false.
4256 2018-03-28 Martin Liska <mliska@suse.cz>
4259 * calls.c (initialize_argument_information): Fix trailing space.
4260 * common.opt: Fix typo and provide better explanation for
4261 -fsanitize-coverage option.
4262 * config/i386/i386.opt: Fix typo.
4264 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4265 Martin Liska <mliska@suse.cz>
4268 * gimplify.c (asan_poison_variable): Don't do the check for
4269 gimplify_omp_ctxp here.
4270 (gimplify_decl_expr): Do it here.
4271 (gimplify_target_expr): Likewise.
4273 2018-03-28 Martin Liska <mliska@suse.cz>
4276 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4277 chkp_type_bounds_count if MPX is not enabled.
4279 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4281 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4283 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4286 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4287 function to create the function decl for complex long double
4288 multiply and divide for -mabi=ieeelongdouble.
4289 (init_float128_ieee): Call it.
4291 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4294 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4295 -fcf-protection=branch -mibt.
4296 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4298 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4301 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4303 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4306 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4307 extern array declarations.
4309 2018-03-27 Richard Biener <rguenther@suse.de>
4312 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4313 explicit single_use checks.
4315 2018-03-27 Richard Biener <rguenther@suse.de>
4317 PR tree-optimization/85082
4318 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4321 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4323 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4324 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4325 Turn on fasynchronous-unwind-tables and funwind-tables.
4327 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4330 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4331 (*bmi_blsr_<mode>_ccz): Ditto.
4333 2018-03-26 Tom de Vries <tom@codesourcery.com>
4335 PR tree-optimization/85063
4336 * omp-general.c (offloading_function_p): New function. Factor out
4338 * omp-offload.c (pass_omp_target_link::gate): ... here.
4339 * omp-general.h (offloading_function_p): Declare.
4340 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4341 with attribute omp declare target for offloading functions.
4343 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4345 PR tree-optimization/84005
4346 * tree-data-ref.h (get_base_for_alignment): Declare.
4347 * tree-data-ref.c (get_base_for_alignment_1): New function.
4348 (get_base_for_alignment): Likewise.
4349 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4350 get_base_for_alignment to find a suitable base object, instead
4351 of always using drb->base_address.
4353 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4356 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4357 known size by default.
4359 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4362 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4363 and non BLKmode operands.
4365 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4368 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4369 Clean up attributes.
4371 2018-03-23 Richard Biener <rguenther@suse.de>
4374 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4375 we are going to emit early debug for LTO.
4377 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4380 * function.c (match_asm_constraints_1): Don't optimize if input
4381 doesn't satisfy general_operand predicate for output's mode.
4384 * alias.c (write_dependence_p): Don't require for x_canonicalized
4385 non-VOIDmode if x has VOIDmode.
4388 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4389 just don't try to optimize it rather than assert it never happens.
4391 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4393 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4394 macro expansions for definition of ST_INTERNAL_<mode> and
4395 LD_INTERNAL_<mode> builtins.
4396 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4398 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4400 (altivec_expand_st_builtin): Likewise.
4401 (altivec_expand_builtin): Remove calls to deleted functions.
4402 (rs6000_address_for_altivec): Delete this function.
4403 * config/rs6000/vector.md: Remove expands for
4404 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4406 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4409 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4410 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4411 re-computing once computed.
4412 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4413 (arm_init_machine_status): Initialize
4414 machine->static_chain_stack_bytes.
4416 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4419 * doc/extend.texi: Add four new prototypes for vec_ld.
4420 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4421 definitions for more logical presentation.
4422 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4423 entries for V1TI variants of __builtin_altivec_ld builtin.
4424 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4425 handling of V1TI variant of LVX icode pattern.
4426 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4427 (rs6000_gimple_fold_builtin): Likewise.
4428 (altivec_init_builtins): Add code to define
4429 __builtin_altivec_lvx_v1ti function.
4431 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4434 * function.c (match_asm_constraints_1): Don't do the optimization
4435 if input isn't a REG, SUBREG, MEM or constant.
4437 2018-03-22 Tom de Vries <tom@codesourcery.com>
4439 PR tree-optimization/84956
4440 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4441 bb_has_abnormal_pred.
4443 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4446 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4447 DECL_INITIAL (decl) to decl at the end.
4448 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4451 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4453 * doc/extend.texi (__builtin_tgmath): Document when complex
4454 integer types are treated as _Complex _Float64.
4456 2018-03-21 Tom de Vries <tom@codesourcery.com>
4458 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4460 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4462 PR tree-optimization/84960
4463 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4464 if it is ENTRY block, move them into single succ of ENTRY in that case.
4466 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4468 PR tree-optimization/84811
4469 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4470 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4471 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4472 (known_subrange_p): Update accordingly. Cast each value involved
4473 in the size comparison, rather than casting the result of the
4476 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4478 PR tree-optimization/84982
4479 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4480 by flipping the least significant bit rather than all bits from
4481 bitpos to bitpos + bitsize - 1.
4483 2018-03-21 Nathan Sidwell <nathan@acm.org>
4485 * doc/extend.texi (Deprecated Features): Remove mention of
4486 long-deleted deprecations.
4488 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4491 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4492 * configure: Regenerate.
4494 2018-03-21 Tom de Vries <tom@codesourcery.com>
4496 PR tree-optimization/83126
4497 * tree-parloops.c (num_phis): New function.
4498 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4500 2018-03-21 Nathan Sidwell <nathan@acm.org>
4502 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4503 mention anon-struct/union members and trailing attributes.
4505 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4507 PR tree-optimization/84969
4508 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4509 builtin memset partitions if they set different rhs values.
4511 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4513 PR rtl-optimization/84989
4514 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4515 VEC_DUPLICATE with scalar result mode.
4517 2018-03-21 Martin Liska <mliska@suse.cz>
4520 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4521 not intended return statement.
4523 2018-03-21 Martin Liska <mliska@suse.cz>
4526 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4527 (chkp_find_bound_slots_1): Limit number of iterations.
4529 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4532 * Minor grammar fixes for x86 options.
4534 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4537 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4538 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4540 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4543 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4544 (altivec_lvx_<mode>_1op): Likewise.
4545 (altivec_stvx_<mode>_2op): Likewise.
4546 (altivec_stvx_<mode>_1op): Likewise.
4547 (altivec_lvx_<VM2:mode>): New define_expand.
4548 (altivec_stvx_<VM2:mode>): Likewise.
4549 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4550 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4551 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4552 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4553 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4554 (rs6000_gen_lvx): Likewise.
4555 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4556 (altivec_expand_stv_builtin): Likewise.
4557 (altivec_expand_builtin): Likewise.
4558 * config/rs6000/vector.md: Likewise.
4560 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4563 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4566 2018-03-20 Richard Biener <rguenther@suse.de>
4569 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4570 sign-conversions as zero, fall back to standard scalar_stmt
4573 2018-03-20 Martin Liska <mliska@suse.cz>
4576 * predict.c (rebuild_frequencies): Handle case when we have
4577 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4579 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4582 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4583 flag_section_anchors.
4584 * varasm.c (use_blocks_for_decl_p): Remove hack for
4585 dw2_force_const_mem.
4588 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4590 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4591 be created, use lowpart_subreg of operands[0] rather than operands[0]
4593 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4594 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4595 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4596 and n constraint instead of aarch64_shift_imm_di and Usd.
4597 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4598 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4600 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4603 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4604 to favor GPR over NEON registers.
4605 (<shift>di3_neon): Likewise.
4607 2018-03-20 Tom de Vries <tom@codesourcery.com>
4610 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4611 (nvptx_process_pars): Emit bar.sync asap and alap.
4613 2018-03-20 Tom de Vries <tom@codesourcery.com>
4616 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4617 seen_label if seen_label is already set.
4619 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4622 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4623 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4624 Use 1U instead of 1. Formatting fixes.
4627 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4628 instead of TREE_TYPE (s1) for the return value.
4630 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4632 PR tree-optimization/84946
4633 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4634 bitsize + bitsize in poly_uint64 rather than poly_int64.
4637 * dwarf2asm.c: Include fold-const.c.
4638 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4639 of decl rather than decl itself.
4641 PR rtl-optimization/84643
4642 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4644 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4647 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4648 calling assemble_variable.
4650 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4653 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4654 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4656 2018-03-19 Jim Wilson <jimw@sifive.com>
4659 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4660 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4661 (riscv_first_stack_step): Likewise.
4662 (riscv_option_override): Use STACK_BOUNDARY instead of
4664 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4666 (BIGGEST_ALIGNMENT): Set to 128.
4667 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4668 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4671 2018-03-19 Richard Biener <rguenther@suse.de>
4673 PR tree-optimization/84933
4674 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4675 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4677 2018-03-19 Richard Biener <rguenther@suse.de>
4679 PR tree-optimization/84859
4680 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4681 (cond_if_else_store_replacement): Perform sinking operation on
4682 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4683 Generalize what a BB with a single eligible store is.
4685 2018-03-19 Richard Biener <rguenther@suse.de>
4687 PR tree-optimization/84929
4688 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4689 chrec_is_positive against non-chrec arg.
4691 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4694 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4696 2018-03-18 Martin Liska <mliska@suse.cz>
4698 PR rtl-optimization/84635
4699 * regrename.c (build_def_use): Use matches_mode only when
4702 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4704 PR tree-optimization/84913
4705 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4706 vectorize chains of COND_EXPRs.
4708 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4710 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4712 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4714 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4716 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4718 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4720 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4721 Kito Cheng <kito.cheng@gmail.com>
4723 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4724 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4725 (nds32_adjust_reg_alloc_order): New function.
4726 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4728 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4730 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4731 nds32_print_operand, nds32_print_operand_address): Use
4732 HOST_WIDE_INT_PRINT_DEC instead.
4734 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4736 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4738 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4741 * config/i386/i386.c (initial_ix86_tune_features,
4742 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4744 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4745 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4746 rather than 1u << ix86_tune. Formatting fix.
4747 (ix86_option_override_internal): Change ix86_arch_mask from
4748 unsigned int to unsigned HOST_WIDE_INT, initialize to
4749 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4750 (ix86_function_specific_restore): Likewise.
4752 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4755 * postreload.c (reload_combine_recognize_pattern): Perform
4756 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4757 truncate_int_for_mode the result for the destination's mode.
4760 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4761 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4763 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4766 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4769 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4770 regno_allocno_class_array and sorted_pseudos.
4771 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4772 insns where regno is used.
4774 2018-03-16 Martin Liska <mliska@suse.cz>
4777 * multiple_target.c (create_dispatcher_calls): Redirect
4778 reference in the symbol table.
4780 2018-03-16 Martin Liska <mliska@suse.cz>
4783 * multiple_target.c (create_dispatcher_calls): Redirect also
4786 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4790 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4791 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4792 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4794 2018-03-16 Julia Koval <julia.koval@intel.com>
4796 * doc/invoke.texi (Skylake Server): Add CLWB.
4797 Cannonlake): Remove CLWB.
4799 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4801 PR tree-optimization/84841
4802 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4804 (FLOAT_ONE_CONST_TYPE): Define.
4805 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4806 (sort_by_operand_rank): Put entries with higher constant_type last
4807 rather than first to match comments.
4809 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4811 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4814 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4817 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4818 check and use address of target always.
4820 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4823 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4824 (indirect_thunk_bnd_needed): Likewise.
4825 (indirect_thunks_used): Likewise.
4826 (indirect_thunks_bnd_used): Likewise.
4827 (indirect_return_needed): New.
4828 (indirect_return_bnd_needed): Likewise.
4829 (output_indirect_thunk_function): Add a bool argument for
4831 (output_indirect_thunk_function): Don't generate alias for
4832 function return thunk.
4833 (ix86_code_end): Call output_indirect_thunk_function to generate
4834 function return thunks.
4835 (ix86_output_function_return): Set indirect_return_bnd_needed
4836 and indirect_return_needed instead of indirect_thunk_bnd_needed
4837 and indirect_thunk_needed.
4839 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4841 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4842 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4843 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4845 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4846 Paul Hua <paul.hua.gm@gmail.com>
4849 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4851 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4853 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4854 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4857 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4860 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4861 instead of GET_MODE_SIZE when comparing Units.
4863 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4866 * varasm.c (hash_section): Return an unchangeble hash value
4867 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4868 Return !aarch64_can_use_per_function_literal_pools_p ().
4870 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4873 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4874 rather than address of cmode as last argument to prepare_cmp_insn.
4876 2018-03-15 Julia Koval <julia.koval@intel.com>
4878 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4879 F_AVX512VNNI, F_AVX512BITALG): New.
4881 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4884 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4885 insn for floating-point loads and stores.
4887 2018-03-14 Carl Love <cel@us.ibm.com>
4889 * config/rs6000/rs6000-c.c: Add macro definitions for
4890 ALTIVEC_BUILTIN_VEC_PERMXOR.
4891 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4892 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4893 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4894 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4896 * config/doc/extend.texi: Add prototypes for vec_permxor.
4898 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4901 * diagnostic-show-locus.c (class layout_point): Convert m_line
4902 from int to linenum_type.
4903 (line_span::comparator): Use linenum "compare" function when
4904 comparing line numbers.
4905 (test_line_span): New function.
4906 (layout_range::contains_point): Convert param "row" from int to
4908 (layout_range::intersects_line_p): Likewise.
4909 (layout::will_show_line_p): Likewise.
4910 (layout::print_source_line): Likewise.
4911 (layout::should_print_annotation_line_p): Likewise.
4912 (layout::print_annotation_line): Likewise.
4913 (layout::print_leading_fixits): Likewise.
4914 (layout::annotation_line_showed_range_p): Likewise.
4915 (struct line_corrections): Likewise for field m_row.
4916 (line_corrections::line_corrections): Likewise for param "row".
4917 (layout::print_trailing_fixits): Likewise.
4918 (layout::get_state_at_point): Likewise.
4919 (layout::get_x_bound_for_row): Likewise.
4920 (layout::print_line): Likewise.
4921 (diagnostic_show_locus): Likewise for locals "last_line" and
4923 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4924 * input.c (selftest::test_linenum_comparisons): New function.
4925 (selftest::input_c_tests): Call it.
4926 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4927 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4928 * selftest.h (ASSERT_GT): New macro.
4929 (ASSERT_GT_AT): New macro.
4930 (ASSERT_LT): New macro.
4931 (ASSERT_LT_AT): New macro.
4933 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4935 PR rtl-optimization/84780
4936 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4938 2018-03-14 Martin Liska <mliska@suse.cz>
4940 * tree.c (record_node_allocation_statistics): Use
4941 get_stats_node_kind.
4942 (get_stats_node_kind): New function extracted from
4943 record_node_allocation_statistics.
4944 (free_node): Use get_stats_node_kind.
4946 2018-03-14 Richard Biener <rguenther@suse.de>
4948 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4949 that the value-set of ANTIC_IN doesn't grow.
4952 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4954 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4955 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4957 2018-03-14 Julia Koval <julia.koval@intel.com>
4959 * config.gcc (icelake-client, icelake-server): New.
4961 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4962 (initial_ix86_arch_features): Ditto.
4963 (PTA_SKYLAKE): Add SGX.
4964 (PTA_ICELAKE): Remove.
4965 (PTA_ICELAKE_CLIENT): New.
4966 (PTA_ICELAKE_SERVER): New.
4967 (ix86_option_override_internal): Split up icelake on icelake client and
4969 (get_builtin_code_for_version): Ditto.
4970 (fold_builtin_cpu): Ditto.
4971 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4972 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4973 * config/i386/i386.h (processor_type): Ditto.
4974 * doc/invoke.texi: Ditto.
4976 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4979 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4980 INTEGER_CST offset, add it together with bitpos / 8 and
4981 sign extend based on POINTER_SIZE.
4985 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4988 * config/i386/constraints.md (Yc): New register constraint.
4989 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4990 Use Yc constraint for alternative 2 of operand 0. Remove
4991 preferred_for_speed attribute.
4993 2018-03-14 Richard Biener <rguenther@suse.de>
4995 PR tree-optimization/84830
4996 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4997 with the old one to avoid oscillations.
4999 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5002 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5004 (assign_by_spills): Return a flag of reload assignment failure.
5005 Do not process the reload assignment failures. Do not spill other
5006 reload pseudos if they has the same reg class. Update n if
5008 (lra_assign): Add a return arg. Set up from the result of
5009 assign_by_spills call.
5010 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5011 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5012 usage_insns if it is not NULL.
5013 (spill_hard_reg_in_range): New function.
5014 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5015 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5016 function prototypes.
5017 (lra_assign): Change prototype.
5018 * lra.c (lra): Add code to deal with fails by splitting hard reg
5021 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5023 * config/riscv/riscv.opt (mrelax): New option.
5024 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5025 "norelax" when riscv_mrelax is disabled.
5026 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5028 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5031 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5032 reassociation for int modes.
5034 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5036 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5037 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5039 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5040 * config/aarch64/aarch64-sve.md
5041 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5042 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5043 (*extend<mode><Vwide>2): Rename to...
5044 (aarch64_sve_extend<mode><Vwide>2): ...this.
5045 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5046 renaming the old pattern to...
5047 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5049 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5050 define_expand, renaming the old pattern to...
5051 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5052 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5053 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5054 account when deciding which SVE instruction the optab should use.
5055 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5057 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5059 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5060 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5061 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5062 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5063 (tlsdesc_small_<mode>): Turn a define_expand and use
5064 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5065 (tlsdesc_small_advsimd_<mode>): ...this.
5066 (tlsdesc_small_sve_<mode>): New pattern.
5068 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5070 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5071 (UNSPEC_UMUL_HIGHPART): New constants.
5072 (MUL_HIGHPART): New int iteraor.
5073 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5074 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5076 (*<su>mul<mode>3_highpart): New define_insn.
5078 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5081 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5084 2018-03-13 Martin Liska <mliska@suse.cz>
5087 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5089 (sem_item_optimizer::~sem_item_optimizer): Release it.
5090 (sem_item_optimizer::merge_classes): Register variable aliases.
5091 (sem_item_optimizer::fixup_pt_set): New function.
5092 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5093 * ipa-icf.h: Declare new variables and functions.
5095 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5098 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5099 integer_pow2p@2 and test integer_pow2p in condition.
5100 (A < 0 ? C : 0): Similarly for @1.
5103 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5104 characters starting at p contain '\0' character, don't look beyond
5108 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5109 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5112 * reg-stack.c (change_stack): Change update_end var from int to
5113 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5114 also call set_block_for_insn on the newly added insns and rescan.
5117 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5118 on the last operand.
5121 * tree.c (stabilize_reference_1): Return save_expr (e) for
5122 STATEMENT_LIST even if it doesn't have side-effects.
5124 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5126 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5128 2018-03-12 Renlin Li <renlin.li@arm.com>
5130 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5131 aarch64_output_scalar_simd_mov_immediate.
5133 2018-03-12 Martin Sebor <msebor@redhat.com>
5135 PR tree-optimization/83456
5136 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5137 for perfectly overlapping calls to memcpy.
5138 (gimple_fold_builtin_memory_chk): Same.
5139 (gimple_fold_builtin_strcpy): Handle no-warning.
5140 (gimple_fold_builtin_stxcpy_chk): Same.
5141 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5143 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5145 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5146 parameter. Use it for SFmode.
5147 (rs6000_function_arg_advance_1): Adjust.
5148 (rs6000_function_arg): Adjust.
5149 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5151 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5153 PR rtl-optimization/84169
5154 PR rtl-optimization/84780
5155 * combine.c (can_combine_p): Check for a 2-insn combination whether
5156 the destination register is used between the two insns, too.
5158 2018-03-12 Richard Biener <rguenther@suse.de>
5160 PR tree-optimization/84803
5161 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5162 for refs DR analysis didn't process.
5164 2018-03-12 Richard Biener <rguenther@suse.de>
5166 PR tree-optimization/84777
5167 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5168 force-vectorize loops ignore whether we are optimizing for size.
5170 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5172 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5173 (TARGET_MD_ASM_ADJUST): Define.
5175 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5176 Kito Cheng <kito.cheng@gmail.com>
5177 Chung-Ju Wu <jasonwucj@gmail.com>
5179 * config/nds32/nds32.c (nds32_compute_stack_frame,
5180 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5181 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5182 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5183 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5184 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5185 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5186 * config/nds32/nds32.md (prologue, epilogue): Use macro
5187 NDS32_V3PUSH_AVAILABLE_P to do checking.
5189 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5192 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5193 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5194 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5195 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5196 addition of most attributes on !orig_type_die or the attribute not
5197 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5199 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5200 Chung-Ju Wu <jasonwucj@gmail.com>
5202 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5204 * config/nds32/nds32.opt (mvh): New option.
5206 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5207 Chung-Ju Wu <jasonwucj@gmail.com>
5209 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5211 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5212 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5215 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5216 Chung-Ju Wu <jasonwucj@gmail.com>
5218 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5220 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5221 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5223 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5224 Kito Cheng <kito.cheng@gmail.com>
5225 Chung-Ju Wu <jasonwucj@gmail.com>
5227 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5228 UNSPEC_FFMISM and UNSPEC_FLMISM.
5229 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5230 for ffb, ffmism and flmism.
5231 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5232 (unspec_ffmism): Ditto.
5233 (unspec_flmism): Ditto.
5234 (nds32_expand_builtin_impl): Check if string extension is available.
5235 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5236 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5238 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5241 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5244 * lra-assigns.c (assign_by_spills): Return a flag of reload
5245 assignment failure. Do not process the reload assignment
5246 failures. Do not spill other reload pseudos if they has the same
5248 (lra_assign): Add a return arg. Set up from the result of
5249 assign_by_spills call.
5250 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5251 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5252 usage_insns if it is not NULL.
5253 (spill_hard_reg_in_range): New function.
5254 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5255 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5256 function prototypes.
5257 (lra_assign): Change prototype.
5258 * lra.c (lra): Add code to deal with fails by splitting hard reg
5261 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5264 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5266 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5269 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5270 (dw_val_node): Add val_symbolic_view.
5271 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5272 (symview_upper_bound): New.
5273 (new_line_info_table): Initialize symviews_since_reset.
5274 (dwarf2out_source_line): Count symviews_since_reset and set
5275 symview_upper_bound.
5276 (dw_val_equal_p): Handle symview.
5277 (add_AT_symview): New.
5278 (print_dw_val): Handle symview.
5279 (attr_checksum, attr_checksum_ordered): Likewise.
5280 (same_dw_val_p, size_of_die): Likewise.
5281 (value_format, output_die): Likewise.
5282 (add_high_low_attributes): Use add_AT_symview for entry_view.
5283 (dwarf2out_finish): Reset symview_upper_bound, clear
5286 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5289 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5290 Add strict argument and use it.
5291 (rs6000_split_multireg_move): Update for new strict argument.
5292 (mem_operand_gpr): Disallow all non-offsettable addresses.
5293 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5295 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5298 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5299 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5300 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5303 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5304 decl, use remap_type if we want to use the type.
5306 2018-03-09 Martin Sebor <msebor@redhat.com>
5308 PR tree-optimization/84526
5309 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5311 (builtin_access::generic_overlap): Be prepared to handle non-array
5314 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5316 PR rtl-optimization/84682
5317 * lra-constraints.c (process_address_1): Check is_address flag
5318 for address constraints.
5319 (process_alt_operands): Likewise.
5320 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5321 preprocess_constraints.
5322 * recog.h (preprocess_constraints): Add oploc parameter.
5324 * recog.c (preprocess_constraints): Test address_operand for
5325 CT_ADDRESS constraints.
5327 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5330 * lra-assigns.c (assign_by_spills): Return a flag of reload
5331 assignment failure. Do not process the reload assignment
5332 failures. Do not spill other reload pseudos if they has the same
5334 (lra_assign): Add a return arg. Set up from the result of
5335 assign_by_spills call.
5336 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5337 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5338 usage_insns if it is not NULL.
5339 (spill_hard_reg_in_range): New function.
5340 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5341 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5342 function prototypes.
5343 (lra_assign): Change prototype.
5344 * lra.c (lra): Add code to deal with fails by splitting hard reg
5347 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5350 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5351 Accept complain bool parameter. Only emit errors if it is true.
5352 (arm_parse_cpu_option_name): Likewise.
5353 (arm_target_thumb_only): Adjust callers of the above.
5354 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5355 prototype to take a default true bool parameter.
5356 (arm_parse_arch_option_name): Likewise.
5358 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5359 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5363 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5364 * configure: Regenerate.
5365 * configure.ac ("linker --version-script option"): New.
5366 ("linker soname option"): New.
5368 2018-03-09 Richard Biener <rguenther@suse.de>
5370 PR tree-optimization/84775
5371 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5372 immediate uses of predicate stmts and mark them modified.
5375 PR tree-optimization/84178
5376 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5378 (version_loop_for_if_conversion): Delay update_ssa call.
5379 (tree_if_conversion): Delay update_ssa until after predicate
5382 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5385 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5386 when the function accesses prior frames.
5388 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5391 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5392 gen_llsym, otherwise call maybe_gen_llsym.
5395 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5396 has ',' character inside of it.
5398 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5401 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5402 as clobbering CC_REGNUM.
5404 2018-03-08 Richard Biener <rguenther@suse.de>
5407 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5408 (follow_copies_to_constant): Do not follow SSA names registered
5411 2018-03-08 Richard Biener <rguenther@suse.de>
5413 PR tree-optimization/84178
5414 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5416 (version_loop_for_if_conversion): Delay update_ssa call.
5417 (tree_if_conversion): Delay update_ssa until after predicate
5420 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5422 PR tree-optimization/84178
5423 * tree-if-conv.c (release_bb_predicate): Remove the
5424 the assertion that the stmts have NULL use_ops.
5425 Discard the statements, asserting that they haven't
5426 yet been added to a BB.
5428 2018-03-08 Richard Biener <rguenther@suse.de>
5430 PR tree-optimization/84746
5431 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5432 (phi_translate): Pass in destination ANTIC_OUT set.
5433 (phi_translate_1): Likewise. For a simplified result lookup
5434 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5435 (phi_translate_set): Adjust.
5436 (do_pre_regular_insertion): Likewise.
5437 (do_pre_partial_partial_insertion): Likewise.
5439 2018-03-08 Martin Liska <mliska@suse.cz>
5441 PR gcov-profile/84735
5442 * doc/gcov.texi: Document usage of profile files.
5443 * gcov-io.h: Document changes in the format.
5445 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5449 * dwarf2out.c (struct dw_line_info_table): Update comments for
5451 (FORCE_RESET_NEXT_VIEW): New.
5452 (FORCE_RESETTING_VIEW_P): New.
5453 (RESETTING_VIEW_P): Check for -1 too.
5454 (ZERO_VIEW_P): Likewise.
5455 (new_line_info_table): Force-reset next view.
5456 (dwarf2out_begin_function): Likewise.
5457 (dwarf2out_source_line): Simplify zero_view_p initialization.
5458 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5459 view directly. Omit view when omitting .loc at line 0.
5461 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5463 PR tree-optimization/84740
5464 * tree-switch-conversion.c (process_switch): Call build_constructors
5465 only if info.phi_count is non-zero.
5467 PR tree-optimization/84739
5468 * tree-tailcall.c (find_tail_calls): Check call arguments against
5469 DECL_ARGUMENTS (current_function_decl) rather than
5470 DECL_ARGUMENTS (func) when checking for tail recursion.
5472 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5474 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5475 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5476 Volker Reichelt's entry and add entries for people that perform
5477 GCC fuzzy testing and report numerous bugs.
5479 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5482 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5483 readonly data in sdata, if that is disabled.
5484 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5485 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5486 -mreadonly-in-sdata option.
5488 2018-03-07 Martin Sebor <msebor@redhat.com>
5490 PR tree-optimization/84468
5491 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5492 basic block when looking for nul assignment.
5494 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5497 * except.h (output_function_exception_table): Adjust prototype.
5498 * except.c (output_function_exception_table): Remove FNNAME parameter
5499 and add SECTION parameter. Ouput one part of the table at a time.
5500 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5501 the first part of the exception table and emit unwind directives.
5502 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5503 (i386_pe_seh_cold_init): Likewise.
5504 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5505 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5506 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5507 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5508 * config/i386/winnt.c: Include except.h.
5509 (struct seh_frame_state): Add reg_offset, after_prologue and
5510 in_cold_section fields.
5511 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5512 (i386_pe_seh_cold_init): New function.
5513 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5514 to seh->in_cold_section.
5515 (seh_emit_push): Record the offset of the push.
5516 (seh_emit_save): Record the offet of the save.
5517 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5518 Test seh->after_prologue to disregard the epilogue.
5519 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5520 (i386_pe_end_cold_function): New function.
5522 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5525 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5526 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5529 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5531 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5532 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5535 * multiple_target.c: Include tree-inline.h and intl.h.
5536 (expand_target_clones): Diagnose and fail if node->definition and
5537 !tree_versionable_function_p (node->decl).
5539 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5541 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5543 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5544 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5545 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5547 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5550 * combine.c (try_combine): Use reg_or_subregno instead of handling
5551 just paradoxical SUBREGs and REGs.
5553 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5555 * config/arc/arc.c (arc_finalize_pic): Remove function.
5556 (arc_must_save_register): We use single base PIC register, remove
5557 checks to save/restore the PIC register.
5558 (arc_expand_prologue): Likewise.
5559 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5561 (arc_verify_short): Likewise.
5562 (arc_attr_type): Likewise.
5563 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5564 (walk_stores): Likewise.
5565 (arc_address_cost): Make it static.
5566 (arc_verify_short): Likewise.
5567 (branch_dest): Likewise.
5568 (arc_attr_type): Likewise.
5569 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5570 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5571 (arc_final_prescan_insn): Remove inserting the nops due to
5572 hardware hazards. It is done in reorg step.
5573 (insn_length_variant_t): Remove.
5574 (insn_length_parameters_t): Likewise.
5575 (arc_insn_length_parameters): Likewise.
5576 (arc_get_insn_variants): Likewise.
5577 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5579 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5582 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5585 PR tree-optimization/84687
5586 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5588 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5590 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5592 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5593 Rename to ppc_speculation_barrier.
5594 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5595 __builtin_ppc_speculation_barrier.
5597 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5600 * combine.c (combine_simplify_rtx): Don't try to simplify if
5601 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5604 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5606 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5607 to 32 bytes when compiling for POWER9.
5609 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5612 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5613 regparm >= 3 with no arg reg available also for calls with
5614 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5617 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5619 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5621 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5624 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5626 2018-03-05 Richard Biener <rguenther@suse.de>
5628 PR tree-optimization/84486
5629 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5630 When inserting a __builtin_assume_aligned call set the LHS
5631 SSA name alignment info accordingly.
5633 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5635 PR tree-optimization/84114
5636 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5637 Avoid reassociation of FLOAT_MODE addition.
5639 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5641 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5642 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5643 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5644 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5645 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5646 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5647 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5649 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5650 __builtin_ia32_wbinvd): New builtins.
5651 (SPECIAL_ARGS2): New.
5652 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5653 (SPECIAL_ARGS2): New.
5654 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5655 (ix86_valid_target_attribute_inner_p): Ditto.
5656 (ix86_init_mmx_sse_builtins): Add special_args2.
5657 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5658 TARGET_WBNOINVD_P): New.
5659 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5660 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5661 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5662 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5663 * config/i386/pconfigintrin.h: New file.
5664 * config/i386/wbnoinvdintrin.h: Ditto.
5665 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5666 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5668 2018-03-05 Richard Biener <rguenther@suse.de>
5670 PR tree-optimization/84670
5671 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5673 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5674 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5675 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5676 doesn't grow if all successors (recursively) were visited at least
5679 2018-03-05 Richard Biener <rguenther@suse.de>
5681 PR tree-optimization/84650
5682 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5683 if executed in the loop pipeline.
5685 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5687 * doc/configfiles.texi (Configuration Files): Move info about
5688 conditionalizing $target-protos.h to...
5689 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5690 differs from $target-protos.h.
5692 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5693 Chung-Ju Wu <jasonwucj@gmail.com>
5695 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5696 * config/nds32/nds32-multiple.md (setmemsi): Define.
5697 * config/nds32/nds32-memory-manipulation.c
5698 (nds32_gen_dup_4_byte_to_word_value): New.
5699 (emit_setmem_word_loop): New.
5700 (emit_setmem_byte_loop): New.
5701 (nds32_expand_setmem_loop): New.
5702 (nds32_expand_setmem_loop_v3m): New.
5703 (nds32_expand_setmem_unroll): New.
5704 (nds32_expand_setmem): New.
5706 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5707 Chung-Ju Wu <jasonwucj@gmail.com>
5709 * config/nds32/nds32-memory-manipulation.c
5710 (nds32_emit_load_store): New.
5711 (nds32_emit_post_inc_load_store): New.
5712 (nds32_emit_mem_move): New.
5713 (nds32_emit_mem_move_block): New.
5714 (nds32_expand_movmemsi_loop_unknown_size): New.
5715 (nds32_expand_movmemsi_loop_known_size): New.
5716 (nds32_expand_movmemsi_loop): New.
5717 (nds32_expand_movmemsi_unroll): New.
5718 (nds32_expand_movmemqi): Rename ...
5719 (nds32_expand_movmemsi): ... to this.
5720 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5721 (movmemsi): ... to this.
5722 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5723 (nds32_expand_movmemsi): ... to this.
5725 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5726 Monk Chiang <sh.chiang04@gmail.com>
5727 Chung-Ju Wu <jasonwucj@gmail.com>
5729 * config/nds32/nds32-protos.h
5730 (nds32_expand_load_multiple): New arguments.
5731 (nds32_expand_store_multiple): Ditto.
5732 (nds32_valid_multiple_load_store): Rename ...
5733 (nds32_valid_multiple_load_store_p): ... to this.
5734 * config/nds32/nds32-memory-manipulation.c
5735 (nds32_expand_load_multiple): Refine implementation.
5736 (nds32_expand_store_multiple): Ditto.
5737 * config/nds32/nds32-multiple.md
5738 (load_multiple): Update nds32_expand_load_multiple interface.
5739 (store_multiple): Update nds32_expand_store_multiple interface.
5740 * config/nds32/nds32-predicates.c
5741 (nds32_valid_multiple_load_store): Rename ...
5742 (nds32_valid_multiple_load_store_p): ... to this and refine
5744 * config/nds32/predicates.md
5745 (nds32_load_multiple_and_update_address_operation): New predicate.
5746 (nds32_store_multiple_and_update_address_operation): New predicate.
5748 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5749 Chung-Ju Wu <jasonwucj@gmail.com>
5751 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5752 (combo): New attribute.
5753 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5755 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5757 * config/nds32/nds32.opt: Change -mcmodel= default value.
5759 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5760 Monk Chiang <sh.chiang04@gmail.com>
5761 Chung-Ju Wu <jasonwucj@gmail.com>
5763 * config/nds32/constants.md (unspec_element): New enum.
5764 * config/nds32/constraints.md (Umw): New constraint.
5765 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5766 * config/nds32/nds32-intrinsic.md: Likewise.
5767 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5768 (nds32_valid_smw_lwm_base_p): New.
5769 (nds32_output_smw_single_word): New.
5770 (nds32_output_lmw_single_word): New.
5771 (nds32_expand_unaligned_load): New.
5772 (nds32_expand_unaligned_store): New.
5773 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5774 (nds32_output_smw_single_word): Declare.
5775 (nds32_output_lmw_single_word): Declare.
5776 (nds32_expand_unaligned_load): Declare.
5777 (nds32_expand_unaligned_store): Declare.
5778 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5779 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5780 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5781 NDS32_BUILTIN_UASTORE_DW.
5782 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5785 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5786 Kito Cheng <kito.cheng@gmail.com>
5787 Chung-Ju Wu <jasonwucj@gmail.com>
5789 * config/nds32/nds32-intrinsic.c
5790 (nds32_expand_builtin_null_ftype_reg): Delete.
5791 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5792 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5793 (nds32_read_argument): New.
5794 (nds32_legitimize_target): Ditto.
5795 (nds32_legitimize_argument): Ditto.
5796 (nds32_check_constant_argument): Ditto.
5797 (nds32_expand_unop_builtin): Ditto.
5798 (nds32_expand_unopimm_builtin): Ditto.
5799 (nds32_expand_binop_builtin): Ditto.
5800 (nds32_builtin_decl_impl): Ditto.
5801 (builtin_description): Ditto.
5802 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5803 (nds32_init_builtins_impl): Ditto.
5804 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5805 (nds32_builtin_decl): New.
5806 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5807 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5809 2018-03-02 Jeff Law <law@redhat.com>
5811 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5812 (redundant_insn, fill_simple_delay_slots): Likewise.
5813 (fill_slots_from_thread): Likewise.
5814 * resource.c (mark_referenced_resources): Likewise.
5815 (mark_set_resources, find_dead_or_set_registers): Likewise.
5817 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5819 * substring-locations.h (format_warning_va): Formatting fix for
5821 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5823 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5824 * substring-locations.c: Include intl.h.
5825 (format_warning_va): Turned into small wrapper around
5826 format_warning_n_va, renamed to ...
5827 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5828 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5830 (format_warning_at_substring_n): New function.
5831 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5832 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5833 format_warning_at_substring with just a shorter name instead of
5834 const function pointer.
5835 (fmtwarn_n): New function.
5836 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5837 appropriate, get rid of all the fmtstr temporaries, move conditionals
5838 with G_() wrapped string literals directly into fmtwarn arguments,
5839 cast dir.len to (int), formatting fixes.
5841 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5843 * doc/invoke.texi: Remove "Cilk Plus" references.
5845 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5846 Richard Biener <rguenther@suse.de>
5849 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5850 for error or warning attributes if CALL_FROM_THUNK_P is set.
5853 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5856 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5857 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5860 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5861 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5863 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5864 instead of -1U in last predictors element's probability member.
5866 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5869 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5870 arguments if they are comparable.
5872 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5874 PR tree-optimization/84634
5875 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5876 masks and masked_loop_p with a single loop_masks, making sure it's
5877 null for bb vectorization.
5879 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5881 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5882 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5883 loop->force_vectorize to check whether there is no alias.
5885 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5888 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5890 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5892 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5893 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5894 instead of a loop around prev_real_insn.
5895 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5899 * config/i386/i386.c (ix86_print_operand): Use conditional
5900 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5903 2018-03-02 Richard Biener <rguenther@suse.de>
5905 PR tree-optimization/84427
5906 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5907 (bitmap_set_subtract_values): Rewrite to handle multiple
5910 (prune_clobbered_mems): Likewise.
5911 (phi_translate): Take edge instead of pred/phiblock.
5912 (phi_translate_1): Likewise.
5913 (phi_translate_set): Likewise. Insert all translated
5914 exprs for a value into the set, keeping possibly multiple
5915 expressions per value.
5916 (compute_antic_aux): Adjust for phi_translate changes.
5917 When intersecting union the expressions and prune those
5918 not in the final value set, keeping possibly multiple
5919 expressions per value. Do not use value-insertion
5920 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5921 all expressions. Add verification that the value-sets
5922 only shrink during iteration.
5923 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5924 (do_pre_regular_insertion): Likewise.
5925 (do_pre_partial_partial_insertion): Likewise.
5927 2018-03-02 Richard Biener <rguenther@suse.de>
5930 * config/darwin.c (saved_debug_info_level): New static global.
5931 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5932 (darwin_asm_lto_end): Restore debug info generation settings.
5934 2018-03-01 Martin Liska <mliska@suse.cz>
5937 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5940 2018-03-01 Richard Biener <rguenther@suse.de>
5943 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5944 type annotation with LTO when debug was disabled at compile-time.
5946 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5948 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5950 (mips_final_postscan_insn): Likewise.
5952 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5954 PR rtl-optimization/84528
5955 * alias.c (init_alias_target): Add commentary.
5956 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5957 a unique base value if the frame pointer is not eliminated
5958 to the stack pointer.
5960 2018-03-01 Tom de Vries <tom@codesourcery.com>
5962 PR rtl-optimization/83327
5963 * lra-int.h (hard_regs_spilled_into): Declare.
5964 * lra.c (hard_regs_spilled_into): Define.
5965 (init_reg_info): Init hard_regs_spilled_into.
5966 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5967 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5968 (process_bb_lives): Handle hard_regs_spilled_into.
5969 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5970 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5972 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5974 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5975 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5976 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5977 * config/rs6000/aix72.h: New file.
5979 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5981 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5982 instead of warning_at with conditional singular and plural messages
5986 * stor-layout.c (update_alignment_for_field): For
5987 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5988 && !DECL_PACKED (field), do the alignment update, just use
5989 only desired_align instead of MAX (type_align, desired_align)
5991 (place_field): Don't do known_align < desired_align handling
5992 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5993 is non-NULL, instead do it after rli->prev_field handling and
5994 only if not within a bitfield word. For DECL_PACKED (field)
5995 use type_align of BITS_PER_UNIT.
5997 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5999 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6000 superfluous parentheses and trailing spaces.
6002 2018-02-28 Richard Biener <rguenther@suse.de>
6004 PR tree-optimization/84584
6005 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6006 SCoPs with fake exit edge.
6008 2018-02-28 Martin Liska <mliska@suse.cz>
6011 * timevar.c (timer::print): Fix format to properly print 100%
6014 2018-02-28 Richard Biener <rguenther@suse.de>
6017 * genmatch.c (capture_info::walk_match): Do not mark
6018 captured expressions without operands as expr_p given
6019 they act more like predicates and should be subject to
6020 "lost tail" side-effect preserving.
6022 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6024 PR rtl-optimization/81611
6025 * auto-inc-dec.c (attempt_change): Move dead note from
6026 mem_insn if it's the next use of regno
6027 (find_address): Take address use of reg holding
6028 non-incremented value. Add parm to limit search to the named
6030 (merge_in_block): Attempt to use a mem insn that is the next
6031 use of the original regno.
6033 2018-02-27 Martin Sebor <msebor@redhat.com>
6036 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6037 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6039 2018-02-27 Martin Sebor <msebor@redhat.com>
6041 PR translation/84207
6042 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6043 n argument to unsigned HOST_WIDE_INT.
6044 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6045 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6046 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6047 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6049 2018-02-27 Richard Biener <rguenther@suse.de>
6051 PR tree-optimization/84512
6052 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6053 Do not use the estimate returned from record_stmt_cost for
6054 the scalar iteration cost but sum properly using add_stmt_cost.
6056 2018-02-27 Richard Biener <rguenther@suse.de>
6058 PR tree-optimization/84466
6059 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6060 Adjust last change to less strictly validate use operands.
6062 2018-02-27 Martin Liska <mliska@suse.cz>
6064 PR gcov-profile/84548
6065 * gcov.c (process_file): Allow partial overlap and consider it
6066 also as group functions.
6067 (output_lines): Properly calculate range of lines for a group.
6069 2018-02-27 Martin Liska <mliska@suse.cz>
6071 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6072 'ggc' suffixes. Change first column width.
6073 (timer::print): Fix formatting of the column.
6075 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6077 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6078 preserve inline entry blocks for the sake of debug inline
6079 entry point markers alone.
6080 (remove_unused_locals): Suggest in comments a better place to
6081 force the preservation of inline entry blocks that are
6082 otherwise unused, but do not preserve them.
6084 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6086 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6088 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6091 * config/i386/constraints.md (Bs): Replace
6092 ix86_indirect_branch_register with
6093 TARGET_INDIRECT_BRANCH_REGISTER.
6095 * config/i386/i386.md (indirect_jump): Likewise.
6096 (tablejump): Likewise.
6097 (*sibcall_memory): Likewise.
6098 (*sibcall_value_memory): Likewise.
6099 Peepholes of indirect call and jump via memory: Likewise.
6100 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6101 (*sibcall_value_GOT_32): Likewise.
6102 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6103 (GOT_memory_operand): Likewise.
6104 (call_insn_operand): Likewise.
6105 (sibcall_insn_operand): Likewise.
6106 (GOT32_symbol_operand): Likewise.
6107 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6109 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6111 PR rtl-optimization/83496
6112 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6113 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6114 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6115 redundant insn, if any.
6116 (relax_delay_slots): Likewise.
6117 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6119 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6121 PR tree-optimization/83965
6122 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6123 that grouped statements are part of a reduction chain. Return
6124 true if the statement is not marked as a reduction itself but
6126 (vect_recog_dot_prod_pattern): Don't check whether the statement
6127 is part of a group here.
6128 (vect_recog_sad_pattern): Likewise.
6129 (vect_recog_widen_sum_pattern): Likewise.
6131 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6134 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6136 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6139 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6141 (ix86_output_indirect_function_return): New prototype.
6142 (ix86_split_simple_return_pop_internal): Likewise.
6143 * config/i386/i386.c (indirect_return_via_cx): New.
6144 (indirect_return_via_cx_bnd): Likewise.
6145 (indirect_thunk_name): Handle return va CX_REG.
6146 (output_indirect_thunk_function): Create alias for
6147 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6148 (ix86_output_indirect_jmp): Remove the bool argument.
6149 (ix86_output_indirect_function_return): New function.
6150 (ix86_split_simple_return_pop_internal): Likewise.
6151 * config/i386/i386.md (*indirect_jump): Don't pass false
6152 to ix86_output_indirect_jmp.
6153 (*tablejump_1): Likewise.
6154 (simple_return_pop_internal): Change it to define_insn_and_split.
6155 Call ix86_split_simple_return_pop_internal to split it for
6157 (simple_return_indirect_internal): Call
6158 ix86_output_indirect_function_return instead of
6159 ix86_output_indirect_jmp.
6161 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6164 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6165 memset and value initialization afterwards.
6167 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6169 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6171 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6174 * common/config/aarch64/aarch64-common.c
6175 (aarch_option_optimization_table[]): Switch
6176 off fomit-frame-pointer
6178 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6179 Chung-Ju Wu <jasonwucj@gmail.com>
6181 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6183 (store_multiple): Ditto.
6185 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6187 * config.gcc: Add --with-cpu support for nds32 target.
6188 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6189 * config/nds32/nds32.opt: Add -mcpu= option.
6191 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6193 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6194 isel=yes): Warn for these deprecated options.
6196 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6198 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6199 ISA_2_5_MASKS_EMBEDDED.
6201 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6203 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6204 p->max as pointers rather than using iterative_hash_expr.
6206 2018-02-23 Carl Love <cel@us.ibm.com>
6208 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6209 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6211 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6212 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6213 P8V_BUILTIN_VEC_VUNSIGNED2.
6215 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6218 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6219 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6221 * lra-constraints.c (curr_insn_transform): Set up
6222 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6224 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6225 * lra-eliminations.c (spill_pseudos): Ditto.
6226 (process_insn_for_elimination): Ditto.
6227 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6228 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6229 LRA_NON_CLOBBERED_ALT.
6231 2018-02-22 Martin Sebor <msebor@redhat.com>
6233 PR tree-optimization/84480
6234 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6235 to maybe_diag_stxncpy_trunc. Call it.
6236 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6237 from gimple_fold_builtin_strcpy. Print inlining stack.
6238 (handle_builtin_stxncpy): Print inlining stack.
6239 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6241 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6244 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6245 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6246 and -fcheck-pointer-bounds are used together.
6247 (indirect_thunk_prefix): New enum.
6248 (indirect_thunk_need_prefix): New function.
6249 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6250 "_nt" instead of "_bnd" for NOTRACK prefix.
6251 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6252 (output_indirect_thunk_function): Likewise.
6254 (ix86_code_end): Update output_indirect_thunk_function calls.
6255 (ix86_output_indirect_branch_via_reg): Replace
6256 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6257 (ix86_output_indirect_branch_via_push): Likewise.
6258 (ix86_output_function_return): Likewise.
6259 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6260 incompatible with -fcf-protection=branch and
6261 -fcheck-pointer-bounds.
6263 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6266 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6267 Change gcc_assert call to output_operand_lossage.
6269 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6271 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6273 2018-02-22 DJ Delorie <dj@redhat.com>
6274 Sebastian Perta <sebastian.perta@renesas.com>
6275 Oleg Endo <olegendo@gcc.gnu.org>
6277 * config/rx/rx.c (rx_rtx_costs): New function.
6278 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6280 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6282 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6284 2018-02-22 Martin Liska <mliska@suse.cz>
6287 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6288 Add "native" as a possible value.
6290 2018-02-22 Martin Liska <mliska@suse.cz>
6293 * config/i386/i386.c (ix86_option_override_internal):
6294 Add "native" as a possible value for -march and -mtune.
6296 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6299 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6300 to all type variants.
6302 PR tree-optimization/84503
6303 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6304 width as info->bitpos + info->bitsize - start.
6305 (merged_store_group::merge_overlapping): Simplify width computation.
6306 (check_no_overlap): New function.
6307 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6308 start + width and last_order of the group, fail if check_no_overlap
6310 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6311 to group if check_no_overlap fails.
6313 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6315 * config/rs6000/altivec.md: Delete contraint arguments to
6316 define_expand, define_split, and define_peephole2, and in
6317 define_insn_and_split if always unused.
6318 * config/rs6000/darwin.md: Ditto.
6319 * config/rs6000/dfp.md: Ditto.
6320 * config/rs6000/rs6000.md: Ditto.
6321 * config/rs6000/sync.md: Ditto.
6322 * config/rs6000/vector.md: Ditto.
6323 * config/rs6000/vsx.md: Ditto.
6325 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6327 * config/rs6000/altivec.md: Write output control strings as braced
6328 blocks instead of double-quoted strings.
6329 * config/rs6000/darwin.md: Ditto.
6330 * config/rs6000/rs6000.md: Ditto.
6331 * config/rs6000/vector.md: Ditto.
6332 * config/rs6000/vsx.md: Ditto.
6334 2018-02-21 Jason Merrill <jason@redhat.com>
6336 PR c++/84314 - ICE with templates and fastcall attribute.
6337 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6339 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6341 * ipa-cp.c (determine_versionability): Fix comment typos.
6343 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6346 * ipa-cp.c (determine_versionability): Do not version functions caling
6349 2018-02-21 Martin Liska <mliska@suse.cz>
6352 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6353 Add "native" as a possible value.
6354 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6355 the macro when native cpu detection is available.
6357 2018-02-21 Martin Liska <mliska@suse.cz>
6360 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6361 Add "native" as a possible value.
6362 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6363 when native cpu detection is available.
6365 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6366 Martin Sebor <msebor@redhat.com>
6368 PR tree-optimization/84478
6369 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6371 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6372 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6373 support which is conservatively correct, for 2 only stay conservative
6374 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6375 argument to the 2 argument get_range_strlen, adjust 6 arg
6376 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6378 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6379 (gimple_fold_builtin_strlen): Pass true as last argument to
6382 2018-02-20 Martin Sebor <msebor@redhat.com>
6385 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6386 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6387 (builtin_memref::builtin_memref): Factor out parts into
6388 set_base_and_offset and call it.
6390 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6393 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6394 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6395 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6396 search at the associated MODE_INT.
6398 2018-02-20 Jeff Law <law@redhat.com>
6401 PR tree-optimization/81592
6403 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6404 for range data rather than using global data.
6405 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6406 range data rather than using global data.
6407 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6408 pass it to children as needed.
6409 (struct directive::fmtresult): Similarly.
6410 (struct directive::set_width): Similarly.
6411 (struct directive::set_precision): Similarly.
6412 (format_integer, format_directive, parse_directive): Similarly.
6413 (format_none): Accept unnamed vr_values parameter.
6414 (format_percent, format_floating, format_character): Similarly.
6415 (format_string, format_plain): Similarly.
6416 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6417 the EVRP range analyzer for range data rather than using global data.
6418 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6419 gimple-ssa-evrp-analyze.h
6420 (class sprintf_dom_walker): Add after_dom_children member function.
6421 Add evrp_range_analyzer member.
6422 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6423 range analyzer as needed.
6424 (sprintf_dom_walker::after_dom_children): New member function.
6425 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6427 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6428 (evrp_range_analyzer::pop_to_marker): Likewise.
6430 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6432 PR tree-optimization/84419
6433 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6434 with the required type if its current type is compatible but
6437 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6440 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6441 after vectorization.
6443 2018-02-20 Martin Liska <mliska@suse.cz>
6446 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6447 possible values if we don't have a hint.
6449 2018-02-20 Martin Liska <mliska@suse.cz>
6453 * final.c (shorten_branches): Build align_tab array with one
6455 * opts.c (finish_options): Add alignment option limit check.
6456 (MAX_CODE_ALIGN): Likewise.
6457 (MAX_CODE_ALIGN_VALUE): Likewise.
6458 * doc/invoke.texi: Document maximum allowed option value for
6459 all -falign-* options.
6461 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6464 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6465 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6466 * var-tracking.c (emit_note_insn_var_location): Remove all references
6467 to NOTE_INSN_CALL_ARG_LOCATION.
6468 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6469 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6470 Use copy_rtx_if_shared.
6471 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6472 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6473 (dwarf2out_var_location): Remove handling of
6474 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6476 * final.c (final_scan_insn): Remove all references to
6477 NOTE_INSN_CALL_ARG_LOCATION.
6478 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6479 before dumping final insns.
6480 * except.c (emit_note_eh_region_end): Remove all references to
6481 NOTE_INSN_CALL_ARG_LOCATION.
6482 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6483 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6484 * config/arc/arc.c (hwloop_optimize): Likewise.
6485 * config/arm/arm.c (create_fix_barrier): Likewise.
6486 * config/s390/s390.c (s390_chunkify_start): Likewise.
6487 * config/sh/sh.c (find_barrier): Likewise.
6488 * config/i386/i386.c (rest_of_insert_endbranch,
6489 ix86_seh_fixup_eh_fallthru): Likewise.
6490 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6491 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6492 * config/frv/frv.c (frv_function_prologue): Likewise.
6493 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6495 (note_outside_basic_block_p): Remove all references to
6496 NOTE_INSN_CALL_ARG_LOCATION.
6497 * gengtype.c (adjust_field_rtx_def): Likewise.
6498 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6500 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6501 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6504 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6507 PR tree-optimization/84452
6508 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6509 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6512 2018-02-19 Martin Liska <mliska@suse.cz>
6515 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6517 2018-02-19 Martin Liska <mliska@suse.cz>
6518 Richard Sandiford <richard.sandiford@linaro.org>
6520 PR tree-optimization/82491
6521 * gimple-fold.c (get_base_constructor): Make earlier bail out
6524 2018-02-19 Carl Love <cel@us.ibm.com>
6526 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6527 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6529 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6530 P8V_BUILTIN_VEC_NEG.
6532 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6534 * config/rl78/rl78.md (movdf): New define expand.
6536 2018-02-19 Martin Liska <mliska@suse.cz>
6539 * doc/invoke.texi: Fix typo.
6540 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6542 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6544 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6545 handle rs6000_single_float and rs6000_double_float specially for
6548 2018-02-16 Jeff Law <law@redhat.com>
6550 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6553 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6556 * ipa-inline.c (inline_small_functions): Fix a typo.
6558 2018-02-16 Nathan Sidwell <nathan@acm.org>
6560 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6562 2018-02-16 Carl Love <cel@us.ibm.com>
6564 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6565 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6566 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6567 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6568 expansion to P8V_BUILTIN_VEC_FLOAT2.
6570 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6572 PR rtl-optimization/70023
6573 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6574 src_regno into account.
6576 2018-02-16 Carl Love <cel@us.ibm.com>
6578 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6579 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6580 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6581 * config/rs6000/rs6000.c: Remove case statements for
6582 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6583 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6584 and P9V_BUILTIN_VEC_VINSERT4B.
6585 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6586 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6587 * config/rs6000/vsx.md:
6588 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6591 2018-02-16 Carl Love <cel@us.ibm.com>
6593 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6595 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6597 * config/rs6000/rs6000-c.c: Add the definitions for
6598 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6599 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6600 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6601 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6602 definition for insert4b and define insn *insert3b_internal.
6603 * doc/extend.texi: Add documentation for vec_extract4b.
6605 2018-02-16 Nathan Sidwell <nathan@acm.org>
6607 * doc/extend.texi (Backwards Compatibility): Mention friend
6608 injection. Note for-scope is deprecated.
6609 * doc/invoke.texi (-ffriend-injection): Deprecate.
6611 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6613 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6614 that moved to I2, also allow destinations that are a paradoxical
6615 subreg (instead of a normal reg).
6617 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6620 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6623 2018-02-16 Richard Biener <rguenther@suse.de>
6625 PR tree-optimization/84037
6626 PR tree-optimization/84016
6628 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6629 Adjust vec_construct for the fact we need additional higher latency
6630 128bit inserts for AVX256 and AVX512 vector builds.
6631 (ix86_add_stmt_cost): Scale vector construction cost for
6634 2018-02-16 Richard Biener <rguenther@suse.de>
6636 PR tree-optimization/84417
6637 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6638 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6639 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6641 2018-02-16 Martin Liska <mliska@suse.cz>
6644 * internal-fn.def (ASAN_CHECK): Set proper flags.
6645 (ASAN_MARK): Likewise.
6647 2018-02-16 Julia Koval <julia.koval@intel.com>
6649 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6650 from PTA_CANNONLAKE.
6652 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6655 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6656 Use ++iter rather than iter++ for std::list iterators.
6657 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6658 defer deleting them until all nodes in the forest are processed. Do
6659 free even leaf nodes. Change to_process into auto_vec.
6662 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6663 * vec.h (vec_default_construct): Use memset instead of placement new
6664 if BROKEN_VALUE_INITIALIZATION is defined.
6665 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6666 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6669 PR rtl-optimization/83723
6670 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6671 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6672 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6674 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6676 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6678 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6680 PR rtl-optimization/81443
6681 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6682 from inner REGs to paradoxical SUBREGs.
6684 2018-02-16 Richard Biener <rguenther@suse.de>
6686 PR tree-optimization/84399
6687 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6688 For operands we can analyze at their definition make sure we can
6689 analyze them at each use as well.
6691 2018-02-16 Richard Biener <rguenther@suse.de>
6693 PR tree-optimization/84190
6694 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6695 volatile accesses if the decl isn't volatile.
6697 2018-02-15 Jason Merrill <jason@redhat.com>
6699 PR c++/84314 - ICE with templates and fastcall attribute.
6700 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6701 TYPE_CANONICAL on an existing type.
6703 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6705 PR tree-optimization/84383
6706 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6707 dstoff nor call operand_equal_p if dstbase is NULL.
6709 PR tree-optimization/84334
6710 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6711 also a CONSTANT_CLASS_P, punt.
6713 2018-02-14 Jim Wilson <jimw@sifive.com>
6715 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6716 first SMALL_OPERAND check. New local min_second_step. Move assert
6717 to where locals are set. Add TARGET_RVC support.
6718 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6720 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6722 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6724 2018-02-14 Martin Sebor <msebor@redhat.com>
6726 PR tree-optimization/83698
6727 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6728 arrays constrain the offset range to their bounds.
6729 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6730 (builtin_access::overlap): Avoid setting the size of overlap if it's
6732 (maybe_diag_overlap): Also consider arrays when deciding what values
6733 of offsets to include in diagnostics.
6735 2018-02-14 Martin Sebor <msebor@redhat.com>
6738 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6739 that correspond to the kind of a declaration.
6741 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6744 * config/pa/pa.md: Load address of PIC label using the linkage table
6745 if the label is nonlocal.
6747 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6749 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6750 warning message if user requests -maltivec=be.
6751 * doc/invoke.texi: Document deprecation of -maltivec=be.
6753 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6756 * config/rs6000/rs6000-c.c: Update definitions for
6757 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6758 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6760 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6763 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6764 add _get_ssp intrinsics. Remove argument from
6765 __builtin_ia32_rdssp[d|q].
6766 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6767 * config/i386/i386-builtin.def: Remove argument from
6768 __builtin_ia32_rdssp[d|q].
6769 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6770 ix86_expand_special_args_builtin for _rdssp[d|q].
6771 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6772 Clear register before usage.
6773 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6774 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6776 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6778 PR tree-optimization/84357
6779 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6780 operand 1 of an ARRAY_REF too.
6782 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6785 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6786 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6788 (set_of_reg): New struct.
6789 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6790 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6791 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6793 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6794 Split into bitclr, bitset, bitinvert patterns if appropriate.
6795 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6796 use rx_fuse_in_memory_bitop.
6797 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6798 to named insn, correct maximum insn length.
6800 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6803 * machmode.def: Define a complex mode for PARTIAL_INT.
6804 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6806 * doc/rtl.texi: Document CSPImode.
6807 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6809 (msp430_hard_regno_nregs_with_padding): Likewise.
6811 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6814 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6816 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6818 PR rtl-optimization/84169
6819 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6820 we generated a parallel as new i3 and we split that to new i2 and i3
6821 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6822 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6823 those to i2, not i1. Partially rewrite this scan code.
6825 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6828 * stor-layout.c (place_field): For variable length fields, adjust
6829 offset_align afterwards not just based on the field's alignment,
6830 but also on the size.
6833 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6834 of exps and logs in the use_exp2 case.
6836 2018-02-13 Jeff Law <law@redhat.com>
6838 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6841 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6844 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6848 * common.opt (gas-loc-support, gas-locview-support): New.
6849 (ginline-points, ginternal-reset-location-views): New.
6850 * doc/invoke.texi: Document them. Use @itemx where intended.
6851 (gvariable-location-views): Adjust.
6852 * target.def (reset_location_view): New.
6853 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6854 (TARGET_RESET_LOCATION_VIEW): New.
6855 * doc/tm.texi: Rebuilt.
6856 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6857 (dwarf2out_default_as_locview_support): New.
6858 (output_asm_line_debug_info): Use option variables.
6859 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6860 (output_loc_list): Likewise.
6861 (add_high_low_attributes): Check option variables.
6862 Don't output entry view attribute in strict mode.
6863 (gen_inlined_subroutine_die): Check option variables.
6864 (dwarf2out_inline_entry): Likewise.
6865 (init_sections_and_labels): Likewise.
6866 (dwarf2out_early_finish): Likewise.
6867 (maybe_reset_location_view): New, from...
6868 (dwarf2out_var_location): ... here. Call it.
6869 * debug.h (dwarf2out_default_as_loc_support): Declare.
6870 (dwarf2out_default_as_locview_support): Declare.
6871 * hooks.c (hook_int_rtx_insn_0): New.
6872 * hooks.h (hook_int_rtx_insn_0): Declare.
6873 * toplev.c (process_options): Take -gas-loc-support and
6874 -gas-locview-support from dwarf2out. Enable
6875 -gvariable-location-views by default only with locview
6876 assembler support. Enable -ginternal-reset-location-views by
6877 default only if the target defines the corresponding hook.
6878 Enable -ginline-points by default if location views are
6879 enabled; force it disabled if statement frontiers are
6881 * tree-inline.c (expand_call_inline): Check option variables.
6882 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6884 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6886 PR tree-optimization/84321
6887 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6888 handling. Also check whether the anti-range contains any values
6889 that satisfy the mask; switch to a VR_RANGE if not.
6891 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6894 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6896 2018-02-13 Martin Jambor <mjambor@suse.cz>
6899 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6900 of call statements, also set location of a load to a temporary.
6902 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6904 * config/rl78/rl78.c (add_vector_labels): New function.
6905 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6906 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6907 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6908 which checks that no arguments are passed.
6909 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6910 * doc/extend.texi: Documentation for the new attribute.
6912 2018-02-13 Andreas Schwab <schwab@suse.de>
6914 * config/riscv/linux.h (CPP_SPEC): Define.
6916 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6919 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6920 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6921 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6922 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6923 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6924 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6925 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6926 temporarily for AES and PCLMUL builtins.
6928 PR tree-optimization/84339
6929 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6930 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6934 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6935 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6936 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6938 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6940 * omp-simd-clone.h: New file.
6941 * omp-simd-clone.c: Include omp-simd-clone.h.
6942 (expand_simd_clones): No longer static.
6943 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6944 cgraph.h and omp-simd-clone.h.
6945 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6946 (vect_recog_widen_shift_pattern): Formatting fix.
6947 (vect_pattern_recog_1): Don't check optab for calls.
6950 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6951 operands[2] into a REG before using gen_lowpart on it.
6953 2018-02-12 Jeff Law <law@redhat.com>
6956 * config/sh/sh.c (find_barrier): Consider a sibling call
6959 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6960 successfully back substituting a reg.
6962 2018-02-12 Richard Biener <rguenther@suse.de>
6964 PR tree-optimization/84037
6965 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6966 parameter, move visited init to caller.
6967 (vect_slp_analyze_operations): Separate cost from validity
6968 check, initialize visited once for all instances.
6969 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6971 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6973 (vect_model_promotion_demotion_cost): Likewise.
6974 (vectorizable_bswap): Guard cost modeling with !slp_node
6975 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6977 (vectorizable_call): Likewise.
6978 (vectorizable_conversion): Likewise.
6979 (vectorizable_assignment): Likewise.
6980 (vectorizable_shift): Likewise.
6981 (vectorizable_operation): Likewise.
6982 (vectorizable_store): Likewise.
6983 (vectorizable_load): Likewise.
6984 (vectorizable_condition): Likewise.
6985 (vectorizable_comparison): Likewise.
6987 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6990 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6991 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6992 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6994 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6997 * params.def (inline-min-speedup): Increase from 8 to 15.
6998 (max-inline-insns-auto): Decrease from 40 to 30.
6999 * ipa-split.c (consider_split): Add some buffer for function to
7000 be considered inlining candidate.
7001 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7004 2018-02-12 Richard Biener <rguenther@suse.de>
7006 PR tree-optimization/84037
7007 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7008 matched stmts if we cannot swap the non-matched ones.
7010 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7012 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7013 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7014 _mm_maskz_scalef_round_ss): New intrinsics.
7015 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7016 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7017 __builtin_ia32_scalefss_round): Remove.
7018 (__builtin_ia32_scalefsd_mask_round,
7019 __builtin_ia32_scalefss_mask_round): New intrinsics.
7020 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7021 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7022 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7023 "<round_constraint>")): Changed to ...
7024 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7025 "<round_scalar_constraint>")): ... this.
7026 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7027 %0, %1, %2<round_op3>}"): Changed to ...
7028 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7029 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7030 %2<round_scalar_mask_op3>}"): ... this.
7031 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7033 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7035 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7036 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7037 (_mm_maskz_sqrt_round_ss): New intrinsics.
7038 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7039 (__builtin_ia32_sqrtsd_mask_round)
7040 (__builtin_ia32_sqrtss_mask_round): New builtins.
7041 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7042 (__builtin_ia32_sqrtss_round): Remove.
7043 (__builtin_ia32_sqrtsd_mask_round)
7044 (__builtin_ia32_sqrtss_mask_round): New builtins.
7045 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7046 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7047 ((match_operand:VF_128 1 "vector_operand"
7048 "xBm,<round_constraint>")): Changed to ...
7049 ((match_operand:VF_128 1 "vector_operand"
7050 "xBm,<round_scalar_constraint>")): ... this.
7051 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7052 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7053 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7054 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7055 %<iptr>1<round_scalar_mask_op3>}): ... this.
7056 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7057 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7059 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7062 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7063 Cast vec_cmpeq result to correct type.
7064 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7065 Cast vec_cmpgt result to correct type.
7067 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7069 * final.c (final_scan_insn_1): Renamed from...
7070 (final_scan_insn): ... this. New wrapper, to recover
7071 seen from the outermost call in recursive ones.
7072 * config/sparc/sparc.c (output_return): Drop seen from call.
7073 (output_sibcall): Likewise.
7074 * config/visium/visium.c (output_branch): Likewise.
7076 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7078 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7081 2018-02-10 Alan Modra <amodra@gmail.com>
7084 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7085 Specify LR as an input.
7087 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7090 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7091 remove_member_access_dummy_vars): New functions.
7092 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7093 lower_omp_1, execute_lower_omp): Use them.
7095 PR rtl-optimization/84308
7096 * shrink-wrap.c (spread_components): Release todo vector.
7098 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7100 PR rtl-optimization/57193
7101 * ira-color.c (struct allocno_color_data): Add member
7102 conflict_allocno_hard_prefs.
7103 (update_conflict_allocno_hard_prefs): New.
7104 (bucket_allocno_compare_func): Add a preference based on
7105 conflict_allocno_hard_prefs.
7106 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7107 (color_allocnos): Remove a dead code. Initiate
7108 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7110 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7113 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7114 constraint from =wa to wa. Avoid a subreg on the output operand,
7115 instead use a pseudo and subreg it in a move.
7116 (p9_xxbrd_<mode>): Changed to ...
7117 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7118 (p9_xxbrd_v2df): New expander.
7119 (p9_xxbrw_<mode>): Changed to ...
7120 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7121 (p9_xxbrw_v4sf): New expander.
7123 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7125 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7127 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7130 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7131 multiply in 32-bit mode.
7132 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7133 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7136 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7138 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7139 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7140 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7141 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7143 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7146 * dwarf2out.c (is_trivial_indirect_ref): New function.
7147 (dwarf2out_late_global_decl): Do not generate a location
7148 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7149 and that are not defined in the current unit.
7151 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7153 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7154 instead of a libcall for UNORDERED.
7156 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7159 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7160 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7162 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7165 * config/s390/s390.c (s390_set_current_function): Invoke
7166 s390_indirect_branch_settings also if fndecl didn't change.
7168 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7170 * config/rs6000/rs6000.md (blockage): Set length to zero.
7172 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7174 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7176 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7179 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7180 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7184 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7185 PARALLEL incoming that failed vt_get_decl_and_offset check.
7188 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7189 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7191 (get_variable_section): For decls in named .bss* sections pass true as
7192 second argument to bss_initializer_p.
7194 2018-02-09 Marek Polacek <polacek@redhat.com>
7195 Jakub Jelinek <jakub@redhat.com>
7198 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7199 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7200 Sync some changes from cxx_fold_indirect_ref.
7202 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7204 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7206 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7207 (BLOCK_INLINE_ENTRY_LABEL): New.
7208 (dwarf2out_var_location): Disregard inline entry markers.
7209 (inline_entry_data): New struct.
7210 (inline_entry_data_hasher): New hashtable type.
7211 (inline_entry_data_hasher::hash): New.
7212 (inline_entry_data_hasher::equal): New.
7213 (inline_entry_data_table): New variable.
7214 (add_high_low_attributes): Add DW_AT_entry_pc and
7215 DW_AT_GNU_entry_view attributes if a pending entry is found
7216 in inline_entry_data_table. Add old entry_pc attribute only
7217 if debug nonbinding markers are disabled.
7218 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7219 markers are enabled.
7220 (block_within_block_p, dwarf2out_inline_entry): New.
7221 (dwarf2out_finish): Check that no entries remained in
7222 inline_entry_data_table.
7223 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7224 (final_scan_insn, notice_source_line): Likewise.
7225 (rest_of_clean_state): Skip inline entry markers.
7226 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7228 * gimple.c (gimple_build_debug_inline_entry): New.
7229 * gimple.h (enum gimple_debug_subcode): Add
7230 GIMPLE_DEBUG_INLINE_ENTRY.
7231 (gimple_build_debug_inline_entry): Declare.
7232 (gimple_debug_inline_entry_p): New.
7233 (gimple_debug_nonbind_marker_p): Adjust.
7234 * insn-notes.def (INLINE_ENTRY): New.
7235 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7236 inline entry marker notes.
7237 (print_insn): Likewise.
7238 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7239 (INSN_DEBUG_MARKER_KIND): Likewise.
7240 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7241 * tree-inline.c (expand_call_inline): Build and insert
7242 debug_inline_entry stmt.
7243 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7244 inline entry blocks early, if nonbind markers are enabled.
7245 (dump_scope_block): Dump fragment info.
7246 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7247 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7248 (gimple_build_debug_inline_entry): New.
7249 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7250 Enable/disable inline entry points too.
7251 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7252 (DEBUG_INSN): Describe inline entry markers.
7254 * common.opt (gvariable-location-views): New.
7255 (gvariable-location-views=incompat5): New.
7256 * config.in: Rebuilt.
7257 * configure: Rebuilt.
7258 * configure.ac: Test assembler for view support.
7259 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7260 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7261 * dwarf2out.c (var_loc_view): New typedef.
7262 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7263 (dwarf2out_locviews_in_attribute): New.
7264 (dwarf2out_locviews_in_loclist): New.
7265 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7266 (enum dw_line_info_opcode): Add LI_adv_address.
7267 (struct dw_line_info_table): Add view.
7268 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7269 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7270 (zero_view_p): New variable.
7271 (ZERO_VIEW_P): New macro.
7272 (output_asm_line_debug_info): New.
7273 (struct var_loc_node): Add view.
7274 (add_AT_view_list, AT_loc_list): New.
7275 (add_var_loc_to_decl): Add view param. Test it against last.
7276 (new_loc_list): Add view params. Record them.
7277 (AT_loc_list_ptr): Handle loc and view lists.
7278 (view_list_to_loc_list_val_node): New.
7279 (print_dw_val): Handle dw_val_class_view_list.
7280 (size_of_die): Likewise.
7281 (value_format): Likewise.
7282 (loc_list_has_views): New.
7283 (gen_llsym): Set vl_symbol too.
7284 (maybe_gen_llsym, skip_loc_list_entry): New.
7285 (dwarf2out_maybe_output_loclist_view_pair): New.
7286 (output_loc_list): Output view list or entries too.
7287 (output_view_list_offset): New.
7288 (output_die): Handle dw_val_class_view_list.
7289 (output_dwarf_version): New.
7290 (output_compilation_unit_header): Use it.
7291 (output_skeleton_debug_sections): Likewise.
7292 (output_rnglists, output_line_info): Likewise.
7293 (output_pubnames, output_aranges): Update version comments.
7294 (output_one_line_info_table): Output view numbers in asm comments.
7295 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7296 Call maybe_gen_llsym.
7297 (loc_list_from_tree_1): Adjust.
7298 (add_AT_location_description): Create view list attribute if
7299 needed, check it's absent otherwise.
7300 (convert_cfa_to_fb_loc_list): Adjust.
7301 (maybe_emit_file): Call output_asm_line_debug_info for test.
7302 (dwarf2out_var_location): Reset views as needed. Precompute
7303 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7304 attribute. Set view.
7305 (new_line_info_table): Reset next view.
7306 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7307 (dwarf2out_source_line): Likewise. Output view resets and labels to
7308 the assembler, or select appropriate line info opcodes.
7309 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7310 (optimize_string_length): Catch it. Adjust.
7311 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7312 dw_val_class_view_list, and remove it if no longer needed.
7313 (hash_loc_list): Hash view numbers.
7314 (loc_list_hasher::equal): Compare them.
7315 (optimize_location_lists): Check whether a view list symbol is
7316 needed, and whether the locview attribute is present, and
7317 whether they match. Remove the locview attribute if no longer
7319 (index_location_lists): Call skip_loc_list_entry for test.
7320 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7321 Use output_dwarf_version.
7322 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7323 (struct dw_val_node): Add val_view_list.
7324 * final.c (SEEN_NEXT_VIEW): New.
7325 (set_next_view_needed): New.
7326 (clear_next_view_needed): New.
7327 (maybe_output_next_view): New.
7328 (final_start_function): Rename to...
7329 (final_start_function_1): ... this. Take pointer to FIRST,
7330 add SEEN parameter. Emit param bindings in the initial view.
7331 (final_start_function): Reintroduce SEEN-less interface.
7332 (final): Rename to...
7333 (final_1): ... this. Take SEEN parameter. Output final pending
7334 next view at the end.
7335 (final): Reintroduce seen-less interface.
7336 (final_scan_insn): Output pending next view before switching
7337 sections or ending a block. Mark the next view as needed when
7338 outputting variable locations. Notify debug backend of section
7339 changes, and of location view changes.
7340 (rest_of_handle_final): Adjust.
7341 * toplev.c (process_options): Autodetect value for debug variable
7342 location views option. Warn on incompat5 without -gdwarf-5.
7343 * doc/invoke.texi (gvariable-location-views): New.
7344 (gvariable-location-views=incompat5): New.
7345 (gno-variable-location-views): New.
7347 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7349 PR tree-optimization/84136
7350 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7351 that the result of find_edge is non-NULL.
7353 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7356 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7357 storing integer register in SImode. Fix cost of 256 and 512
7358 byte aligned SSE register store.
7360 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7362 * config/i386/i386.c (ix86_multiplication_cost): Fix
7363 multiplication cost for TARGET_AVX512DQ.
7365 2018-02-08 Marek Polacek <polacek@redhat.com>
7367 PR tree-optimization/84238
7368 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7371 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7373 PR tree-optimization/84265
7374 * tree-vect-stmts.c (vectorizable_store): Don't treat
7375 VMAT_CONTIGUOUS accesses as grouped.
7376 (vectorizable_load): Likewise.
7378 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7380 PR tree-optimization/81635
7381 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7382 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7383 (test_round_for_mask): New functions.
7384 (wide_int_cc_tests): Call test_round_for_mask.
7385 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7386 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7387 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7388 range returned by get_range_info.
7390 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7393 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7394 * symtab.c: Include builtins.h
7395 (symtab_node::output_to_lto_symbol_table_p): Move here
7396 from lto-streamer-out.c:output_symbol_p.
7397 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7398 (output_symbol_p): Move all logic to symtab.c
7399 (produce_symtab): Update.
7401 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7403 * config/s390/s390-opts.h (enum indirect_branch): Define.
7404 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7405 (s390_indirect_branch_via_thunk)
7406 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7407 (enum s390_indirect_branch_type): Define.
7408 * config/s390/s390.c (struct s390_frame_layout, struct
7409 machine_function): Remove.
7410 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7411 (indirect_branch_table_label_no, indirect_branch_table_name):
7413 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7414 (enum s390_indirect_branch_option): Define.
7415 (s390_return_addr_from_memory): New function.
7416 (s390_handle_string_attribute): New function.
7417 (s390_attribute_table): Add new attribute handler.
7418 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7419 (s390_indirect_branch_via_thunk): New function.
7420 (s390_indirect_branch_via_inline_thunk): New function.
7421 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7422 sibling call optimization for non z10 compiles.
7423 (s390_emit_call): Force indirect branch target to be a single
7424 register. Add r1 clobber for non-z10 compiles.
7425 (s390_emit_epilogue): Emit return jump via return_use expander.
7426 (s390_reorg): Handle JUMP_INSNs as execute targets.
7427 (s390_option_override_internal): Perform validity checks for the
7428 new command line options.
7429 (s390_indirect_branch_attrvalue): New function.
7430 (s390_indirect_branch_settings): New function.
7431 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7432 (s390_output_indirect_thunk_function): New function.
7433 (s390_code_end): Implement target hook.
7434 (s390_case_values_threshold): Implement target hook.
7435 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7437 * config/s390/s390.h (struct s390_frame_layout)
7438 (struct machine_function): Move here from s390.c.
7439 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7440 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7441 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7442 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7443 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7444 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7445 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7446 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7447 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7448 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7449 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7450 (mnemonic attribute): Add values which aren't recognized
7452 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7453 pattern for branch conversion. Fix mnemonic attribute.
7454 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7455 indirect branch via thunk if requested.
7456 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7457 ("*indirect_jump"): Disable for branch conversion using out of
7459 ("indirect_jump_via_thunk<mode>_z10")
7460 ("indirect_jump_via_thunk<mode>")
7461 ("indirect_jump_via_inlinethunk<mode>_z10")
7462 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7463 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7464 ("casesi_jump_via_inlinethunk<mode>_z10")
7465 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7466 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7467 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7468 ("*indirect2_jump"): Disable for branch conversion.
7469 ("casesi_jump"): Turn into expander and expand patterns for branch
7471 ("return_use"): New expander.
7472 ("*return"): Emit return via thunk and rename it to ...
7473 ("*return<mode>"): ... this one.
7474 * config/s390/s390.opt: Add new options and and enum for the
7477 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7479 * lra-constraints.c (match_reload): Unconditionally use
7480 gen_lowpart_SUBREG, rather than selecting between that
7481 and equivalent gen_rtx_SUBREG code.
7483 2018-02-08 Richard Biener <rguenther@suse.de>
7485 PR tree-optimization/84233
7486 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7487 changed flag instead of boguously re-using phi_inserted.
7489 2018-02-08 Martin Jambor <mjambor@suse.cz>
7491 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7492 static local variables.
7494 2018-02-08 Richard Biener <rguenther@suse.de>
7496 PR tree-optimization/84278
7497 * tree-vect-stmts.c (vectorizable_store): When looking for
7498 smaller vector types to perform grouped strided loads/stores
7499 make sure the mode is supported by the target.
7500 (vectorizable_load): Likewise.
7502 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7504 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7505 Increase LDP/STP opportunities by adding adjacent callee-saves.
7507 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7509 PR rtl-optimization/84068
7510 PR rtl-optimization/83459
7511 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7513 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7515 PR tree-optimization/84224
7516 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7517 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7520 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7523 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7524 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7525 count, remove one USE.
7527 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7529 * doc/install.texi (Configuration): Document the
7530 --with-long-double-format={ibm,ieee} PowerPC configuration
7534 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7535 Convert from define_expand to be define_insn_and_split. Rework
7536 float/double/_Float128 conversions to QI/HI/SImode to work with
7537 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7538 conversions to QI/HImode types did a store and then a load to
7539 truncate the value. For conversions to VSX registers, don't split
7540 the insn, instead emit the code directly. Use the code iterator
7541 any_fix to combine signed and unsigned conversions.
7542 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7543 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7544 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7545 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7546 (fix_<mode>di2_hw): Likewise.
7547 (fixuns_<mode>di2_hw): Likewise.
7548 (fix_<mode>si2_hw): Likewise.
7549 (fixuns_<mode>si2_hw): Likewise.
7550 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7551 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7552 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7553 fix<uns>_trunc<SFDF:mode>si2_p8.
7554 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7556 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7557 (fix<uns>_<mode>_mem): Likewise.
7558 (fctiw<u>z_<mode>_mem): Likewise.
7559 (fix<uns>_<mode>_mem): Likewise.
7560 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7561 the register allocator from doing a direct move to the GPRs to do
7562 a store, and instead use the ISA 3.0 store byte/half-word from
7563 vector register instruction. For IEEE 128-bit floating point,
7564 also optimize stores of 32-bit ints.
7565 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7567 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7569 * genextract.c (push_pathstr_operand): New function to support
7571 (walk_rtx): Call push_pathstr_operand.
7572 (print_path): Support [a-zA-Z].
7574 2018-02-07 Richard Biener <rguenther@suse.de>
7576 PR tree-optimization/84037
7577 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7578 (cse_and_gimplify_to_preheader): Declare.
7579 (vect_get_place_in_interleaving_chain): Likewise.
7580 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7582 (_loop_vec_info::~_loop_vec_info): Delete it.
7583 (cse_and_gimplify_to_preheader): New function.
7584 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7585 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7586 (vectorizable_load): Likewise. For grouped stores always base
7587 the IV on the first element.
7588 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7589 condition before gimplifying.
7591 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7593 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7594 *DIV_EXPR and *MOD_EXPR.
7596 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7599 * config/i386/i386.c (ix86_option_override_internal): Mask out
7600 the CF_SET bit when checking -fcf-protection.
7602 2018-02-07 Tom de Vries <tom@codesourcery.com>
7605 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7608 2018-02-07 Richard Biener <rguenther@suse.de>
7610 PR tree-optimization/84204
7611 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7614 PR tree-optimization/84205
7615 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7616 special-case isl_ast_op_zdiv_r.
7618 PR tree-optimization/84223
7619 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7620 Only add conditions from within the region.
7621 (gather_bbs::after_dom_children): Adjust.
7623 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7626 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7627 * config/avr/avr.md: Only post-reload split REG-REG moves if
7628 either register is GENERAL_REG_P.
7630 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7632 PR tree-optimization/84235
7633 * tree-ssa-scopedtables.c
7634 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7635 if the subtraction is performed in floating point type where NaNs are
7636 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7637 build 1. Formatting fix.
7639 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7642 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7643 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7644 and skip it regardless of bb boundaries. Use CALL_P macro,
7645 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7646 unnecessarily, formatting fix.
7648 2018-02-06 Michael Collison <michael.collison@arm.com>
7650 * config/arm/thumb2.md:
7651 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7652 (*thumb_mov_notscc): Ditto.
7654 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7657 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7658 unsigned_fix, not "s".
7660 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7662 * configure.ac (gcc_fn_eh_frame_ro): New function.
7663 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7664 correct .eh_frame permissions.
7665 * configure: Regenerate.
7667 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7669 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7672 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7674 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7675 Display warning message for -mno-speculate-indirect-jumps.
7677 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7679 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7681 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7683 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7685 PR tree-optimization/84225
7686 * tree-eh.c (find_trapping_overflow): Only call
7687 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7689 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7692 * config/i386/i386.c: Reimplement the check of possible options
7693 -mibt/-mshstk conbination. Change error messages.
7694 * doc/invoke.texi: Fix a typo: remove extra '='.
7696 2018-02-06 Marek Polacek <polacek@redhat.com>
7698 PR tree-optimization/84228
7699 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7701 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7704 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7705 emitted arch directives.
7706 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7707 __ARM_FEATURE_COPROC before changing architectures.
7709 2018-02-06 Richard Biener <rguenther@suse.de>
7711 * config/i386/i386.c (print_reg): Fix typo.
7712 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7714 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7716 * configure: Regenerate.
7718 2018-02-05 Martin Sebor <msebor@redhat.com>
7720 PR tree-optimization/83369
7721 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7724 2018-02-05 Martin Liska <mliska@suse.cz>
7726 * doc/invoke.texi: Cherry-pick upstream r323995.
7728 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7730 * ira.c (ira_init_register_move_cost): Adjust comment.
7732 2018-02-05 Martin Liska <mliska@suse.cz>
7734 PR gcov-profile/84137
7735 * doc/gcov.texi: Fix typo in documentation.
7737 2018-02-05 Martin Liska <mliska@suse.cz>
7739 PR gcov-profile/83879
7740 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7741 using dlopen functionality.
7743 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7745 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7746 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7747 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7748 _mm_maskz_range_round_ss): New intrinsics.
7749 (__builtin_ia32_rangesd128_round)
7750 (__builtin_ia32_rangess128_round): Remove.
7751 (__builtin_ia32_rangesd128_mask_round,
7752 __builtin_ia32_rangess128_mask_round): New builtins.
7753 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7754 __builtin_ia32_rangess128_round): Remove.
7755 (__builtin_ia32_rangesd128_mask_round,
7756 __builtin_ia32_rangess128_mask_round): New builtins.
7757 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7758 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7759 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7760 "<round_saeonly_constraint>")): Changed to ...
7761 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7762 "<round_saeonly_scalar_constraint>")): ... this.
7763 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7764 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7765 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7766 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7767 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7769 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7771 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7773 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7774 Remove all values except native, 8540 and 8548.
7776 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7778 * config/i386/i386.c (ix86_output_function_return): Pass
7779 INVALID_REGNUM, instead of -1, as invalid register number to
7780 indirect_thunk_name and output_indirect_thunk.
7782 2018-02-02 Julia Koval <julia.koval@intel.com>
7784 * config.gcc: Add -march=icelake.
7785 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7786 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7787 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7788 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7789 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7790 (processor_target_table): Add icelake.
7791 (ix86_option_override_internal): Handle new PTAs.
7792 (get_builtin_code_for_version): Handle icelake.
7793 (M_INTEL_COREI7_ICELAKE): New.
7794 (fold_builtin_cpu): Handle icelake.
7795 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7796 * doc/invoke.texi: Add -march=icelake.
7798 2018-02-02 Julia Koval <julia.koval@intel.com>
7800 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7801 to wide_int_bitmask.
7802 * wide-int-bitmask.h: New.
7804 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7807 * config/i386/i386.md: Replace Pmode with word_mode in
7808 builtin_setjmp_setup and builtin_longjmp to support x32.
7810 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7814 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7816 (rs6000_supported_cpu_names): New static variable.
7817 (linux_cpu_translation_table): Likewise.
7818 (elf_platform) <cpu>: Define new static variable and use it.
7819 Translate kernel AT_PLATFORM name to canonical name if needed.
7820 Error if platform name is unknown.
7822 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7825 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7827 2018-02-01 Jeff Law <law@redhat.com>
7830 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7831 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7832 the scratch if RELEASE_VIA_POP is false.
7833 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7834 If we have to save a temporary register, decrement SIZE appropriately.
7835 Pass new arguments to release_scratch_register_on_entry.
7836 (ix86_adjust_stack_and_probe): Likewise.
7837 (ix86_emit_probe_stack_range): Pass new arguments to
7838 release_scratch_register_on_entry.
7840 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7842 PR rtl-optimization/84157
7843 * combine.c (change_zero_ext): Use REG_P predicate in
7844 front of HARD_REGISTER_P predicate.
7846 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7848 * config/avr/avr.c (avr_option_override): Move disabling of
7849 -fdelete-null-pointer-checks to...
7850 * common/config/avr/avr-common.c (avr_option_optimization_table):
7853 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7855 PR tree-optimization/81635
7856 * tree-data-ref.c (split_constant_offset_1): For types that
7857 wrap on overflow, try to use range info to prove that wrapping
7860 2018-02-01 Renlin Li <renlin.li@arm.com>
7863 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7865 (aarch64_register_move_cost): Likewise.
7866 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7868 (REG_CLASS_NAMES): Likewise.
7869 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7870 TAILCALL_ADDR_REGS. Remove IP registers.
7871 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7873 2018-02-01 Richard Biener <rguenther@suse.de>
7875 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7876 for specifying RPO order and allow NULL for that.
7877 * domwalk.c (dom_walker::dom_walker): Likewise.
7878 (dom_walker::walk): Handle NULL RPO order.
7879 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7881 (rewrite_update_dom_walker): Likewise.
7882 (mark_def_dom_walker): Likewise.
7884 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7886 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7887 (aarch64_maybe_expand_sve_subreg_move): Declare.
7888 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7889 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7891 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7892 that are semantically a reverse operation.
7893 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7894 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7895 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7897 (aarch64_can_change_mode_class): For big-endian, forbid changes
7898 between two SVE modes if they have different element sizes.
7900 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7902 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7903 the TImode handling for big-endian targets.
7905 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7907 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7908 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7910 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7911 Remove BSWAP handing for big-endian targets and use the form of
7912 LD1RQ appropariate for the mode.
7914 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7916 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7917 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7920 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7923 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7924 check for operands that need to go through aarch64_sve_reload_be.
7926 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7928 PR tree-optimization/81661
7929 PR tree-optimization/84117
7930 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7931 * tree-eh.c: Include gimplify.h.
7932 (find_trapping_overflow, replace_trapping_overflow,
7933 rewrite_to_non_trapping_overflow): New functions.
7934 * tree-vect-loop.c: Include tree-eh.h.
7935 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7936 * tree-data-ref.c: Include tree-eh.h.
7937 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7939 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7941 PR rtl-optimization/84123
7942 * combine.c (change_zero_ext): Check if hard register satisfies
7943 can_change_dest_mode before calling gen_lowpart_SUBREG.
7945 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7948 * ira.c (ira_init_register_move_cost): Remove assert.
7950 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7952 PR rtl-optimization/84071
7953 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7954 * doc/tm.texi: Regenerate.
7956 2018-01-31 Richard Biener <rguenther@suse.de>
7958 PR tree-optimization/84132
7959 * tree-data-ref.c (analyze_miv_subscript): Properly
7960 check whether evolution_function_is_affine_multivariate_p
7961 before calling gcd_of_steps_may_divide_p.
7963 2018-01-31 Julia Koval <julia.koval@intel.com>
7966 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7967 * config/i386/i386.md (rdpid_rex64) New.
7968 (rdpid): Make 32bit only.
7970 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7973 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7974 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7976 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7979 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7981 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7983 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7985 PR rtl-optimization/84071
7986 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7987 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7989 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7991 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7992 (arc_attribute_table): Add 'aux' attribute.
7993 (arc_in_small_data_p): Consider aux like variables.
7994 (arc_is_aux_reg_p): New function.
7995 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7996 (arc_get_aux_arg): New function.
7997 (prepare_move_operands): Handle aux-register access.
7998 (arc_handle_aux_attribute): New function.
7999 * doc/extend.texi (ARC Variable attributes): Add subsection.
8001 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8003 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8004 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8005 (arc_attribute_table): Add 'uncached' attribute.
8006 (arc_print_operand): Print '.di' flag for uncached memory
8008 (arc_in_small_data_p): Do not consider for small data the uncached
8010 (arc_is_uncached_mem_p): New function.
8011 * config/arc/predicates.md (compact_store_memory_operand): Check
8012 for uncached memory accesses.
8013 (nonvol_nonimm_operand): Likewise.
8014 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8016 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8019 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8020 falign-loops=): Add Optimization flag.
8022 2018-01-30 Jeff Law <law@redhat.com>
8025 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8026 INT_REGISTERS_SAVED. Check it prior to calling
8027 get_scratch_register_on_entry.
8028 (ix86_adjust_stack_and_probe): Similarly.
8029 (ix86_emit_probe_stack_range): Similarly.
8030 (ix86_expand_prologue): Corresponding changes.
8032 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8035 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8036 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8038 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8041 * lra-constraints.c (curr_insn_transform): Process AND in the
8044 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8046 PR rtl-optimization/83986
8047 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8048 dependence against last_pending_memory_flush in addition to
8051 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8053 PR tree-optimization/81611
8054 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8057 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8060 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8063 2018-01-30 Richard Biener <rguenther@suse.de>
8064 Jakub Jelinek <jakub@redhat.com>
8066 PR tree-optimization/84111
8067 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8068 inner loops added during recursion, as they don't have up-to-date
8071 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8074 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8075 (can_inline_edge_by_limits_p): ... here.
8076 (can_early_inline_edge_p, check_callers,
8077 update_caller_keys, update_callee_keys, recursive_inlining,
8078 add_new_edges_to_heap, speculation_useful_p,
8079 inline_small_functions,
8080 inline_small_functions, flatten_function,
8081 inline_to_all_callers_1): Update.
8083 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8085 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8088 2018-01-30 Richard Biener <rguenther@suse.de>
8090 PR tree-optimization/83008
8091 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8092 invariant and constant vector uses in stmts when they need
8095 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8098 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8099 * configure: Regenerate.
8101 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8103 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8105 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8106 Use gen_rtx_REG rather than gen_lowpart.
8108 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8110 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8111 rather than 0 when creating partial subregs.
8113 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8115 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8118 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8121 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8122 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8123 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8124 flags. This restores the settings used before the 2017-07-24.
8125 Turning off pre increment/decrement/modify allows IVOPTS to
8126 optimize DF/SF loops where the index is an int.
8128 2018-01-29 Richard Biener <rguenther@suse.de>
8129 Kelvin Nilsen <kelvin@gcc.gnu.org>
8132 * tree-vect-stmts.c (vectorizable_call): Don't call
8133 targetm.vectorize_builtin_md_vectorized_function if callee is
8136 2018-01-22 Carl Love <cel@us.ibm.com>
8138 * doc/extend.tex: Fix typo in second arg in
8139 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8141 2018-01-29 Richard Biener <rguenther@suse.de>
8143 PR tree-optimization/84086
8144 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8145 (flush_ssaname_freelist): When SSA names were released reset
8146 the SCEV hash table.
8148 2018-01-29 Richard Biener <rguenther@suse.de>
8150 PR tree-optimization/84057
8151 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8152 removed paths when removing edges.
8154 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8156 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8157 -mfunction-return=@var{choice}.
8159 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8162 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8164 (layout::print_source_line): Likewise.
8165 (test_get_line_width_without_trailing_whitespace): Add test cases.
8167 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8170 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8173 2018-01-26 Jim Wilson <jimw@sifive.com>
8175 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8177 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8180 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8182 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8183 and CMP + SUB-immediate -> SUBS.
8185 2018-01-26 Martin Sebor <msebor@redhat.com>
8187 PR tree-optimization/83896
8188 * tree-ssa-strlen.c (get_string_len): Rename...
8189 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8190 Avoid assuming length is constant.
8191 (handle_char_store): Use HOST_WIDE_INT for string length.
8193 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8196 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8197 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8199 2018-01-26 Richard Biener <rguenther@suse.de>
8201 PR rtl-optimization/84003
8202 * dse.c (record_store): Only record redundant stores when
8203 the earlier store aliases at least all accesses the later one does.
8205 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8207 PR rtl-optimization/83985
8208 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8209 REG_CFA_RESTORE insns.
8210 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8211 REG_CFA_RESTORE insns here.
8214 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8215 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8217 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8219 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8220 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8221 (arc_init): Likewise.
8222 (arc_override_options): Likewise.
8223 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8225 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8227 * config/arc/arc.h (TARGET_DBNZ): Define.
8228 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8229 properly set the tune attribute.
8230 (dbnz): Use TARGET_DBNZ guard.
8231 * config/arc/arc.opt (mtune): Add core3 option.
8233 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8235 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8236 recognize new pic like addresses.
8237 (arc_delegitimize_address): Clean up.
8239 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8241 * config/arc/arc-arches.def: Option mrf16 valid for all
8243 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8244 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8245 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8246 * config/arc/arc-tables.opt: Regenerate.
8247 * config/arc/arc.c (arc_conditional_register_usage): Handle
8248 reduced register file case.
8249 (arc_file_start): Set must have build attributes.
8250 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8252 * config/arc/arc.opt (mrf16): Add new option.
8253 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8254 * config/arc/genmultilib.awk: Handle new mrf16 option.
8255 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8256 * config/arc/t-multilib: Regenerate.
8257 * doc/invoke.texi (ARC Options): Document mrf16 option.
8259 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8261 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8262 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8263 (arc_attribute_table): Add 'secure_call' attribute.
8264 (arc_print_operand): Print secure call operand.
8265 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8267 (arc_is_secure_call_p): New function. * config/arc/arc.md
8268 (call_i): Add support for sjli instruction.
8269 (call_value_i): Likewise.
8270 * config/arc/constraints.md (Csc): New constraint.
8272 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8273 John Eric Martin <John.Martin@emmicro-us.com>
8275 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8276 * config/arc/arc.c (_arc_jli_section): New struct.
8277 (arc_jli_section): New type.
8278 (rc_jli_sections): New static variable.
8279 (arc_handle_jli_attribute): New function.
8280 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8281 (arc_file_end): New function.
8282 (TARGET_ASM_FILE_END): Define.
8283 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8284 (arc_add_jli_section): New function.
8285 (jli_call_scan): Likewise.
8286 (arc_reorg): Call jli_call_scan.
8287 (arc_output_addsi): Remove 'S' from printing asm operand.
8288 (arc_is_jli_call_p): New function.
8289 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8291 (movhi_insn): Likewise.
8292 (movsi_insn): Likewise.
8293 (movsi_set_cc_insn): Likewise.
8294 (loadqi_update): Likewise.
8295 (load_zeroextendqisi_update): Likewise.
8296 (load_signextendqisi_update): Likewise.
8297 (loadhi_update): Likewise.
8298 (load_zeroextendhisi_update): Likewise.
8299 (load_signextendhisi_update): Likewise.
8300 (loadsi_update): Likewise.
8301 (loadsf_update): Likewise.
8302 (movsicc_insn): Likewise.
8303 (bset_insn): Likewise.
8304 (bxor_insn): Likewise.
8305 (bclr_insn): Likewise.
8306 (bmsk_insn): Likewise.
8307 (bicsi3_insn): Likewise.
8308 (cmpsi_cc_c_insn): Likewise.
8309 (movsi_ne): Likewise.
8310 (movsi_cond_exec): Likewise.
8311 (clrsbsi2): Likewise.
8322 (call_i): Remove 'S' asm letter, add jli instruction.
8323 (call_value_i): Likewise.
8324 * config/arc/arc.op (mjli-always): New option.
8325 * config/arc/constraints.md (Cji): New constraint.
8326 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8328 (subsf3_fpx): Likewise.
8329 (mulsf3_fpx): Likewise.
8330 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8332 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8334 * doc/invoke.texi (ARC): Document mjli-always option.
8336 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8338 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8339 avoid addition with 0 and use incw and decw where possible.
8341 2018-01-26 Richard Biener <rguenther@suse.de>
8343 PR tree-optimization/81082
8344 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8345 association if it requires casting to unsigned.
8346 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8347 from fold_plusminus_mult_expr to catch important cases late when
8348 range info is available.
8350 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8352 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8353 * configure.ac (hidden_linkonce): New test.
8354 * configure: Regenerate.
8355 * config.in: Regenerate.
8357 2018-01-26 Julia Koval <julia.koval@intel.com>
8359 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8360 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8361 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8362 _mm_mask_bitshuffle_epi64_mask): Fix type.
8363 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8364 USI_FTYPE_V4DI_V4DI_USI): Remove.
8365 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8366 __builtin_ia32_vpshufbitqmb256_mask,
8367 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8368 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8369 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8371 2018-01-26 Alan Modra <amodra@gmail.com>
8374 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8375 UNSPEC_VBPERMQ. Sort other unspecs.
8377 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8379 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8381 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8384 * predict.c (drop_profile): Do not push/pop cfun; update also
8386 (handle_missing_profiles): Fix logic looking for zero profiles.
8388 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8391 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8392 on functions with #pragma omp declare simd or functions with simd
8394 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8395 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8396 Remove trailing \n from warning_at calls.
8398 2018-01-25 Tom de Vries <tom@codesourcery.com>
8401 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8402 for neutered workers.
8404 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8407 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8408 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8410 2018-01-24 Jeff Law <law@redhat.com>
8413 * i386.c (get_probe_interval): Move to earlier point.
8414 (ix86_compute_frame_layout): If -fstack-clash-protection and
8415 the frame is larger than the probe interval, then use pushes
8416 to save registers rather than reg->mem moves.
8417 (ix86_expand_prologue): Remove conditional for int_registers_saved
8420 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8423 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8424 min/max for never referenced object.
8426 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8429 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8431 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8432 attributes from DECL_ATTRIBUTES (decl) without affecting
8433 DECL_ATTRIBUTES (current_function_decl).
8434 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8435 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8437 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8439 PR tree-optimization/83979
8440 * fold-const.c (fold_comparison): Use constant_boolean_node
8441 instead of boolean_{true,false}_node.
8443 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8445 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8448 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8450 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8451 Simplify the clause that sets the length attribute.
8452 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8453 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8454 clause that sets the length attribute.
8455 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8457 2018-01-24 Tom de Vries <tom@codesourcery.com>
8460 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8461 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8462 Add strict parameter.
8463 (prevent_branch_around_nothing): Insert dummy insn between branch to
8464 label and label with no ptx insn inbetween.
8465 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8467 2018-01-24 Tom de Vries <tom@codesourcery.com>
8470 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8471 for neutered threads in warp.
8472 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8474 2018-01-24 Richard Biener <rguenther@suse.de>
8476 PR tree-optimization/83176
8477 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8480 2018-01-24 Richard Biener <rguenther@suse.de>
8482 PR tree-optimization/82819
8483 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8484 code generating pluses that are no-ops in the target precision.
8486 2018-01-24 Richard Biener <rguenther@suse.de>
8489 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8491 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8493 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8494 to merge probabilities.
8495 * predict.c (probably_never_executed): Also mark as cold functions
8496 with global 0 profile and guessed local profile.
8497 * profile-count.c (profile_probability::combine_with_count): New
8499 * profile-count.h (profile_probability::operator*,
8500 profile_probability::operator*=, profile_probability::operator/,
8501 profile_probability::operator/=): Reduce precision to adjusted
8502 and set value to guessed on contradictory divisions.
8503 (profile_probability::combine_with_freq): Remove.
8504 (profile_probability::combine_wiht_count): Declare.
8505 (profile_count::force_nonzero):: Set to adjusted.
8506 (profile_count::probability_in):: Set quality to adjusted.
8507 * tree-ssa-tail-merge.c (replace_block_by): Use
8510 2018-01-23 Andrew Waterman <andrew@sifive.com>
8511 Jim Wilson <jimw@sifive.com>
8513 * config/riscv/riscv.c (riscv_stack_boundary): New.
8514 (riscv_option_override): Set riscv_stack_boundary. Handle
8515 riscv_preferred_stack_boundary_arg.
8516 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8517 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8518 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8519 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8520 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8521 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8523 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8526 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8527 of struct ix86_frame.
8528 (ix86_expand_epilogue): Likewise. Add a local variable for
8529 the reg_save_offset field in struct ix86_frame.
8531 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8533 PR tree-optimization/82604
8534 * tree-loop-distribution.c (enum partition_kind): New enum item
8535 PKIND_PARTIAL_MEMSET.
8536 (partition_builtin_p): Support above new enum item.
8537 (generate_code_for_partition): Ditto.
8538 (compute_access_range): Differentiate cases that equality can be
8539 proven at all loops, the innermost loops or no loops.
8540 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8541 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8542 (finalize_partitions, distribute_loop): Don't fuse partition of
8543 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8544 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8547 2018-01-23 Martin Liska <mliska@suse.cz>
8549 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8550 order to ignore the predictor.
8551 (PRED_POLYMORPHIC_CALL): Likewise.
8552 (PRED_RECURSIVE_CALL): Likewise.
8554 2018-01-23 Martin Liska <mliska@suse.cz>
8556 * tree-profile.c (tree_profiling): Print function header to
8557 aware reader which function we are working on.
8558 * value-prof.c (gimple_find_values_to_profile): Do not print
8559 not interesting value histograms.
8561 2018-01-23 Martin Liska <mliska@suse.cz>
8563 * profile-count.h (enum profile_quality): Add
8564 profile_uninitialized as the first value. Do not number values
8565 as they are zero based.
8566 (profile_count::verify): Update sanity check.
8567 (profile_probability::verify): Likewise.
8569 2018-01-23 Nathan Sidwell <nathan@acm.org>
8571 * doc/invoke.texi (ffor-scope): Deprecate.
8573 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8575 PR tree-optimization/83510
8576 * domwalk.c (set_all_edges_as_executable): New function.
8577 (dom_walker::dom_walker): Convert bool param
8578 "skip_unreachable_blocks" to enum reachability. Move setup of
8579 edge flags to set_all_edges_as_executable and only do it when
8580 reachability is REACHABLE_BLOCKS.
8581 * domwalk.h (enum dom_walker::reachability): New enum.
8582 (dom_walker::dom_walker): Convert bool param
8583 "skip_unreachable_blocks" to enum reachability.
8584 (set_all_edges_as_executable): New decl.
8585 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8586 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8588 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8589 but converting true to REACHABLE_BLOCKS.
8590 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8592 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8593 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8594 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8596 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8597 if check_all_array_refs will be called.
8599 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8601 * tree.c (selftest::test_location_wrappers): Add more test
8604 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8606 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8607 (selftest::test_bit_in_range): Likewise.
8609 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8612 * doc/sourcebuild.texi (vect_float): Say that the selector
8613 only describes the situation when -funsafe-math-optimizations is on.
8614 (vect_float_strict): Document.
8616 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8618 PR tree-optimization/83965
8619 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8620 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8621 instead of checking only for a reduction.
8622 (vect_recog_widen_sum_pattern): Likewise.
8624 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8626 * predict.c (probably_never_executed): Only use precise profile info.
8627 (compute_function_frequency): Skip after inlining hack since we now
8628 have quality checking.
8630 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8632 * profile-count.h (profile_probability::very_unlikely,
8633 profile_probability::unlikely, profile_probability::even): Set
8634 precision to guessed.
8636 2018-01-23 Richard Biener <rguenther@suse.de>
8638 PR tree-optimization/83963
8639 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8640 Properly terminate dominator walk when crossing the exit edge not
8641 when visiting its source block.
8643 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8646 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8647 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8649 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8651 PR tree-optimization/83957
8652 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8653 semicolon after for body surrounded by braces.
8655 PR tree-optimization/83081
8656 * profile-count.h (profile_probability::split): New method.
8657 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8658 Use profile_probability::split.
8659 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8660 when splitting a single conditional jump into 2.
8662 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8664 PR tree-optimization/69452
8665 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8668 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8670 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8671 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8672 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8674 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8676 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8677 * config/rl78/rl78.md (movdi): New define_expand.
8678 * config/rl78/rl78.c (rl78_split_movdi): New function.
8680 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8683 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8685 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8686 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8687 128-bit to produce an UNSPEC move to get the double word with the
8688 signbit and then a shift directly to do signbit.
8689 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8690 implementation with a new version that just does either a direct
8691 move or a regular move. Move memory interface to separate insns.
8692 Move insns so they are next to the expander.
8693 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8694 with signbit move. Split big and little endian case.
8695 (signbit<mode>2_dm_mem_le): Likewise.
8696 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8697 (signbit<mode>2_dm2): Likewise.
8699 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8701 * config/rl78/rl78.md (anddi3): New define_expand.
8703 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8705 * config/rl78/rl78.md (umindi3): New define_expand.
8707 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8709 * config/rl78/rl78.md (smindi3): New define_expand.
8711 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8713 * config/rl78/rl78.md (smaxdi3): New define_expand.
8715 2018-01-22 Carl Love <cel@us.ibm.com>
8717 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8718 LVX_V1TI): Add macro expansion.
8719 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8720 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8721 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8722 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8723 Change check to determine if the instruction is a byte reversing
8724 entry. Fix typo in comment.
8725 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8726 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8727 Add def_builtin calls for new builtins.
8728 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8729 Add define_insn expansion.
8731 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8733 * config/rl78/rl78.md (umaxdi3): New define_expand.
8735 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8737 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8738 for non-QImode registers.
8740 2018-01-22 Richard Biener <rguenther@suse.de>
8742 PR tree-optimization/83963
8743 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8744 including the loop exit block.
8745 (scop_detection::merge_sese): Likewise.
8746 (scop_detection::add_scop): Do it here instead.
8748 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8750 * doc/sourcebuild.texi (arm_softfloat): Document.
8752 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8755 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8756 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8757 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8759 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8760 David Edelsohn <dje.gcc@gmail.com>
8763 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8764 Change "crset eq" to "crset 2".
8765 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8766 (*call_indirect_aix<mode>_nospec): Likewise.
8767 (*call_value_indirect_aix<mode>_nospec): Likewise.
8768 (*call_indirect_elfv2<mode>_nospec): Likewise.
8769 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8770 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8771 change assembly output from . to $.
8772 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8773 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8774 (*tablejump<mode>_internal1_nospec): Likewise.
8776 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8779 * config/sh/sh_optimize_sett_clrt.cc:
8780 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8782 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8784 PR tree-optimization/83940
8785 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8786 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8787 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8788 use it to pass back the definition type.
8789 (vect_check_store_rhs): Likewise rhs_dt_out.
8790 (vect_build_gather_load_calls): Add a mask_dt argument and use
8791 it instead of a call to vect_is_simple_use.
8792 (vectorizable_store): Update calls to vect_check_load_store_mask
8793 and vect_check_store_rhs. Use the dt returned by the latter instead
8794 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8795 instead of calls to vect_is_simple_use. Pass the scalar rather
8796 than the vector operand to vect_is_simple_use when handling
8797 second and subsequent copies of an rhs value.
8798 (vectorizable_load): Update calls to vect_check_load_store_mask
8799 and vect_build_gather_load_calls. Use the cached mask_dt and
8800 gs_info.offset_dt instead of calls to vect_is_simple_use.
8802 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8805 * tree-emutls.c: Include gimplify.h.
8806 (lower_emutls_2): New function.
8807 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8808 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8809 it before further processing.
8812 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8813 UINTVAL (trueop1) instead of INTVAL (op1).
8815 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8819 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8820 INCOMING_FRAME_SP_OFFSET if not defined.
8821 (scan_trace): Add ENTRY argument. If true and
8822 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8823 emit a note to adjust the CFA offset.
8824 (create_cfi_notes): Adjust scan_trace callers.
8825 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8826 INCOMING_FRAME_SP_OFFSET in the CIE.
8827 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8828 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8830 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8831 * doc/tm.texi: Regenerated.
8833 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8835 PR rtl-optimization/83147
8836 * lra-constraints.c (remove_inheritance_pseudos): Use
8837 lra_substitute_pseudo_within_insn.
8839 2018-01-19 Tom de Vries <tom@codesourcery.com>
8840 Cesar Philippidis <cesar@codesourcery.com>
8843 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8845 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8848 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8849 spaces for function labels.
8851 2018-01-19 Martin Liska <mliska@suse.cz>
8853 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8854 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8855 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8856 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8857 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8858 (PRED_CONST_RETURN): Change from 69 to 65.
8859 (PRED_NULL_RETURN): Change from 91 to 71.
8860 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8861 (PRED_LOOP_GUARD): Change from 66 to 73.
8863 2018-01-19 Martin Liska <mliska@suse.cz>
8865 * predict.c (predict_insn_def): Add new assert.
8866 (struct branch_predictor): Change type to signed integer.
8867 (test_prediction_value_range): Amend test to cover
8869 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8870 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8871 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8872 (PRED_LOOP_IV_COMPARE): Likewise.
8873 * predict.h (PROB_UNINITIALIZED): Define new constant.
8875 2018-01-19 Martin Liska <mliska@suse.cz>
8877 * predict.c (dump_prediction): Add new format for
8878 analyze_brprob.py script which is enabled with -details
8880 * profile-count.h (precise_p): New function.
8882 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8884 PR tree-optimization/83922
8885 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8886 there are no statements that need masking.
8887 (vect_active_double_reduction_p): New function.
8888 (vect_analyze_loop_operations): Use it when handling phis that
8889 are not in the loop header.
8891 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8893 PR tree-optimization/83914
8894 * tree-vect-loop.c (vectorizable_induction): Don't convert
8895 init_expr or apply the peeling adjustment for inductions
8896 that are nested within the vectorized loop.
8898 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8900 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8903 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8907 * function.h (gimplify_parameters): Add gimple_seq * argument.
8908 * function.c: Include gimple.h and options.h.
8909 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8910 for the added local temporaries if needed.
8911 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8912 if there are any parameter cleanups, wrap whole body into a
8913 try/finally with the cleanups.
8915 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8918 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8919 Use GET_MODE_CLASS for scalar floating point.
8921 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8925 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8926 Fix call of call_cgraph_insertion_hooks.
8928 2018-01-18 Martin Sebor <msebor@redhat.com>
8930 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8932 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8935 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8938 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8941 * common.opt: (-ffile-prefix-map): New option.
8942 * opts.c (common_handle_option): Defer it.
8943 * opts-global.c (handle_common_deferred_options): Handle it.
8944 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8945 * file-prefix-map.h: New file.
8946 (remap_debug_filename, add_debug_prefix_map): ...here.
8947 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8948 * final.c (debug_prefix_map, add_debug_prefix_map
8949 remap_debug_filename): Move to...
8950 * file-prefix-map.c: New file.
8951 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8952 generalize, get rid of alloca(), use strrchr() instead of strchr().
8953 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8954 Implement in terms of add_prefix_map().
8955 (remap_macro_filename, remap_debug_filename): Implement in term of
8957 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8958 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8959 * dbxout.c: Include file-prefix-map.h.
8960 * varasm.c: Likewise.
8961 * vmsdbgout.c: Likewise.
8962 * xcoffout.c: Likewise.
8963 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8964 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8965 * doc/invoke.texi (-ffile-prefix-map): Document.
8966 (-fdebug-prefix-map): Update description.
8968 2018-01-18 Martin Liska <mliska@suse.cz>
8970 * config/i386/i386.c (indirect_thunk_name): Document that also
8972 (output_indirect_thunk): Document why both instructions
8973 (pause and lfence) are generated.
8975 2018-01-18 Richard Biener <rguenther@suse.de>
8977 PR tree-optimization/83887
8978 * graphite-scop-detection.c
8979 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8980 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8981 (scop_detection::merge_sese): Re-implement with a flood-fill
8982 algorithm that properly finds a SESE region if it exists.
8984 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8987 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8988 pointer_diff optimizations use view_convert instead of convert.
8990 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8992 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8993 Generate different code for -mno-speculate-indirect-jumps.
8994 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8995 (*call_indirect_aix<mode>): Disable for
8996 -mno-speculate-indirect-jumps.
8997 (*call_indirect_aix<mode>_nospec): New define_insn.
8998 (*call_value_indirect_aix<mode>): Disable for
8999 -mno-speculate-indirect-jumps.
9000 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9001 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9002 -mno-speculate-indirect-jumps.
9003 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9005 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9007 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9008 long double type, set the flags for noting the default long double
9009 type, even if we don't pass or return a long double type.
9011 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9014 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9017 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9019 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9020 support for merge[hl].
9021 (fold_mergehl_helper): New helper function.
9022 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9023 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9024 (altivec_vmrglw_direct): Add xxmrglw insn.
9026 2018-01-17 Andrew Waterman <andrew@sifive.com>
9028 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9029 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9031 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9034 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9035 call the lto_location_cache before reading the
9036 DECL_SOURCE_LOCATION of the types.
9038 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9039 Richard Sandiford <richard.sandiford@linaro.org>
9041 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9042 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9043 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9044 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9045 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9047 * config/aarch64/constraints.md (aarch64_movti_operand):
9049 * config/aarch64/predicates.md (Uti): Add new constraint.
9051 2018-01-17 Carl Love <cel@us.ibm.com>
9053 * config/rs6000/vsx.md (define_expand xl_len_r,
9054 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9055 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9057 (define_expand, define_insn): Move the shift left from the
9058 define_insn to the define_expand for lxvl and stxvl instructions.
9059 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9060 and XL_LEN_R definitions to PURE.
9062 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9064 * config/i386/i386.c (indirect_thunk_name): Declare regno
9065 as unsigned int. Compare regno with INVALID_REGNUM.
9066 (output_indirect_thunk): Ditto.
9067 (output_indirect_thunk_function): Ditto.
9068 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9069 in the call to output_indirect_thunk_function.
9071 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9074 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9075 rather than the size of inner_type to determine the stack slot size
9076 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9078 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9081 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9084 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9086 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9087 endian Linux systems to optionally enable multilibs for selecting
9088 the long double type if the user configured an explicit type.
9089 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9090 have no long double multilibs if not defined.
9091 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9092 warn if the user used -mabi={ieee,ibm}longdouble and we built
9093 multilibs for long double.
9094 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9095 appropriate multilib option.
9096 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9098 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9099 for building long double multilibs.
9100 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9102 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9104 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9107 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9109 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9112 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9115 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9118 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9120 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9121 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9123 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9125 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9126 different rtl trees depending on TARGET_64BIT.
9127 (rs6000_gen_lvx): Likewise.
9129 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9131 * config/visium/visium.md (nop): Tweak comment.
9132 (hazard_nop): Likewise.
9134 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9136 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9137 -mspeculate-indirect-jumps.
9138 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9139 for -mno-speculate-indirect-jumps.
9140 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9141 (*call_value_indirect_elfv2<mode>): Disable for
9142 -mno-speculate-indirect-jumps.
9143 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9144 (indirect_jump): Emit different RTL for
9145 -mno-speculate-indirect-jumps.
9146 (*indirect_jump<mode>): Disable for
9147 -mno-speculate-indirect-jumps.
9148 (*indirect_jump<mode>_nospec): New define_insn.
9149 (tablejump): Emit different RTL for
9150 -mno-speculate-indirect-jumps.
9151 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9152 (tablejumpsi_nospec): New define_expand.
9153 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9154 (tablejumpdi_nospec): New define_expand.
9155 (*tablejump<mode>_internal1): Disable for
9156 -mno-speculate-indirect-jumps.
9157 (*tablejump<mode>_internal1_nospec): New define_insn.
9158 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9161 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9163 * caller-save.c (insert_save): Drop unnecessary parameter. All
9166 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9167 Richard Biener <rguenth@suse.de>
9170 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9171 return early, inline manually is_gimple_sizepos. Make sure if we
9172 call gimplify_expr we don't end up with a gimple constant.
9173 * tree.c (variably_modified_type_p): Don't return true for
9174 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9175 * gimplify.h (is_gimple_sizepos): Remove.
9177 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9179 PR tree-optimization/83857
9180 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9181 vectorizable_live_operation for pure SLP statements.
9182 (vectorizable_live_operation): Handle PHIs.
9184 2018-01-16 Richard Biener <rguenther@suse.de>
9186 PR tree-optimization/83867
9187 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9188 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9190 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9193 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9194 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9195 If off is not INTEGER_CST, issue a may not be aligned warning
9196 rather than isn't aligned. Use isn%'t rather than isn't.
9197 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9199 <case MULT_EXPR>: Improve the case when bottom and one of the
9200 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9201 operand, in that case check if the other operand is multiple of
9202 bottom divided by the INTEGER_CST operand.
9204 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9207 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9208 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9209 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9210 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9211 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9212 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9213 (pa_function_arg_size): New function.
9215 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9217 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9218 in a separate statement.
9220 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9222 PR tree-optimization/83847
9223 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9224 group gathers and scatters.
9226 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9228 PR rtl-optimization/86620
9229 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9231 PR rtl-optimization/83213
9232 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9233 to last if both are JUMP_INSNs.
9235 PR tree-optimization/83843
9236 * gimple-ssa-store-merging.c
9237 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9238 store_immediate_info for bswap/nop orig_stores.
9240 2018-01-15 Andrew Waterman <andrew@sifive.com>
9242 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9244 <UDIV>: Increase cost if !TARGET_DIV.
9246 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9248 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9249 (define_attr "cr_logical_3op"): New.
9250 (cceq_ior_compare): Adjust.
9251 (cceq_ior_compare_complement): Adjust.
9252 (*cceq_rev_compare): Adjust.
9253 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9254 (is_cracked_insn): Adjust.
9255 (insn_must_be_first_in_group): Adjust.
9256 * config/rs6000/40x.md: Adjust.
9257 * config/rs6000/440.md: Adjust.
9258 * config/rs6000/476.md: Adjust.
9259 * config/rs6000/601.md: Adjust.
9260 * config/rs6000/603.md: Adjust.
9261 * config/rs6000/6xx.md: Adjust.
9262 * config/rs6000/7450.md: Adjust.
9263 * config/rs6000/7xx.md: Adjust.
9264 * config/rs6000/8540.md: Adjust.
9265 * config/rs6000/cell.md: Adjust.
9266 * config/rs6000/e300c2c3.md: Adjust.
9267 * config/rs6000/e500mc.md: Adjust.
9268 * config/rs6000/e500mc64.md: Adjust.
9269 * config/rs6000/e5500.md: Adjust.
9270 * config/rs6000/e6500.md: Adjust.
9271 * config/rs6000/mpc.md: Adjust.
9272 * config/rs6000/power4.md: Adjust.
9273 * config/rs6000/power5.md: Adjust.
9274 * config/rs6000/power6.md: Adjust.
9275 * config/rs6000/power7.md: Adjust.
9276 * config/rs6000/power8.md: Adjust.
9277 * config/rs6000/power9.md: Adjust.
9278 * config/rs6000/rs64.md: Adjust.
9279 * config/rs6000/titan.md: Adjust.
9281 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9283 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9284 ix86_indirect_branch_register logic.
9286 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9288 * config/i386/constraints.md (Bs): Update
9289 ix86_indirect_branch_register check. Don't check
9290 ix86_indirect_branch_register with GOT_memory_operand.
9292 * config/i386/predicates.md (GOT_memory_operand): Don't check
9293 ix86_indirect_branch_register here.
9294 (GOT32_symbol_operand): Likewise.
9296 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9298 * config/i386/predicates.md (constant_call_address_operand):
9299 Rewrite ix86_indirect_branch_register logic.
9300 (sibcall_insn_operand): Likewise.
9302 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9304 * config/i386/constraints.md (Bs): Replace
9305 ix86_indirect_branch_thunk_register with
9306 ix86_indirect_branch_register.
9308 * config/i386/i386.md (indirect_jump): Likewise.
9309 (tablejump): Likewise.
9310 (*sibcall_memory): Likewise.
9311 (*sibcall_value_memory): Likewise.
9312 Peepholes of indirect call and jump via memory: Likewise.
9313 * config/i386/i386.opt: Likewise.
9314 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9315 (GOT_memory_operand): Likewise.
9316 (call_insn_operand): Likewise.
9317 (sibcall_insn_operand): Likewise.
9318 (GOT32_symbol_operand): Likewise.
9320 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9323 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9324 type rather than type addr's type points to.
9325 (expand_omp_atomic_mutex): Likewise.
9326 (expand_omp_atomic): Likewise.
9328 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9331 * config/i386/i386.c (output_indirect_thunk_function): Use
9332 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9333 for __x86_return_thunk.
9335 2018-01-15 Richard Biener <rguenther@suse.de>
9338 * expmed.c (extract_bit_field_1): Fix typo.
9340 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9343 * config/arm/iterators.md (VF): New mode iterator.
9344 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9345 Remove integer-related logic from pattern.
9346 (neon_vabd<mode>_3): Likewise.
9348 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9351 * common.opt (fstrict-overflow): No longer an alias.
9352 (fwrapv-pointer): New option.
9353 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9354 also for pointer types based on flag_wrapv_pointer.
9355 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9356 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9357 opts->x_flag_wrapv got set.
9358 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9359 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9360 POINTER_TYPE_OVERFLOW_UNDEFINED.
9361 * match.pd: Likewise in address comparison pattern.
9362 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9364 2018-01-15 Richard Biener <rguenther@suse.de>
9367 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9368 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9369 Reset type names to their identifier if their TYPE_DECL doesn't
9370 have linkage (and thus is used for ODR and devirt).
9371 (save_debug_info_for_decl): Remove.
9372 (save_debug_info_for_type): Likewise.
9373 (add_tree_to_fld_list): Adjust.
9374 * tree-pretty-print.c (dump_generic_node): Make dumping of
9375 type names more robust.
9377 2018-01-15 Richard Biener <rguenther@suse.de>
9379 * BASE-VER: Bump to 8.0.1.
9381 2018-01-14 Martin Sebor <msebor@redhat.com>
9384 * builtins.c (check_access): Avoid warning when the no-warning bit
9387 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9389 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9390 * ira-color (allocno_hard_regs_compare): Likewise.
9392 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9395 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9396 Use .pushsection/.popsection.
9398 2018-01-14 Martin Sebor <msebor@redhat.com>
9401 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9403 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9405 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9406 entry from extra_headers.
9407 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9408 extra_headers, make the list bitwise identical to the i?86-*-* one.
9410 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9412 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9413 -mcmodel=large with -mindirect-branch=thunk,
9414 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9415 -mfunction-return=thunk-extern.
9416 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9417 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9418 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9420 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9422 * config/i386/i386.c (print_reg): Print the name of the full
9423 integer register without '%'.
9424 (ix86_print_operand): Handle 'V'.
9425 * doc/extend.texi: Document 'V' modifier.
9427 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9429 * config/i386/constraints.md (Bs): Disallow memory operand for
9430 -mindirect-branch-register.
9432 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9433 (GOT_memory_operand): Likewise.
9434 (call_insn_operand): Likewise.
9435 (sibcall_insn_operand): Likewise.
9436 (GOT32_symbol_operand): Likewise.
9437 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9438 for -mindirect-branch-register.
9439 (tablejump): Likewise.
9440 (*sibcall_memory): Likewise.
9441 (*sibcall_value_memory): Likewise.
9442 Disallow peepholes of indirect call and jump via memory for
9443 -mindirect-branch-register.
9444 (*call_pop): Replace m with Bw.
9445 (*call_value_pop): Likewise.
9446 (*sibcall_pop_memory): Replace m with Bs.
9447 * config/i386/i386.opt (mindirect-branch-register): New option.
9448 * doc/invoke.texi: Document -mindirect-branch-register option.
9450 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9452 * config/i386/i386-protos.h (ix86_output_function_return): New.
9453 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9454 set function_return_type.
9455 (indirect_thunk_name): Add ret_p to indicate thunk for function
9457 (output_indirect_thunk_function): Pass false to
9458 indirect_thunk_name.
9459 (ix86_output_indirect_branch_via_reg): Likewise.
9460 (ix86_output_indirect_branch_via_push): Likewise.
9461 (output_indirect_thunk_function): Create alias for function
9462 return thunk if regno < 0.
9463 (ix86_output_function_return): New function.
9464 (ix86_handle_fndecl_attribute): Handle function_return.
9465 (ix86_attribute_table): Add function_return.
9466 * config/i386/i386.h (machine_function): Add
9467 function_return_type.
9468 * config/i386/i386.md (simple_return_internal): Use
9469 ix86_output_function_return.
9470 (simple_return_internal_long): Likewise.
9471 * config/i386/i386.opt (mfunction-return=): New option.
9472 (indirect_branch): Mention -mfunction-return=.
9473 * doc/extend.texi: Document function_return function attribute.
9474 * doc/invoke.texi: Document -mfunction-return= option.
9476 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9478 * config/i386/i386-opts.h (indirect_branch): New.
9479 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9480 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9481 with local indirect jump when converting indirect call and jump.
9482 (ix86_set_indirect_branch_type): New.
9483 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9484 (indirectlabelno): New.
9485 (indirect_thunk_needed): Likewise.
9486 (indirect_thunk_bnd_needed): Likewise.
9487 (indirect_thunks_used): Likewise.
9488 (indirect_thunks_bnd_used): Likewise.
9489 (INDIRECT_LABEL): Likewise.
9490 (indirect_thunk_name): Likewise.
9491 (output_indirect_thunk): Likewise.
9492 (output_indirect_thunk_function): Likewise.
9493 (ix86_output_indirect_branch_via_reg): Likewise.
9494 (ix86_output_indirect_branch_via_push): Likewise.
9495 (ix86_output_indirect_branch): Likewise.
9496 (ix86_output_indirect_jmp): Likewise.
9497 (ix86_code_end): Call output_indirect_thunk_function if needed.
9498 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9500 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9501 (ix86_attribute_table): Add indirect_branch.
9502 * config/i386/i386.h (machine_function): Add indirect_branch_type
9503 and has_local_indirect_jump.
9504 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9506 (tablejump): Likewise.
9507 (*indirect_jump): Use ix86_output_indirect_jmp.
9508 (*tablejump_1): Likewise.
9509 (simple_return_indirect_internal): Likewise.
9510 * config/i386/i386.opt (mindirect-branch=): New option.
9511 (indirect_branch): New.
9514 (thunk-inline): Likewise.
9515 (thunk-extern): Likewise.
9516 * doc/extend.texi: Document indirect_branch function attribute.
9517 * doc/invoke.texi: Document -mindirect-branch= option.
9519 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9522 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9524 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9526 * ipa-inline.c (want_inline_small_function_p): Return false if
9527 inlining has already failed with CIF_FINAL_ERROR.
9528 (update_caller_keys): Call want_inline_small_function_p before
9530 (update_callee_keys): Likewise.
9532 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9534 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9536 (rs6000_quadword_masked_address_p): Likewise.
9537 (quad_aligned_load_p): Likewise.
9538 (quad_aligned_store_p): Likewise.
9539 (const_load_sequence_p): Add comment to describe the outer-most loop.
9540 (mimic_memory_attributes_and_flags): New function.
9541 (rs6000_gen_stvx): Likewise.
9542 (replace_swapped_aligned_store): Likewise.
9543 (rs6000_gen_lvx): Likewise.
9544 (replace_swapped_aligned_load): Likewise.
9545 (replace_swapped_load_constant): Capitalize argument name in
9546 comment describing this function.
9547 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9548 and stores that access quad-word aligned addresses and replace
9549 with stvx or lvx instructions when appropriate.
9550 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9551 New function prototype.
9552 (rs6000_quadword_masked_address_p): Likewise.
9553 (rs6000_gen_lvx): Likewise.
9554 (rs6000_gen_stvx): Likewise.
9555 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9556 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9557 when memory address is aligned.
9558 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9559 this split to select lvx instruction when memory address is aligned.
9560 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9561 instruction when memory address is aligned.
9562 (*vsx_le_perm_load_v16qi): Likewise.
9563 (four unnamed splitters): Modify to select the stvx instruction
9564 when memory is aligned.
9566 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9568 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9569 which appears in the queue multiple times.
9571 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9572 Alan Hayward <alan.hayward@arm.com>
9573 David Sherwood <david.sherwood@arm.com>
9575 * tree-vectorizer.h (vec_lower_bound): New structure.
9576 (_loop_vec_info): Add check_nonzero and lower_bounds.
9577 (LOOP_VINFO_CHECK_NONZERO): New macro.
9578 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9579 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9580 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9581 fields. Make seg_len the distance travelled, not including the
9583 (dr_direction_indicator): Declare.
9584 (dr_zero_step_indicator): Likewise.
9585 (dr_known_forward_stride_p): Likewise.
9586 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9588 (runtime_alias_check_p): Allow runtime alias checks with
9590 (operator ==): Compare access_size and align.
9591 (prune_runtime_alias_test_list): Rework for new distinction between
9592 the access_size and seg_len.
9593 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9595 (get_segment_min_max): New function.
9596 (create_intersect_range_checks): Use it.
9597 (dr_step_indicator): New function.
9598 (dr_direction_indicator): Likewise.
9599 (dr_zero_step_indicator): Likewise.
9600 (dr_known_forward_stride_p): Likewise.
9601 * tree-loop-distribution.c (data_ref_segment_size): Return
9602 DR_STEP * (niters - 1).
9603 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9605 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9606 (vect_preserves_scalar_order_p): New function, split out from...
9607 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9608 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9609 (vect_vfa_access_size): New function.
9610 (vect_vfa_align): Likewise.
9611 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9612 (dump_lower_bound): New function.
9613 (vect_check_lower_bound): Likewise.
9614 (vect_small_gap_p): Likewise.
9615 (vectorizable_with_step_bound_p): Likewise.
9616 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9617 depencies if the vectorization factor is 1. Convert the checks
9618 for nonzero steps into checks on the bounds of DR_STEP. Try using
9619 a bunds check for variable steps if the minimum required step is
9620 relatively small. Update calls to the dr_with_seg_len
9621 constructor and to vect_compile_time_alias.
9622 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9624 (vect_loop_versioning): Call it.
9625 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9627 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9629 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9630 Alan Hayward <alan.hayward@arm.com>
9631 David Sherwood <david.sherwood@arm.com>
9633 * doc/sourcebuild.texi (vect_scatter_store): Document.
9634 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9636 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9638 * genopinit.c (main): Add supports_vec_scatter_store and
9639 supports_vec_scatter_store_cached to target_optabs.
9640 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9641 IFN_MASK_SCATTER_STORE.
9642 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9644 * internal-fn.h (internal_store_fn_p): Declare.
9645 (internal_fn_stored_value_index): Likewise.
9646 * internal-fn.c (scatter_store_direct): New macro.
9647 (expand_scatter_store_optab_fn): New function.
9648 (direct_scatter_store_optab_supported_p): New macro.
9649 (internal_store_fn_p): New function.
9650 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9651 IFN_MASK_SCATTER_STORE.
9652 (internal_fn_mask_index): Likewise.
9653 (internal_fn_stored_value_index): New function.
9654 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9656 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9657 * optabs-query.c (supports_vec_scatter_store_p): New function.
9658 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9659 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9660 true for scatter stores.
9661 (vect_gather_scatter_fn_p): Handle scatter stores too.
9662 (vect_check_gather_scatter): Consider using scatter stores if
9663 supports_vec_scatter_store_p.
9664 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9666 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9667 internal_fn_stored_value_index.
9668 (check_load_store_masking): Handle scatter stores too.
9669 (vect_get_store_rhs): Make public.
9670 (vectorizable_call): Use internal_store_fn_p.
9671 (vectorizable_store): Handle scatter store internal functions.
9672 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9673 when deciding whether the end of the group has been reached.
9674 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9675 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9676 (mask_scatter_store<mode>): New insns.
9678 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9679 Alan Hayward <alan.hayward@arm.com>
9680 David Sherwood <david.sherwood@arm.com>
9682 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9683 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9684 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9686 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9687 Use vect_truncate_gather_scatter_offset if we can't treat the
9688 operation as a normal gather load or scatter store.
9689 (get_group_load_store_type): Take the gather_scatter_info
9690 as argument. Try using a gather load or scatter store for
9691 single-element groups.
9692 (get_load_store_type): Update calls to get_group_load_store_type
9693 and vect_use_strided_gather_scatters_p.
9695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9696 Alan Hayward <alan.hayward@arm.com>
9697 David Sherwood <david.sherwood@arm.com>
9699 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9700 optional tree argument.
9701 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9703 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9704 but continue to use the current value as a fallback.
9705 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9706 to compare the updates.
9707 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9708 (get_load_store_type): Use it when handling a strided access.
9709 (vect_get_strided_load_store_ops): New function.
9710 (vect_get_data_ptr_increment): Likewise.
9711 (vectorizable_load): Handle strided gather loads. Always pass
9712 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9714 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9715 Alan Hayward <alan.hayward@arm.com>
9716 David Sherwood <david.sherwood@arm.com>
9718 * doc/md.texi (gather_load@var{m}): Document.
9719 (mask_gather_load@var{m}): Likewise.
9720 * genopinit.c (main): Add supports_vec_gather_load and
9721 supports_vec_gather_load_cached to target_optabs.
9722 * optabs-tree.c (init_tree_optimization_optabs): Use
9723 ggc_cleared_alloc to allocate target_optabs.
9724 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9725 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9727 * internal-fn.h (internal_load_fn_p): Declare.
9728 (internal_gather_scatter_fn_p): Likewise.
9729 (internal_fn_mask_index): Likewise.
9730 (internal_gather_scatter_fn_supported_p): Likewise.
9731 * internal-fn.c (gather_load_direct): New macro.
9732 (expand_gather_load_optab_fn): New function.
9733 (direct_gather_load_optab_supported_p): New macro.
9734 (direct_internal_fn_optab): New function.
9735 (internal_load_fn_p): Likewise.
9736 (internal_gather_scatter_fn_p): Likewise.
9737 (internal_fn_mask_index): Likewise.
9738 (internal_gather_scatter_fn_supported_p): Likewise.
9739 * optabs-query.c (supports_at_least_one_mode_p): New function.
9740 (supports_vec_gather_load_p): Likewise.
9741 * optabs-query.h (supports_vec_gather_load_p): Declare.
9742 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9743 and memory_type field.
9744 (NUM_PATTERNS): Bump to 15.
9745 * tree-vect-data-refs.c: Include internal-fn.h.
9746 (vect_gather_scatter_fn_p): New function.
9747 (vect_describe_gather_scatter_call): Likewise.
9748 (vect_check_gather_scatter): Try using internal functions for
9749 gather loads. Recognize existing calls to a gather load function.
9750 (vect_analyze_data_refs): Consider using gather loads if
9751 supports_vec_gather_load_p.
9752 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9753 (vect_get_gather_scatter_offset_type): Likewise.
9754 (vect_convert_mask_for_vectype): Likewise.
9755 (vect_add_conversion_to_patterm): Likewise.
9756 (vect_try_gather_scatter_pattern): Likewise.
9757 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9758 (vect_vect_recog_func_ptrs): Add it.
9759 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9760 internal_fn_mask_index and internal_gather_scatter_fn_p.
9761 (check_load_store_masking): Take the gather_scatter_info as an
9762 argument and handle gather loads.
9763 (vect_get_gather_scatter_ops): New function.
9764 (vectorizable_call): Check internal_load_fn_p.
9765 (vectorizable_load): Likewise. Handle gather load internal
9767 (vectorizable_store): Update call to check_load_store_masking.
9768 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9769 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9770 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9771 (aarch64_gather_scale_operand_d): New predicates.
9772 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9773 (mask_gather_load<mode>): New insns.
9775 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9776 Alan Hayward <alan.hayward@arm.com>
9777 David Sherwood <david.sherwood@arm.com>
9779 * optabs.def (fold_left_plus_optab): New optab.
9780 * doc/md.texi (fold_left_plus_@var{m}): Document.
9781 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9782 * internal-fn.c (fold_left_direct): Define.
9783 (expand_fold_left_optab_fn): Likewise.
9784 (direct_fold_left_optab_supported_p): Likewise.
9785 * fold-const-call.c (fold_const_fold_left): New function.
9786 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9787 * tree-parloops.c (valid_reduction_p): New function.
9788 (gather_scalar_reductions): Use it.
9789 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9790 (vect_finish_replace_stmt): Declare.
9791 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9792 (needs_fold_left_reduction_p): New function, split out from...
9793 (vect_is_simple_reduction): ...here. Accept reductions that
9794 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9795 (vect_force_simple_reduction): Also store the reduction type in
9796 the assignment's STMT_VINFO_REDUC_TYPE.
9797 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9798 (merge_with_identity): New function.
9799 (vect_expand_fold_left): Likewise.
9800 (vectorize_fold_left_reduction): Likewise.
9801 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9802 scalar phi in place for it. Check for target support and reject
9803 cases that would reassociate the operation. Defer the transform
9804 phase to vectorize_fold_left_reduction.
9805 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9806 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9807 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9811 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9812 call to ifc_temp_var.
9814 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9815 Alan Hayward <alan.hayward@arm.com>
9816 David Sherwood <david.sherwood@arm.com>
9818 * target.def (legitimize_address_displacement): Take the original
9819 offset as a poly_int.
9820 * targhooks.h (default_legitimize_address_displacement): Update
9822 * targhooks.c (default_legitimize_address_displacement): Likewise.
9823 * doc/tm.texi: Regenerate.
9824 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9825 as an argument, moving assert of ad->disp == ad->disp_term to...
9826 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9827 Try calling targetm.legitimize_address_displacement before expanding
9828 the address rather than afterwards, and adjust for the new interface.
9829 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9830 Match the new hook interface. Handle SVE addresses.
9831 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9836 * Makefile.in (OBJS): Add early-remat.o.
9837 * target.def (select_early_remat_modes): New hook.
9838 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9839 * doc/tm.texi: Regenerate.
9840 * targhooks.h (default_select_early_remat_modes): Declare.
9841 * targhooks.c (default_select_early_remat_modes): New function.
9842 * timevar.def (TV_EARLY_REMAT): New timevar.
9843 * passes.def (pass_early_remat): New pass.
9844 * tree-pass.h (make_pass_early_remat): Declare.
9845 * early-remat.c: New file.
9846 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9848 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9851 Alan Hayward <alan.hayward@arm.com>
9852 David Sherwood <david.sherwood@arm.com>
9854 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9855 vfm1 with a bound_epilog parameter.
9856 (vect_do_peeling): Update calls accordingly, and move the prologue
9857 call earlier in the function. Treat the base bound_epilog as 0 for
9858 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9859 this base when peeling for gaps.
9860 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9861 with fully-masked loops.
9862 (vect_estimate_min_profitable_iters): Handle the single peeled
9863 iteration in that case.
9865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9866 Alan Hayward <alan.hayward@arm.com>
9867 David Sherwood <david.sherwood@arm.com>
9869 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9870 single-element interleaving even if the size is not a power of 2.
9871 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9872 accesses for single-element interleaving if the group size is
9875 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9876 Alan Hayward <alan.hayward@arm.com>
9877 David Sherwood <david.sherwood@arm.com>
9879 * doc/md.texi (fold_extract_last_@var{m}): Document.
9880 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9881 * optabs.def (fold_extract_last_optab): New optab.
9882 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9883 * internal-fn.c (fold_extract_direct): New macro.
9884 (expand_fold_extract_optab_fn): Likewise.
9885 (direct_fold_extract_optab_supported_p): Likewise.
9886 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9887 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9888 EXTRACT_LAST_REDUCTION.
9889 (get_initial_def_for_reduction): Do not create an initial vector
9890 for EXTRACT_LAST_REDUCTION reductions.
9891 (vectorizable_reduction): Leave the scalar phi in place for
9892 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9893 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9894 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9895 transform phase to vectorizable_condition.
9896 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9898 (vect_finish_stmt_generation): ...here.
9899 (vect_finish_replace_stmt): New function.
9900 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9901 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9903 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9906 Alan Hayward <alan.hayward@arm.com>
9907 David Sherwood <david.sherwood@arm.com>
9909 * doc/md.texi (extract_last_@var{m}): Document.
9910 * optabs.def (extract_last_optab): New optab.
9911 * internal-fn.def (EXTRACT_LAST): New internal function.
9912 * internal-fn.c (cond_unary_direct): New macro.
9913 (expand_cond_unary_optab_fn): Likewise.
9914 (direct_cond_unary_optab_supported_p): Likewise.
9915 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9916 loops using EXTRACT_LAST.
9917 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9918 (extract_last_<mode>): ...this optab.
9919 (vec_extract<mode><Vel>): Update accordingly.
9921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9922 Alan Hayward <alan.hayward@arm.com>
9923 David Sherwood <david.sherwood@arm.com>
9925 * target.def (empty_mask_is_expensive): New hook.
9926 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9927 * doc/tm.texi: Regenerate.
9928 * targhooks.h (default_empty_mask_is_expensive): Declare.
9929 * targhooks.c (default_empty_mask_is_expensive): New function.
9930 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9931 if the target says that empty masks are expensive.
9932 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9934 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9936 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9937 Alan Hayward <alan.hayward@arm.com>
9938 David Sherwood <david.sherwood@arm.com>
9940 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9941 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9942 (vect_use_loop_mask_for_alignment_p): New function.
9943 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9944 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9945 niters_skip argument. Make sure that the first niters_skip elements
9946 of the first iteration are inactive.
9947 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9948 Update call to vect_set_loop_masks_directly.
9949 (get_misalign_in_elems): New function, split out from...
9950 (vect_gen_prolog_loop_niters): ...here.
9951 (vect_update_init_of_dr): Take a code argument that specifies whether
9952 the adjustment should be added or subtracted.
9953 (vect_update_init_of_drs): Likewise.
9954 (vect_prepare_for_masked_peels): New function.
9955 (vect_do_peeling): Skip prologue peeling if we're using a mask
9956 instead. Update call to vect_update_inits_of_drs.
9957 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9959 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9960 alignment. Do not include the number of peeled iterations in
9961 the minimum threshold in that case.
9962 (vectorizable_induction): Adjust the start value down by
9963 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9964 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9965 Take the number of skipped iterations into account when calculating
9967 * tree-vect-stmts.c (vect_gen_while_not): New function.
9969 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9970 Alan Hayward <alan.hayward@arm.com>
9971 David Sherwood <david.sherwood@arm.com>
9973 * doc/sourcebuild.texi (vect_fully_masked): Document.
9974 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9976 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9978 (vect_analyze_loop_2): ...here. Don't check the vectorization
9979 factor against the number of loop iterations if the loop is
9982 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9983 Alan Hayward <alan.hayward@arm.com>
9984 David Sherwood <david.sherwood@arm.com>
9986 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9987 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9988 (dump_groups): Update accordingly.
9989 (iv_use::mem_type): New member variable.
9990 (address_p): New function.
9991 (record_use): Add a mem_type argument and initialize the new
9993 (record_group_use): Add a mem_type argument. Use address_p.
9994 Remove obsolete null checks of base_object. Update call to record_use.
9995 (find_interesting_uses_op): Update call to record_group_use.
9996 (find_interesting_uses_cond): Likewise.
9997 (find_interesting_uses_address): Likewise.
9998 (get_mem_type_for_internal_fn): New function.
9999 (find_address_like_use): Likewise.
10000 (find_interesting_uses_stmt): Try find_address_like_use before
10001 calling find_interesting_uses_op.
10002 (addr_offset_valid_p): Use the iv mem_type field as the type
10003 of the addressed memory.
10004 (add_autoinc_candidates): Likewise.
10005 (get_address_cost): Likewise.
10006 (split_small_address_groups_p): Use address_p.
10007 (split_address_groups): Likewise.
10008 (add_iv_candidate_for_use): Likewise.
10009 (autoinc_possible_for_pair): Likewise.
10010 (rewrite_groups): Likewise.
10011 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10012 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10013 (get_alias_ptr_type_for_ptr_address): New function.
10014 (rewrite_use_address): Rewrite address uses in calls that were
10015 identified by find_address_like_use.
10017 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10018 Alan Hayward <alan.hayward@arm.com>
10019 David Sherwood <david.sherwood@arm.com>
10021 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10023 * gimple-expr.h (is_gimple_addressable: Likewise.
10024 * gimple-expr.c (is_gimple_address): Likewise.
10025 * internal-fn.c (expand_call_mem_ref): New function.
10026 (expand_mask_load_optab_fn): Use it.
10027 (expand_mask_store_optab_fn): Likewise.
10029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10030 Alan Hayward <alan.hayward@arm.com>
10031 David Sherwood <david.sherwood@arm.com>
10033 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10034 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10035 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10036 (cond_umax@var{mode}): Document.
10037 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10038 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10039 (cond_umin_optab, cond_umax_optab): New optabs.
10040 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10041 (COND_IOR, COND_XOR): New internal functions.
10042 * internal-fn.h (get_conditional_internal_fn): Declare.
10043 * internal-fn.c (cond_binary_direct): New macro.
10044 (expand_cond_binary_optab_fn): Likewise.
10045 (direct_cond_binary_optab_supported_p): Likewise.
10046 (get_conditional_internal_fn): New function.
10047 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10048 Cope with reduction statements that are vectorized as calls rather
10050 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10051 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10052 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10053 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10054 (UNSPEC_COND_EOR): New unspecs.
10055 (optab): Add mappings for them.
10056 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10057 (sve_int_op, sve_fp_op): New int attributes.
10059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10060 Alan Hayward <alan.hayward@arm.com>
10061 David Sherwood <david.sherwood@arm.com>
10063 * optabs.def (while_ult_optab): New optab.
10064 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10065 * internal-fn.def (WHILE_ULT): New internal function.
10066 * internal-fn.h (direct_internal_fn_supported_p): New override
10067 that takes two types as argument.
10068 * internal-fn.c (while_direct): New macro.
10069 (expand_while_optab_fn): New function.
10070 (convert_optab_supported_p): Likewise.
10071 (direct_while_optab_supported_p): New macro.
10072 * wide-int.h (wi::udiv_ceil): New function.
10073 * tree-vectorizer.h (rgroup_masks): New structure.
10074 (vec_loop_masks): New typedef.
10075 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10076 and fully_masked_p.
10077 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10078 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10079 (vect_max_vf): New function.
10080 (slpeel_make_loop_iterate_ntimes): Delete.
10081 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10082 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10083 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10084 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10085 internal-fn.h, stor-layout.h and optabs-query.h.
10086 (vect_set_loop_mask): New function.
10087 (add_preheader_seq): Likewise.
10088 (add_header_seq): Likewise.
10089 (interleave_supported_p): Likewise.
10090 (vect_maybe_permute_loop_masks): Likewise.
10091 (vect_set_loop_masks_directly): Likewise.
10092 (vect_set_loop_condition_masked): Likewise.
10093 (vect_set_loop_condition_unmasked): New function, split out from
10094 slpeel_make_loop_iterate_ntimes.
10095 (slpeel_make_loop_iterate_ntimes): Rename to..
10096 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10097 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10098 (vect_do_peeling): Update call accordingly.
10099 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10101 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10102 mask_compare_type, can_fully_mask_p and fully_masked_p.
10103 (release_vec_loop_masks): New function.
10104 (_loop_vec_info): Use it to free the loop masks.
10105 (can_produce_all_loop_masks_p): New function.
10106 (vect_get_max_nscalars_per_iter): Likewise.
10107 (vect_verify_full_masking): Likewise.
10108 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10109 retries, and free the mask rgroups before retrying. Check loop-wide
10110 reasons for disallowing fully-masked loops. Make the final decision
10111 about whether use a fully-masked loop or not.
10112 (vect_estimate_min_profitable_iters): Do not assume that peeling
10113 for the number of iterations will be needed for fully-masked loops.
10114 (vectorizable_reduction): Disable fully-masked loops.
10115 (vectorizable_live_operation): Likewise.
10116 (vect_halve_mask_nunits): New function.
10117 (vect_double_mask_nunits): Likewise.
10118 (vect_record_loop_mask): Likewise.
10119 (vect_get_loop_mask): Likewise.
10120 (vect_transform_loop): Handle the case in which the final loop
10121 iteration might handle a partial vector. Call vect_set_loop_condition
10122 instead of slpeel_make_loop_iterate_ntimes.
10123 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10124 (check_load_store_masking): New function.
10125 (prepare_load_store_mask): Likewise.
10126 (vectorizable_store): Handle fully-masked loops.
10127 (vectorizable_load): Likewise.
10128 (supportable_widening_operation): Use vect_halve_mask_nunits for
10130 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10131 (vect_gen_while): New function.
10132 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10133 (aarch64_uqdec<mode>): New insn.
10135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10136 Alan Hayward <alan.hayward@arm.com>
10137 David Sherwood <david.sherwood@arm.com>
10139 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10140 (reduc_xor_scal_optab): New optabs.
10141 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10142 (reduc_xor_scal_@var{m}): Document.
10143 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10144 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10145 internal functions.
10146 * fold-const-call.c (fold_const_call): Handle them.
10147 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10148 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10149 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10150 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10151 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10152 (UNSPEC_XORV): New unspecs.
10153 (optab): Add entries for them.
10154 (BITWISEV): New int iterator.
10155 (bit_reduc_op): New int attributes.
10157 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10158 Alan Hayward <alan.hayward@arm.com>
10159 David Sherwood <david.sherwood@arm.com>
10161 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10162 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10163 * optabs.def (vec_shl_insert_optab): New optab.
10164 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10165 (duplicate_and_interleave): Likewise.
10166 * tree-vect-loop.c: Include internal-fn.h.
10167 (neutral_op_for_slp_reduction): New function, split out from
10168 get_initial_defs_for_reduction.
10169 (get_initial_def_for_reduction): Handle option 2 for variable-length
10170 vectors by loading the neutral value into a vector and then shifting
10171 the initial value into element 0.
10172 (get_initial_defs_for_reduction): Replace the code argument with
10173 the neutral value calculated by neutral_op_for_slp_reduction.
10174 Use gimple_build_vector for constant-length vectors.
10175 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10176 but the first group_size elements have a neutral value.
10177 Use duplicate_and_interleave otherwise.
10178 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10179 Update call to get_initial_defs_for_reduction. Handle SLP
10180 reductions for variable-length vectors by creating one vector
10181 result for each scalar result, with the elements associated
10182 with other scalar results stubbed out with the neutral value.
10183 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10184 Require IFN_VEC_SHL_INSERT for double reductions on
10185 variable-length vectors, or SLP reductions that have
10186 a neutral value. Require can_duplicate_and_interleave_p
10187 support for variable-length unchained SLP reductions if there
10188 is no neutral value, such as for MIN/MAX reductions. Also require
10189 the number of vector elements to be a multiple of the number of
10190 SLP statements when doing variable-length unchained SLP reductions.
10191 Update call to vect_create_epilog_for_reduction.
10192 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10193 and remove initial values.
10194 (duplicate_and_interleave): Make public.
10195 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10196 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10199 Alan Hayward <alan.hayward@arm.com>
10200 David Sherwood <david.sherwood@arm.com>
10202 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10203 (can_duplicate_and_interleave_p): New function.
10204 (vect_get_and_check_slp_defs): Take the vector of statements
10205 rather than just the current one. Remove excess parentheses.
10206 Restriction rejectinon of vect_constant_def and vect_external_def
10207 for variable-length vectors to boolean types, or types for which
10208 can_duplicate_and_interleave_p is false.
10209 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10210 (duplicate_and_interleave): New function.
10211 (vect_get_constant_vectors): Use gimple_build_vector for
10212 constant-length vectors and suitable variable-length constant
10213 vectors. Use duplicate_and_interleave for other variable-length
10214 vectors. Don't defer the update when inserting new statements.
10216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10217 Alan Hayward <alan.hayward@arm.com>
10218 David Sherwood <david.sherwood@arm.com>
10220 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10221 min_profitable_iters doesn't go negative.
10223 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10224 Alan Hayward <alan.hayward@arm.com>
10225 David Sherwood <david.sherwood@arm.com>
10227 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10228 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10229 * optabs.def (vec_mask_load_lanes_optab): New optab.
10230 (vec_mask_store_lanes_optab): Likewise.
10231 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10232 (MASK_STORE_LANES): Likewise.
10233 * internal-fn.c (mask_load_lanes_direct): New macro.
10234 (mask_store_lanes_direct): Likewise.
10235 (expand_mask_load_optab_fn): Handle masked operations.
10236 (expand_mask_load_lanes_optab_fn): New macro.
10237 (expand_mask_store_optab_fn): Handle masked operations.
10238 (expand_mask_store_lanes_optab_fn): New macro.
10239 (direct_mask_load_lanes_optab_supported_p): Likewise.
10240 (direct_mask_store_lanes_optab_supported_p): Likewise.
10241 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10243 (vect_load_lanes_supported): Likewise.
10244 * tree-vect-data-refs.c (strip_conversion): New function.
10245 (can_group_stmts_p): Likewise.
10246 (vect_analyze_data_ref_accesses): Use it instead of checking
10247 for a pair of assignments.
10248 (vect_store_lanes_supported): Take a masked_p parameter.
10249 (vect_load_lanes_supported): Likewise.
10250 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10251 vect_store_lanes_supported and vect_load_lanes_supported.
10252 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10253 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10254 parameter. Don't allow gaps for masked accesses.
10255 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10256 and vect_load_lanes_supported.
10257 (get_load_store_type): Take a masked_p parameter and update
10258 call to get_group_load_store_type.
10259 (vectorizable_store): Update call to get_load_store_type.
10260 Handle IFN_MASK_STORE_LANES.
10261 (vectorizable_load): Update call to get_load_store_type.
10262 Handle IFN_MASK_LOAD_LANES.
10264 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10265 Alan Hayward <alan.hayward@arm.com>
10266 David Sherwood <david.sherwood@arm.com>
10268 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10270 * config/aarch64/aarch64-protos.h
10271 (aarch64_sve_struct_memory_operand_p): Declare.
10272 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10273 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10274 (VPRED, vpred): Handle SVE structure modes.
10275 * config/aarch64/constraints.md (Utx): New constraint.
10276 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10277 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10278 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10279 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10280 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10281 structure modes. Split into pieces after RA.
10282 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10283 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10285 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10286 SVE structure modes.
10287 (aarch64_classify_address): Likewise.
10288 (sizetochar): Move earlier in file.
10289 (aarch64_print_operand): Handle SVE register lists.
10290 (aarch64_array_mode): New function.
10291 (aarch64_sve_struct_memory_operand_p): Likewise.
10292 (TARGET_ARRAY_MODE): Redefine.
10294 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10295 Alan Hayward <alan.hayward@arm.com>
10296 David Sherwood <david.sherwood@arm.com>
10298 * target.def (array_mode): New target hook.
10299 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10300 * doc/tm.texi: Regenerate.
10301 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10302 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10303 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10304 targetm.array_mode.
10305 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10309 Alan Hayward <alan.hayward@arm.com>
10310 David Sherwood <david.sherwood@arm.com>
10312 * fold-const.c (fold_binary_loc): Check the argument types
10313 rather than the result type when testing for a vector operation.
10315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10317 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10318 * doc/tm.texi: Regenerate.
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 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10325 (sve): Document new AArch64 extension.
10326 * doc/md.texi (w): Extend the description of the AArch64
10327 constraint to include SVE vectors.
10328 (Upl, Upa): Document new AArch64 predicate constraints.
10329 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10331 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10332 (msve-vector-bits=): New option.
10333 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10334 SVE when these are disabled.
10335 (sve): New extension.
10336 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10337 modes. Adjust their number of units based on aarch64_sve_vg.
10338 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10339 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10340 aarch64_addr_query_type.
10341 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10342 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10343 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10344 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10345 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10346 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10347 (aarch64_simd_imm_zero_p): Delete.
10348 (aarch64_check_zero_based_sve_index_immediate): Declare.
10349 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10350 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10351 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10352 (aarch64_sve_float_mul_immediate_p): Likewise.
10353 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10354 rather than an rtx.
10355 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10356 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10357 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10358 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10359 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10360 (aarch64_regmode_natural_size): Likewise.
10361 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10362 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10364 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10365 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10366 for VG and the SVE predicate registers.
10367 (V_ALIASES): Add a "z"-prefixed alias.
10368 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10369 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10370 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10371 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10372 (REG_CLASS_NAMES): Add entries for them.
10373 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10374 and the predicate registers.
10375 (aarch64_sve_vg): Declare.
10376 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10377 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10378 (REGMODE_NATURAL_SIZE): Define.
10379 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10381 * config/aarch64/aarch64.c: Include cfgrtl.h.
10382 (simd_immediate_info): Add a constructor for series vectors,
10383 and an associated step field.
10384 (aarch64_sve_vg): New variable.
10385 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10386 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10387 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10388 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10389 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10390 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10391 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10392 (aarch64_get_mask_mode): New functions.
10393 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10394 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10395 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10396 predicate modes and predicate registers. Explicitly restrict
10397 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10398 to store a vector mode if it is recognized by
10399 aarch64_classify_vector_mode.
10400 (aarch64_regmode_natural_size): New function.
10401 (aarch64_hard_regno_caller_save_mode): Return the original mode
10403 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10404 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10405 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10406 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10408 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10409 does not overlap dest if the function is frame-related. Handle
10411 (aarch64_split_add_offset): New function.
10412 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10413 them aarch64_add_offset.
10414 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10415 and update call to aarch64_sub_sp.
10416 (aarch64_add_cfa_expression): New function.
10417 (aarch64_expand_prologue): Pass extra temporary registers to the
10418 functions above. Handle the case in which we need to emit new
10419 DW_CFA_expressions for registers that were originally saved
10420 relative to the stack pointer, but now have to be expressed
10421 relative to the frame pointer.
10422 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10424 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10425 IP0 and IP1 values for SVE frames.
10426 (aarch64_expand_vec_series): New function.
10427 (aarch64_expand_sve_widened_duplicate): Likewise.
10428 (aarch64_expand_sve_const_vector): Likewise.
10429 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10430 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10431 into the register, rather than emitting a SET directly.
10432 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10433 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10434 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10435 (offset_9bit_signed_scaled_p): New functions.
10436 (aarch64_replicate_bitmask_imm): New function.
10437 (aarch64_bitmask_imm): Use it.
10438 (aarch64_cannot_force_const_mem): Reject expressions involving
10439 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10440 (aarch64_classify_index): Handle SVE indices, by requiring
10441 a plain register index with a scale that matches the element size.
10442 (aarch64_classify_address): Handle SVE addresses. Assert that
10443 the mode of the address is VOIDmode or an integer mode.
10444 Update call to aarch64_classify_symbol.
10445 (aarch64_classify_symbolic_expression): Update call to
10446 aarch64_classify_symbol.
10447 (aarch64_const_vec_all_in_range_p): New function.
10448 (aarch64_print_vector_float_operand): Likewise.
10449 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10450 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10451 and the FP immediates 1.0 and 0.5.
10452 (aarch64_print_address_internal): Handle SVE addresses.
10453 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10454 (aarch64_regno_regclass): Handle predicate registers.
10455 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10457 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10458 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10459 (aarch64_convert_sve_vector_bits): New function.
10460 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10461 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10462 rather than an rtx.
10463 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10464 Handle SVE vector and predicate modes. Accept VL-based constants
10465 that need only one temporary register, and VL offsets that require
10466 no temporary registers.
10467 (aarch64_conditional_register_usage): Mark the predicate registers
10468 as fixed if SVE isn't available.
10469 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10470 Return true for SVE vector and predicate modes.
10471 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10472 rather than an unsigned int. Handle SVE modes.
10473 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10475 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10477 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10478 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10479 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10480 (aarch64_sve_float_mul_immediate_p): New functions.
10481 (aarch64_sve_valid_immediate): New function.
10482 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10483 Explicitly reject structure modes. Check for INDEX constants.
10484 Handle PTRUE and PFALSE constants.
10485 (aarch64_check_zero_based_sve_index_immediate): New function.
10486 (aarch64_simd_imm_zero_p): Delete.
10487 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10488 vector modes. Accept constants in the range of CNT[BHWD].
10489 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10490 ask for an Advanced SIMD mode.
10491 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10492 (aarch64_simd_vector_alignment): Handle SVE predicates.
10493 (aarch64_vectorize_preferred_vector_alignment): New function.
10494 (aarch64_simd_vector_alignment_reachable): Use it instead of
10496 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10497 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10499 (MAX_VECT_LEN): Delete.
10500 (expand_vec_perm_d): Add a vec_flags field.
10501 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10502 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10503 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10505 (aarch64_evpc_rev): Rename to...
10506 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10507 (aarch64_evpc_rev_global): New function.
10508 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10509 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10511 (aarch64_evpc_sve_tbl): New function.
10512 (aarch64_expand_vec_perm_const_1): Update after rename of
10513 aarch64_evpc_rev. Handle SVE permutes too, trying
10514 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10515 than aarch64_evpc_tbl.
10516 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10517 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10518 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10519 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10520 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10521 (aarch64_expand_sve_vcond): New functions.
10522 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10523 of aarch64_vector_mode_p.
10524 (aarch64_dwarf_poly_indeterminate_value): New function.
10525 (aarch64_compute_pressure_classes): Likewise.
10526 (aarch64_can_change_mode_class): Likewise.
10527 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10528 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10529 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10530 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10531 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10532 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10533 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10534 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10536 (Dn, Dl, Dr): Accept const as well as const_vector.
10537 (Dz): Likewise. Compare against CONST0_RTX.
10538 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10539 of "vector" where appropriate.
10540 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10541 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10542 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10543 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10544 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10545 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10546 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10547 (v_int_equiv): Extend to SVE modes.
10548 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10550 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10551 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10552 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10553 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10554 (SVE_COND_FP_CMP): New int iterators.
10555 (perm_hilo): Handle the new unpack unspecs.
10556 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10558 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10559 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10560 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10561 (aarch64_equality_operator, aarch64_constant_vector_operand)
10562 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10563 (aarch64_sve_nonimmediate_operand): Likewise.
10564 (aarch64_sve_general_operand): Likewise.
10565 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10566 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10567 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10568 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10569 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10570 (aarch64_sve_float_arith_immediate): Likewise.
10571 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10572 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10573 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10574 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10575 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10576 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10577 (aarch64_sve_float_arith_operand): Likewise.
10578 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10579 (aarch64_sve_float_mul_operand): Likewise.
10580 (aarch64_sve_vec_perm_operand): Likewise.
10581 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10582 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10583 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10584 as well as const_vector.
10585 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10586 in file. Use CONST0_RTX and CONSTM1_RTX.
10587 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10588 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10589 Use aarch64_simd_imm_zero.
10590 * config/aarch64/aarch64-sve.md: New file.
10591 * config/aarch64/aarch64.md: Include it.
10592 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10593 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10594 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10595 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10596 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10597 (sve): New attribute.
10598 (enabled): Disable instructions with the sve attribute unless
10600 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10601 aarch64_expand_mov_immediate.
10602 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10603 CNT[BHSD] immediates.
10604 (movti): Split CONST_POLY_INT moves into two halves.
10605 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10606 Split additions that need a temporary here if the destination
10607 is the stack pointer.
10608 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10609 (*add<mode>3_poly_1): New instruction.
10610 (set_clobber_cc): New expander.
10612 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10614 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10615 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10616 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10617 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10618 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10619 Change innermode from fixed_mode_size to machine_mode.
10620 (simplify_subreg): Update call accordingly. Handle a constant-sized
10621 subreg of a variable-length CONST_VECTOR.
10623 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10624 Alan Hayward <alan.hayward@arm.com>
10625 David Sherwood <david.sherwood@arm.com>
10627 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10628 (add_offset_to_base): New function, split out from...
10629 (create_mem_ref): ...here. When handling a scale other than 1,
10630 check first whether the address is valid without the offset.
10631 Add it into the base if so, leaving the index and scale as-is.
10633 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10636 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10637 fold_for_warn before checking if arg2 is INTEGER_CST.
10639 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10641 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10642 (store_multiple_operation): Delete.
10643 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10644 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10645 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10646 guarded by TARGET_STRING.
10647 (rs6000_output_load_multiple): Delete.
10648 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10649 OPTION_MASK_STRING / TARGET_STRING handling.
10650 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10651 (const rs6000_opt_masks) <"string">: Change mask to 0.
10652 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10653 (MASK_STRING): Delete.
10654 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10656 (load_multiple): Delete.
10663 (store_multiple): Delete.
10670 (movmemsi_8reg): Delete.
10671 (corresponding unnamed define_insn): Delete.
10672 (movmemsi_6reg): Delete.
10673 (corresponding unnamed define_insn): Delete.
10674 (movmemsi_4reg): Delete.
10675 (corresponding unnamed define_insn): Delete.
10676 (movmemsi_2reg): Delete.
10677 (corresponding unnamed define_insn): Delete.
10678 (movmemsi_1reg): Delete.
10679 (corresponding unnamed define_insn): Delete.
10680 * config/rs6000/rs6000.opt (mno-string): New.
10681 (mstring): Replace by deprecation warning stub.
10682 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10684 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10686 * regrename.c (regrename_do_replace): If replacing the same
10687 reg multiple times, try to reuse last created gen_raw_REG.
10690 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10691 main to workaround a bug in GDB.
10693 2018-01-12 Tom de Vries <tom@codesourcery.com>
10696 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10698 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10700 PR rtl-optimization/80481
10701 * ira-color.c (get_cap_member): New function.
10702 (allocnos_conflict_by_live_ranges_p): Use it.
10703 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10704 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10706 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10709 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10710 (*saddl_se_1): Ditto.
10711 (*ssubsi_1): Ditto.
10712 (*ssubl_se_1): Ditto.
10714 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10716 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10717 rather than wi::to_widest for DR_INITs.
10718 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10719 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10720 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10722 (vect_analyze_group_access_1): Note that here.
10724 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10726 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10727 polynomial type sizes.
10729 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10731 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10732 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10733 (gimple_add_tmp_var): Likewise.
10735 2018-01-12 Martin Liska <mliska@suse.cz>
10737 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10738 (gimple_alloc_sizes): Likewise.
10739 (dump_gimple_statistics): Use PRIu64 in printf format.
10740 * gimple.h: Change uint64_t to int.
10742 2018-01-12 Martin Liska <mliska@suse.cz>
10744 * tree-core.h: Use uint64_t instead of int.
10745 * tree.c (tree_node_counts): Likewise.
10746 (tree_node_sizes): Likewise.
10747 (dump_tree_statistics): Use PRIu64 in printf format.
10749 2018-01-12 Martin Liska <mliska@suse.cz>
10751 * Makefile.in: As qsort_chk is implemented in vec.c, add
10752 vec.o to linkage of gencfn-macros.
10753 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10754 passing the info to record_node_allocation_statistics.
10755 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10757 * ggc-common.c (struct ggc_usage): Add operator== and use
10758 it in operator< and compare function.
10759 * mem-stats.h (struct mem_usage): Likewise.
10760 * vec.c (struct vec_usage): Remove operator< and compare
10761 function. Can be simply inherited.
10763 2018-01-12 Martin Jambor <mjambor@suse.cz>
10766 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10767 * tree-ssa-math-opts.c: Include domwalk.h.
10768 (convert_mult_to_fma_1): New function.
10769 (fma_transformation_info): New type.
10770 (fma_deferring_state): Likewise.
10771 (cancel_fma_deferring): New function.
10772 (result_of_phi): Likewise.
10773 (last_fma_candidate_feeds_initial_phi): Likewise.
10774 (convert_mult_to_fma): Added deferring logic, split actual
10775 transformation to convert_mult_to_fma_1.
10776 (math_opts_dom_walker): New type.
10777 (math_opts_dom_walker::after_dom_children): New method, body moved
10778 here from pass_optimize_widening_mul::execute, added deferring logic
10780 (pass_optimize_widening_mul::execute): Moved most of code to
10781 math_opts_dom_walker::after_dom_children.
10782 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10783 * config/i386/i386.c (ix86_option_override_internal): Added
10784 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10786 2018-01-12 Richard Biener <rguenther@suse.de>
10789 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10790 inline instance vars.
10792 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10795 * config/rx/rx.c (rx_is_restricted_memory_address):
10796 Handle SUBREG case.
10798 2018-01-12 Richard Biener <rguenther@suse.de>
10800 PR tree-optimization/80846
10801 * target.def (split_reduction): New target hook.
10802 * targhooks.c (default_split_reduction): New function.
10803 * targhooks.h (default_split_reduction): Declare.
10804 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10805 target requests first reduce vectors by combining low and high
10807 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10808 (get_vectype_for_scalar_type_and_size): Export.
10809 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10810 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10811 * doc/tm.texi: Regenerate.
10812 * config/i386/i386.c (ix86_split_reduction): Implement
10813 TARGET_VECTORIZE_SPLIT_REDUCTION.
10815 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10818 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10819 in PIC mode except for TARGET_VXWORKS_RTP.
10820 * config/sparc/sparc.c: Include cfgrtl.h.
10821 (TARGET_INIT_PIC_REG): Define.
10822 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10823 (sparc_pic_register_p): New predicate.
10824 (sparc_legitimate_address_p): Use it.
10825 (sparc_legitimize_pic_address): Likewise.
10826 (sparc_delegitimize_address): Likewise.
10827 (sparc_mode_dependent_address_p): Likewise.
10828 (gen_load_pcrel_sym): Remove 4th parameter.
10829 (load_got_register): Adjust call to above. Remove obsolete stuff.
10830 (sparc_expand_prologue): Do not call load_got_register here.
10831 (sparc_flat_expand_prologue): Likewise.
10832 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10833 (sparc_use_pseudo_pic_reg): New function.
10834 (sparc_init_pic_reg): Likewise.
10835 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10836 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10838 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10840 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10841 Add item for branch_cost.
10843 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10845 PR rtl-optimization/83565
10846 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10847 not extend the result to a larger mode for rotate operations.
10848 (num_sign_bit_copies1): Likewise.
10850 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10853 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10855 Use values-Xc.o for -pedantic.
10856 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10858 2018-01-12 Martin Liska <mliska@suse.cz>
10861 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10863 (possible_polymorphic_call_targets): Use it.
10864 (ipa_devirt): Likewise.
10866 2018-01-12 Martin Liska <mliska@suse.cz>
10868 * profile-count.h (enum profile_quality): Use 0 as invalid
10869 enum value of profile_quality.
10871 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10873 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10874 -mext-string options.
10876 2018-01-12 Richard Biener <rguenther@suse.de>
10878 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10879 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10880 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10882 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10884 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10886 * configure.ac (--with-long-double-format): Add support for the
10887 configuration option to change the default long double format on
10889 * config.gcc (powerpc*-linux*-*): Likewise.
10890 * configure: Regenerate.
10891 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10892 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10893 used without modification.
10895 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10897 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10898 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10899 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10900 MISC_BUILTIN_SPEC_BARRIER.
10901 (rs6000_init_builtins): Likewise.
10902 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10904 (speculation_barrier): New define_insn.
10905 * doc/extend.texi: Document __builtin_speculation_barrier.
10907 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10910 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10911 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10912 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10914 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10915 integral modes instead of "ss" and "sd".
10916 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10917 vectors with 32-bit and 64-bit elements.
10918 (vecdupssescalarmodesuffix): New mode attribute.
10919 (vec_dup<mode>): Use it.
10921 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10924 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10925 frame if argument is passed on stack.
10927 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10930 * ree.c (combine_reaching_defs): Optimize also
10931 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10932 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10934 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10936 PR middle-end/83189
10937 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10939 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10941 PR middle-end/83718
10942 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10943 after they are computed.
10945 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10947 PR tree-optimization/83695
10948 * gimple-loop-linterchange.cc
10949 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10950 reset cached scev information after interchange.
10951 (pass_linterchange::execute): Remove call to scev_reset_htab.
10953 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10955 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10956 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10957 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10958 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10959 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10960 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10961 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10962 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10963 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10964 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10965 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10966 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10967 (V_lane_reg): Likewise.
10968 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10970 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10971 (vfmal_lane_low<mode>_intrinsic,
10972 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10973 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10974 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10975 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10976 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10977 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10979 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10981 * config/arm/arm-cpus.in (fp16fml): New feature.
10982 (ALL_SIMD): Add fp16fml.
10983 (armv8.2-a): Add fp16fml as an option.
10984 (armv8.3-a): Likewise.
10985 (armv8.4-a): Add fp16fml as part of fp16.
10986 * config/arm/arm.h (TARGET_FP16FML): Define.
10987 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10989 * config/arm/arm-modes.def (V2HF): Define.
10990 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10991 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10992 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10993 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10994 vfmsl_low, vfmsl_high): New set of builtins.
10995 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10996 (vfml_op): New code attribute.
10997 (VFMLHALVES): New int iterator.
10998 (VFML, VFMLSEL): New mode attributes.
10999 (V_reg): Define mapping for V2HF.
11000 (V_hi, V_lo): New mode attributes.
11001 (VF_constraint): Likewise.
11002 (vfml_half, vfml_half_selector): New int attributes.
11003 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11005 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11006 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11008 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11009 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11010 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11011 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11013 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11014 Document new effective target and option set.
11016 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11018 * config/arm/arm-cpus.in (armv8_4): New feature.
11019 (ARMv8_4a): New fgroup.
11020 (armv8.4-a): New arch.
11021 * config/arm/arm-tables.opt: Regenerate.
11022 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11023 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11024 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11025 Add matching rules for -march=armv8.4-a and extensions.
11026 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11028 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11031 * config/rx/rx.md (BW): New mode attribute.
11032 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11034 2018-01-11 Richard Biener <rguenther@suse.de>
11036 PR tree-optimization/83435
11037 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11038 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11039 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11041 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11042 Alan Hayward <alan.hayward@arm.com>
11043 David Sherwood <david.sherwood@arm.com>
11045 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11047 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11048 (aarch64_print_address_internal): Use it to check for a zero offset.
11050 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11051 Alan Hayward <alan.hayward@arm.com>
11052 David Sherwood <david.sherwood@arm.com>
11054 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11055 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11056 Return a poly_int64 rather than a HOST_WIDE_INT.
11057 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11058 rather than a HOST_WIDE_INT.
11059 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11060 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11061 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11062 final_offset from HOST_WIDE_INT to poly_int64.
11063 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11064 to_constant when getting the number of units in an Advanced SIMD
11066 (aarch64_builtin_vectorized_function): Check for a constant number
11068 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11070 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11071 attribute instead of GET_MODE_NUNITS.
11072 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11073 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11074 GET_MODE_SIZE for fixed-size registers.
11075 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11076 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11077 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11078 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11079 (aarch64_print_operand, aarch64_print_address_internal)
11080 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11081 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11082 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11083 Handle polynomial GET_MODE_SIZE.
11084 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11085 wider than SImode without modification.
11086 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11087 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11088 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11089 passing and returning SVE modes.
11090 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11091 rather than GEN_INT.
11092 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11093 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11094 (aarch64_allocate_and_probe_stack_space): Likewise.
11095 (aarch64_layout_frame): Cope with polynomial offsets.
11096 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11097 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11098 polynomial offsets.
11099 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11100 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11101 poly_int64 rather than a HOST_WIDE_INT.
11102 (aarch64_get_separate_components, aarch64_process_components)
11103 (aarch64_expand_prologue, aarch64_expand_epilogue)
11104 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11105 (aarch64_anchor_offset): New function, split out from...
11106 (aarch64_legitimize_address): ...here.
11107 (aarch64_builtin_vectorization_cost): Handle polynomial
11108 TYPE_VECTOR_SUBPARTS.
11109 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11111 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11112 number of elements from the PARALLEL rather than the mode.
11113 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11114 rather than GET_MODE_BITSIZE.
11115 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11116 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11117 (aarch64_expand_vec_perm_const_1): Handle polynomial
11118 d->perm.length () and d->perm elements.
11119 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11120 Apply to_constant to d->perm elements.
11121 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11122 polynomial CONST_VECTOR_NUNITS.
11123 (aarch64_move_pointer): Take amount as a poly_int64 rather
11125 (aarch64_progress_pointer): Avoid temporary variable.
11126 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11127 the mode attribute instead of GET_MODE.
11129 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11130 Alan Hayward <alan.hayward@arm.com>
11131 David Sherwood <david.sherwood@arm.com>
11133 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11134 x exists before using it.
11135 (aarch64_add_constant_internal): Rename to...
11136 (aarch64_add_offset_1): ...this. Replace regnum with separate
11137 src and dest rtxes. Handle the case in which they're different,
11138 including when the offset is zero. Replace scratchreg with an rtx.
11139 Use 2 additions if there is no spare register into which we can
11140 move a 16-bit constant.
11141 (aarch64_add_constant): Delete.
11142 (aarch64_add_offset): Replace reg with separate src and dest
11143 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11144 Use aarch64_add_offset_1.
11145 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11146 an rtx rather than an int. Take the delta as a poly_int64
11147 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11148 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11149 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11150 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11151 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11152 and aarch64_add_sp.
11153 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11154 aarch64_add_constant.
11156 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11158 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11159 Use scalar_float_mode.
11161 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11163 * config/aarch64/aarch64-simd.md
11164 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11165 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11166 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11167 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11168 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11169 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11170 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11171 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11172 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11173 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11175 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11178 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11179 targ_options->x_arm_arch_string is non NULL.
11181 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11183 * config/aarch64/aarch64.h
11184 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11186 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11189 * expmed.c (emit_store_flag_force): Swap if const op0
11190 and change VOIDmode to mode of op0.
11192 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11194 PR rtl-optimization/83761
11195 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11196 than bytes to mode_for_size.
11198 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11200 PR middle-end/83189
11201 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11202 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11205 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11207 PR middle-end/83575
11208 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11209 when in layout mode.
11210 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11211 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11214 2018-01-10 Michael Collison <michael.collison@arm.com>
11216 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11217 * config/aarch64/aarch64-option-extension.def: Add
11218 AARCH64_OPT_EXTENSION of 'fp16fml'.
11219 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11220 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11221 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11222 * config/aarch64/constraints.md (Ui7): New constraint.
11223 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11224 (VFMLA_SEL_W): Ditto.
11227 (VFMLA16_LOW): New int iterator.
11228 (VFMLA16_HIGH): Ditto.
11229 (UNSPEC_FMLAL): New unspec.
11230 (UNSPEC_FMLSL): Ditto.
11231 (UNSPEC_FMLAL2): Ditto.
11232 (UNSPEC_FMLSL2): Ditto.
11233 (f16mac): New code attribute.
11234 * config/aarch64/aarch64-simd-builtins.def
11235 (aarch64_fmlal_lowv2sf): Ditto.
11236 (aarch64_fmlsl_lowv2sf): Ditto.
11237 (aarch64_fmlalq_lowv4sf): Ditto.
11238 (aarch64_fmlslq_lowv4sf): Ditto.
11239 (aarch64_fmlal_highv2sf): Ditto.
11240 (aarch64_fmlsl_highv2sf): Ditto.
11241 (aarch64_fmlalq_highv4sf): Ditto.
11242 (aarch64_fmlslq_highv4sf): Ditto.
11243 (aarch64_fmlal_lane_lowv2sf): Ditto.
11244 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11245 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11246 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11247 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11248 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11249 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11250 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11251 (aarch64_fmlal_lane_highv2sf): Ditto.
11252 (aarch64_fmlsl_lane_highv2sf): Ditto.
11253 (aarch64_fmlal_laneq_highv2sf): Ditto.
11254 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11255 (aarch64_fmlalq_lane_highv4sf): Ditto.
11256 (aarch64_fmlsl_lane_highv4sf): Ditto.
11257 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11258 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11259 * config/aarch64/aarch64-simd.md:
11260 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11261 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11262 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11263 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11264 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11265 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11266 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11267 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11268 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11269 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11270 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11271 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11272 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11273 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11274 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11275 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11276 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11277 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11278 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11279 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11280 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11281 (vfmlsl_low_u32): Ditto.
11282 (vfmlalq_low_u32): Ditto.
11283 (vfmlslq_low_u32): Ditto.
11284 (vfmlal_high_u32): Ditto.
11285 (vfmlsl_high_u32): Ditto.
11286 (vfmlalq_high_u32): Ditto.
11287 (vfmlslq_high_u32): Ditto.
11288 (vfmlal_lane_low_u32): Ditto.
11289 (vfmlsl_lane_low_u32): Ditto.
11290 (vfmlal_laneq_low_u32): Ditto.
11291 (vfmlsl_laneq_low_u32): Ditto.
11292 (vfmlalq_lane_low_u32): Ditto.
11293 (vfmlslq_lane_low_u32): Ditto.
11294 (vfmlalq_laneq_low_u32): Ditto.
11295 (vfmlslq_laneq_low_u32): Ditto.
11296 (vfmlal_lane_high_u32): Ditto.
11297 (vfmlsl_lane_high_u32): Ditto.
11298 (vfmlal_laneq_high_u32): Ditto.
11299 (vfmlsl_laneq_high_u32): Ditto.
11300 (vfmlalq_lane_high_u32): Ditto.
11301 (vfmlslq_lane_high_u32): Ditto.
11302 (vfmlalq_laneq_high_u32): Ditto.
11303 (vfmlslq_laneq_high_u32): Ditto.
11304 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11305 (AARCH64_FL_FOR_ARCH8_4): New.
11306 (AARCH64_ISA_F16FML): New ISA flag.
11307 (TARGET_F16FML): New feature flag for fp16fml.
11308 (doc/invoke.texi): Document new fp16fml option.
11310 2018-01-10 Michael Collison <michael.collison@arm.com>
11312 * config/aarch64/aarch64-builtins.c:
11313 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11314 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11315 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11316 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11317 (AARCH64_ISA_SHA3): New ISA flag.
11318 (TARGET_SHA3): New feature flag for sha3.
11319 * config/aarch64/iterators.md (sha512_op): New int attribute.
11320 (CRYPTO_SHA512): New int iterator.
11321 (UNSPEC_SHA512H): New unspec.
11322 (UNSPEC_SHA512H2): Ditto.
11323 (UNSPEC_SHA512SU0): Ditto.
11324 (UNSPEC_SHA512SU1): Ditto.
11325 * config/aarch64/aarch64-simd-builtins.def
11326 (aarch64_crypto_sha512hqv2di): New builtin.
11327 (aarch64_crypto_sha512h2qv2di): Ditto.
11328 (aarch64_crypto_sha512su0qv2di): Ditto.
11329 (aarch64_crypto_sha512su1qv2di): Ditto.
11330 (aarch64_eor3qv8hi): Ditto.
11331 (aarch64_rax1qv2di): Ditto.
11332 (aarch64_xarqv2di): Ditto.
11333 (aarch64_bcaxqv8hi): Ditto.
11334 * config/aarch64/aarch64-simd.md:
11335 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11336 (aarch64_crypto_sha512su0qv2di): Ditto.
11337 (aarch64_crypto_sha512su1qv2di): Ditto.
11338 (aarch64_eor3qv8hi): Ditto.
11339 (aarch64_rax1qv2di): Ditto.
11340 (aarch64_xarqv2di): Ditto.
11341 (aarch64_bcaxqv8hi): Ditto.
11342 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11343 (vsha512h2q_u64): Ditto.
11344 (vsha512su0q_u64): Ditto.
11345 (vsha512su1q_u64): Ditto.
11346 (veor3q_u16): Ditto.
11347 (vrax1q_u64): Ditto.
11348 (vxarq_u64): Ditto.
11349 (vbcaxq_u16): Ditto.
11350 * config/arm/types.md (crypto_sha512): New type attribute.
11351 (crypto_sha3): Ditto.
11352 (doc/invoke.texi): Document new sha3 option.
11354 2018-01-10 Michael Collison <michael.collison@arm.com>
11356 * config/aarch64/aarch64-builtins.c:
11357 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11358 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11359 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11360 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11361 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11362 (AARCH64_ISA_SM4): New ISA flag.
11363 (TARGET_SM4): New feature flag for sm4.
11364 * config/aarch64/aarch64-simd-builtins.def
11365 (aarch64_sm3ss1qv4si): Ditto.
11366 (aarch64_sm3tt1aq4si): Ditto.
11367 (aarch64_sm3tt1bq4si): Ditto.
11368 (aarch64_sm3tt2aq4si): Ditto.
11369 (aarch64_sm3tt2bq4si): Ditto.
11370 (aarch64_sm3partw1qv4si): Ditto.
11371 (aarch64_sm3partw2qv4si): Ditto.
11372 (aarch64_sm4eqv4si): Ditto.
11373 (aarch64_sm4ekeyqv4si): Ditto.
11374 * config/aarch64/aarch64-simd.md:
11375 (aarch64_sm3ss1qv4si): Ditto.
11376 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11377 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11378 (aarch64_sm4eqv4si): Ditto.
11379 (aarch64_sm4ekeyqv4si): Ditto.
11380 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11381 (sm3part_op): Ditto.
11382 (CRYPTO_SM3TT): Ditto.
11383 (CRYPTO_SM3PART): Ditto.
11384 (UNSPEC_SM3SS1): New unspec.
11385 (UNSPEC_SM3TT1A): Ditto.
11386 (UNSPEC_SM3TT1B): Ditto.
11387 (UNSPEC_SM3TT2A): Ditto.
11388 (UNSPEC_SM3TT2B): Ditto.
11389 (UNSPEC_SM3PARTW1): Ditto.
11390 (UNSPEC_SM3PARTW2): Ditto.
11391 (UNSPEC_SM4E): Ditto.
11392 (UNSPEC_SM4EKEY): Ditto.
11393 * config/aarch64/constraints.md (Ui2): New constraint.
11394 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11395 * config/arm/types.md (crypto_sm3): New type attribute.
11396 (crypto_sm4): Ditto.
11397 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11398 (vsm3tt1aq_u32): Ditto.
11399 (vsm3tt1bq_u32): Ditto.
11400 (vsm3tt2aq_u32): Ditto.
11401 (vsm3tt2bq_u32): Ditto.
11402 (vsm3partw1q_u32): Ditto.
11403 (vsm3partw2q_u32): Ditto.
11404 (vsm4eq_u32): Ditto.
11405 (vsm4ekeyq_u32): Ditto.
11406 (doc/invoke.texi): Document new sm4 option.
11408 2018-01-10 Michael Collison <michael.collison@arm.com>
11410 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11411 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11412 (AARCH64_FL_FOR_ARCH8_4): New.
11413 (AARCH64_FL_V8_4): New flag.
11414 (doc/invoke.texi): Document new armv8.4-a option.
11416 2018-01-10 Michael Collison <michael.collison@arm.com>
11418 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11419 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11420 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11421 * config/aarch64/aarch64-option-extension.def: Add
11422 AARCH64_OPT_EXTENSION of 'sha2'.
11423 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11424 (crypto): Disable sha2 and aes if crypto disabled.
11425 (crypto): Enable aes and sha2 if enabled.
11426 (simd): Disable sha2 and aes if simd disabled.
11427 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11429 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11430 (TARGET_SHA2): New feature flag for sha2.
11431 (TARGET_AES): New feature flag for aes.
11432 * config/aarch64/aarch64-simd.md:
11433 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11434 conditional on TARGET_AES.
11435 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11436 (aarch64_crypto_sha1hsi): Make pattern conditional
11438 (aarch64_crypto_sha1hv4si): Ditto.
11439 (aarch64_be_crypto_sha1hv4si): Ditto.
11440 (aarch64_crypto_sha1su1v4si): Ditto.
11441 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11442 (aarch64_crypto_sha1su0v4si): Ditto.
11443 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11444 (aarch64_crypto_sha256su0v4si): Ditto.
11445 (aarch64_crypto_sha256su1v4si): Ditto.
11446 (doc/invoke.texi): Document new aes and sha2 options.
11448 2018-01-10 Martin Sebor <msebor@redhat.com>
11450 PR tree-optimization/83781
11451 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11454 2018-01-11 Martin Sebor <msebor@gmail.com>
11455 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11457 PR tree-optimization/83501
11458 PR tree-optimization/81703
11460 * tree-ssa-strlen.c (get_string_cst): Rename...
11461 (get_string_len): ...to this. Handle global constants.
11462 (handle_char_store): Adjust.
11464 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11465 Jim Wilson <jimw@sifive.com>
11467 * config/riscv/riscv-protos.h (riscv_output_return): New.
11468 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11469 (riscv_attribute_table, riscv_output_return),
11470 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11471 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11472 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11474 (riscv_expand_prologue): Add early return for naked function.
11475 (riscv_expand_epilogue): Likewise.
11476 (riscv_function_ok_for_sibcall): Return false for naked function.
11477 (riscv_set_current_function): New.
11478 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11479 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11480 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11481 * doc/extend.texi (RISC-V Function Attributes): New.
11483 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11485 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11486 check for 128-bit long double before checking TCmode.
11487 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11488 128-bit long doubles before checking TFmode or TCmode.
11489 (FLOAT128_IBM_P): Likewise.
11491 2018-01-10 Martin Sebor <msebor@redhat.com>
11493 PR tree-optimization/83671
11494 * builtins.c (c_strlen): Unconditionally return zero for the empty
11496 Use -Warray-bounds for warnings.
11497 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11498 for non-constant array indices with COMPONENT_REF, arrays of
11499 arrays, and pointers to arrays.
11500 (gimple_fold_builtin_strlen): Determine and set length range for
11501 non-constant character arrays.
11503 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11505 PR middle-end/81897
11506 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11509 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11511 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11513 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11516 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11517 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11518 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11519 indexed_or_indirect_operand predicate.
11520 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11521 (*vsx_le_perm_load_v8hi): Likewise.
11522 (*vsx_le_perm_load_v16qi): Likewise.
11523 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11524 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11525 (*vsx_le_perm_store_v8hi): Likewise.
11526 (*vsx_le_perm_store_v16qi): Likewise.
11527 (eight unnamed splitters): Likewise.
11529 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11531 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11532 * config/rs6000/emmintrin.h: Likewise.
11533 * config/rs6000/mmintrin.h: Likewise.
11534 * config/rs6000/xmmintrin.h: Likewise.
11536 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11539 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11541 * tree.c (tree_nop_conversion): Return true for location wrapper
11543 (maybe_wrap_with_location): New function.
11544 (selftest::check_strip_nops): New function.
11545 (selftest::test_location_wrappers): New function.
11546 (selftest::tree_c_tests): Call it.
11547 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11548 (maybe_wrap_with_location): New decl.
11549 (EXPR_LOCATION_WRAPPER_P): New macro.
11550 (location_wrapper_p): New inline function.
11551 (tree_strip_any_location_wrapper): New inline function.
11553 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11556 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11557 stack_realign_offset for the largest alignment of stack slot
11559 (ix86_find_max_used_stack_alignment): New function.
11560 (ix86_finalize_stack_frame_flags): Use it. Set
11561 max_used_stack_alignment if we don't realign stack.
11562 * config/i386/i386.h (machine_function): Add
11563 max_used_stack_alignment.
11565 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11567 * config/arm/arm.opt (-mbranch-cost): New option.
11568 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11571 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11574 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11575 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11577 2018-01-10 Richard Biener <rguenther@suse.de>
11580 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11581 early out so it also covers the case where we have a non-NULL
11584 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11586 PR tree-optimization/83753
11587 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11588 for non-strided grouped accesses if the number of elements is 1.
11590 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11593 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11594 * i386.h (TARGET_USE_GATHER): Define.
11595 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11597 2018-01-10 Martin Liska <mliska@suse.cz>
11600 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11601 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11603 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11604 CLEANUP_NO_PARTITIONING is not set.
11606 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11608 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11609 for vectors, as a partial revert of r254296.
11610 * rtl.h (const_vec_p): Delete.
11611 (const_vec_duplicate_p): Don't test for vector CONSTs.
11612 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11613 * expmed.c (make_tree): Likewise.
11616 * common.md (E, F): Use CONSTANT_P instead of checking for
11618 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11619 checking for CONST_VECTOR.
11621 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11623 PR middle-end/83575
11624 * predict.c (force_edge_cold): Handle in more sane way edges
11625 with no prediction.
11627 2018-01-09 Carl Love <cel@us.ibm.com>
11629 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11631 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11632 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11633 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11634 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11635 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11636 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11637 * config/rs6000/rs6000-protos.h: Add extern defition for
11638 rs6000_generate_float2_double_code.
11639 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11641 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11642 (float2_v2df): Add define_expand.
11644 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11647 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11648 op_mode in the force_to_mode call.
11650 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11652 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11653 instead of checking each element individually.
11654 (aarch64_evpc_uzp): Likewise.
11655 (aarch64_evpc_zip): Likewise.
11656 (aarch64_evpc_ext): Likewise.
11657 (aarch64_evpc_rev): Likewise.
11658 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11659 instead of checking each element individually. Return true without
11661 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11662 whether all selected elements come from the same input, instead of
11663 checking each element individually. Remove calls to gen_rtx_REG,
11664 start_sequence and end_sequence and instead assert that no rtl is
11667 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11669 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11670 order of HIGH and CONST checks.
11672 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11674 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11675 if the destination isn't an SSA_NAME.
11677 2018-01-09 Richard Biener <rguenther@suse.de>
11679 PR tree-optimization/83668
11680 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11682 (canonicalize_loop_form): ... here, renamed from ...
11683 (canonicalize_loop_closed_ssa_form): ... this and amended to
11684 swap successor edges for loop exit blocks to make us use
11685 the RPO order we need for initial schedule generation.
11687 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11689 PR tree-optimization/64811
11690 * match.pd: When optimizing comparisons with Inf, avoid
11691 introducing or losing exceptions from comparisons with NaN.
11693 2018-01-09 Martin Liska <mliska@suse.cz>
11696 * asan.c (shadow_mem_size): Add gcc_assert.
11698 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11700 Don't save registers in main().
11703 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11704 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11705 * config/avr/avr.c (avr_set_current_function): Don't error if
11706 naked, OS_task or OS_main are specified at the same time.
11707 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11709 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11711 * common/config/avr/avr-common.c (avr_option_optimization_table):
11712 Switch on -mmain-is-OS_task for optimizing compilations.
11714 2018-01-09 Richard Biener <rguenther@suse.de>
11716 PR tree-optimization/83572
11717 * graphite.c: Include cfganal.h.
11718 (graphite_transform_loops): Connect infinite loops to exit
11719 and remove fake edges at the end.
11721 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11723 * ipa-inline.c (edge_badness): Revert accidental checkin.
11725 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11728 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11729 symbols; not inline clones.
11731 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11734 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11735 hard registers. Formatting fixes.
11737 PR preprocessor/83722
11738 * gcc.c (try_generate_repro): Pass
11739 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11740 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11743 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11744 Kito Cheng <kito.cheng@gmail.com>
11746 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11747 (riscv_leaf_function_p): Delete.
11748 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11750 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11752 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11754 (do_ifelse): New function.
11755 (do_isel): New function.
11756 (do_sub3): New function.
11757 (do_add3): New function.
11758 (do_load_mask_compare): New function.
11759 (do_overlap_load_compare): New function.
11760 (expand_compare_loop): New function.
11761 (expand_block_compare): Call expand_compare_loop() when appropriate.
11762 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11763 option description.
11764 (-mblock-compare-inline-loop-limit): New option.
11766 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11769 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11770 Reverse order of second and third operands in first alternative.
11771 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11772 of first and second elements in UNSPEC_VPERMR vector.
11773 (altivec_expand_vec_perm_le): Likewise.
11775 2018-01-08 Jeff Law <law@redhat.com>
11777 PR rtl-optimizatin/81308
11778 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11779 (process_switch): If group_case_labels makes a change, then set
11781 (pass_convert_switch::execute): If a switch is converted, then
11782 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11784 PR rtl-optimization/81308
11785 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11788 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11790 PR target/83663 - Revert r255946
11791 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11792 generation for cases where splatting a value is not useful.
11793 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11794 across a vec_duplicate and a paradoxical subreg forming a vector
11795 mode to a vec_concat.
11797 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11799 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11800 -march=armv8.3-a variants.
11801 * config/arm/t-multilib: Likewise.
11802 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11804 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11806 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11808 (cceq_ior_compare_complement): Give it a name so I can use it, and
11809 change boolean_or_operator predicate to boolean_operator so it can
11810 be used to generate a crand.
11811 (eqne): New code iterator.
11812 (bd/bd_neg): New code_attrs.
11813 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11814 a single define_insn.
11815 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11816 decrement (bdnzt/bdnzf/bdzt/bdzf).
11817 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11818 with the new names of the branch decrement patterns, and added the
11819 names of the branch decrement conditional patterns.
11821 2018-01-08 Richard Biener <rguenther@suse.de>
11823 PR tree-optimization/83563
11824 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11827 2018-01-08 Richard Biener <rguenther@suse.de>
11829 PR middle-end/83713
11830 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11832 2018-01-08 Richard Biener <rguenther@suse.de>
11834 PR tree-optimization/83685
11835 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11836 references to abnormals.
11838 2018-01-08 Richard Biener <rguenther@suse.de>
11841 * dwarf2out.c (output_indirect_strings): Handle empty
11842 skeleton_debug_str_hash.
11843 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11845 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11847 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11848 (emit_store_direct): Likewise.
11849 (arc_trampoline_adjust_address): Likewise.
11850 (arc_asm_trampoline_template): New function.
11851 (arc_initialize_trampoline): Use asm_trampoline_template.
11852 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11853 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11854 * config/arc/arc.md (flush_icache): Delete pattern.
11856 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11858 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11859 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11862 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11865 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11866 by not USED_FOR_TARGET.
11867 (make_pass_resolve_sw_modes): Likewise.
11869 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11871 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11874 2018-01-08 Richard Biener <rguenther@suse.de>
11876 PR middle-end/83580
11877 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11879 2018-01-08 Richard Biener <rguenther@suse.de>
11881 PR middle-end/83517
11882 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11884 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11886 PR middle-end/81897
11887 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11888 basic blocks with a small number of successors.
11889 (convert_control_dep_chain_into_preds): Improve handling of
11891 (dump_predicates): Split apart into...
11892 (dump_pred_chain): ...here...
11893 (dump_pred_info): ...and here.
11894 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11895 (can_chain_union_be_invalidated_p): Improve check for invalidation
11897 (uninit_uses_cannot_happen): Avoid unnecessary if
11898 convert_control_dep_chain_into_preds yielded nothing.
11900 2018-01-06 Martin Sebor <msebor@redhat.com>
11902 PR tree-optimization/83640
11903 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11904 subtracting negative offset from size.
11905 (builtin_access::overlap): Adjust offset bounds of the access to fall
11906 within the size of the object if possible.
11908 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11910 PR rtl-optimization/83699
11911 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11912 extract_bit_field_as_subreg to cases in which the extracted
11913 value is also a vector.
11915 * lra-constraints.c (process_alt_operands): Test for the equivalence
11916 substitutions when detecting a possible reload cycle.
11918 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11921 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11922 by default if flag_selective_schedling{,2}. Formatting fixes.
11924 PR rtl-optimization/83682
11925 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11926 if it has non-VECTOR_MODE element mode.
11927 (vec_duplicate_p): Likewise.
11929 PR middle-end/83694
11930 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11931 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11933 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11936 * config/i386/i386-builtin.def
11937 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11938 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11939 Require also OPTION_MASK_ISA_AVX512F in addition to
11940 OPTION_MASK_ISA_GFNI.
11941 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11942 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11943 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11944 to OPTION_MASK_ISA_GFNI.
11945 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11946 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11947 OPTION_MASK_ISA_AVX512BW.
11948 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11949 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11950 addition to OPTION_MASK_ISA_GFNI.
11951 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11952 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11953 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11954 to OPTION_MASK_ISA_GFNI.
11955 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11956 a requirement for all ISAs rather than any of them with a few
11958 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11960 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11961 bitmasks to be enabled with 3 exceptions, instead of requiring any
11962 enabled ISA with lots of exceptions.
11963 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11964 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11965 Change avx512bw in isa attribute to avx512f.
11966 * config/i386/sgxintrin.h: Add license boilerplate.
11967 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11968 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11969 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11970 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11972 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11973 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11974 temporarily sse2 rather than sse if not enabled already.
11977 * config/i386/sse.md (VI248_VLBW): Rename to ...
11978 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11979 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11980 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11981 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11982 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11983 mode iterator instead of VI248_VLBW.
11985 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11987 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11988 (record_modified): Skip clobbers; add debug output.
11989 (param_change_prob): Use sreal frequencies.
11991 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11993 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11994 punt for user-aligned variables.
11996 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11998 * tree-chrec.c (chrec_contains_symbols): Return true for
12001 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12004 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12005 of (x|y) == x for BICS pattern.
12007 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12009 PR tree-optimization/83605
12010 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12011 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12014 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12016 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12017 * config/epiphany/rtems.h: New file.
12019 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12020 Uros Bizjak <ubizjak@gmail.com>
12023 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12024 QIreg_operand instead of register_operand predicate.
12025 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12026 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12027 comments instead of -fmitigate[-_]rop.
12029 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12032 * cgraphunit.c (symbol_table::compile): Switch to text_section
12033 before calling assembly_start debug hook.
12034 * run-rtl-passes.c (run_rtl_passes): Likewise.
12037 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12039 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12040 range_int_cst_p rather than !symbolic_range_p before calling
12041 extract_range_from_multiplicative_op_1.
12043 2018-01-04 Jeff Law <law@redhat.com>
12045 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12046 redundant test in assertion.
12048 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12050 * doc/rtl.texi: Document machine_mode wrapper classes.
12052 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12054 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12055 using tree_to_uhwi.
12057 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12059 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12060 the VEC_PERM_EXPR fold to fail.
12062 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12065 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12066 to switched_sections.
12068 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12071 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12072 test for d.testing.
12074 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12077 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12078 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12080 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12083 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12084 is BLKmode and bitpos not zero or mode change is needed.
12086 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12089 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12092 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12095 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12096 instead of MULT rtx. Update all corresponding splitters.
12097 (*saddl_se): Ditto.
12098 (*ssub<modesuffix>): Ditto.
12099 (*ssubl_se): Ditto.
12100 (*cmp_sadd_di): Update split patterns.
12101 (*cmp_sadd_si): Ditto.
12102 (*cmp_sadd_sidi): Ditto.
12103 (*cmp_ssub_di): Ditto.
12104 (*cmp_ssub_si): Ditto.
12105 (*cmp_ssub_sidi): Ditto.
12106 * config/alpha/predicates.md (const23_operand): New predicate.
12107 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12108 Look for ASHIFT, not MULT inner operand.
12109 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12111 2018-01-04 Martin Liska <mliska@suse.cz>
12113 PR gcov-profile/83669
12114 * gcov.c (output_intermediate_file): Add version to intermediate
12116 * doc/gcov.texi: Document new field 'version' in intermediate
12117 file format. Fix location of '-k' option of gcov command.
12119 2018-01-04 Martin Liska <mliska@suse.cz>
12122 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12124 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12126 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12128 2018-01-03 Martin Sebor <msebor@redhat.com>
12130 PR tree-optimization/83655
12131 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12132 checking calls with invalid arguments.
12134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12136 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12137 (vectorizable_mask_load_store): Delete.
12138 (vectorizable_call): Return false for masked loads and stores.
12139 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12140 instead of gimple_assign_rhs1.
12141 (vectorizable_load): Handle IFN_MASK_LOAD.
12142 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12146 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12148 (vectorizable_mask_load_store): ...here.
12149 (vectorizable_load): ...and here.
12151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12153 * tree-vect-stmts.c (vect_build_all_ones_mask)
12154 (vect_build_zero_merge_argument): New functions, split out from...
12155 (vectorizable_load): ...here.
12157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12159 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12161 (vectorizable_mask_load_store): ...here.
12162 (vectorizable_store): ...and here.
12164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12166 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12168 (vectorizable_mask_load_store): ...here.
12170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12172 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12173 (vect_model_store_cost): Take a vec_load_store_type instead of a
12175 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12176 (vect_model_store_cost): Take a vec_load_store_type instead of a
12178 (vectorizable_mask_load_store): Update accordingly.
12179 (vectorizable_store): Likewise.
12180 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12184 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12185 IFN_MASK_LOAD calls here rather than...
12186 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12189 Alan Hayward <alan.hayward@arm.com>
12190 David Sherwood <david.sherwood@arm.com>
12192 * expmed.c (extract_bit_field_1): For vector extracts,
12193 fall back to extract_bit_field_as_subreg if vec_extract
12196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12197 Alan Hayward <alan.hayward@arm.com>
12198 David Sherwood <david.sherwood@arm.com>
12200 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12201 they are variable or constant sized.
12202 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12203 slots for constant-sized data.
12205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12206 Alan Hayward <alan.hayward@arm.com>
12207 David Sherwood <david.sherwood@arm.com>
12209 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12210 handling COND_EXPRs with boolean comparisons, try to find a better
12211 basis for the mask type than the boolean itself.
12213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12215 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12216 is calculated and how it can be overridden.
12217 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12218 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12220 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12224 Alan Hayward <alan.hayward@arm.com>
12225 David Sherwood <david.sherwood@arm.com>
12227 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12228 Remove the mode argument.
12229 (aarch64_simd_valid_immediate): Remove the mode and inverse
12231 * config/aarch64/iterators.md (bitsize): New iterator.
12232 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12233 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12234 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12235 aarch64_simd_valid_immediate.
12236 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12237 (aarch64_reg_or_bic_imm): Likewise.
12238 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12239 with an insn_type enum and msl with a modifier_type enum.
12240 Replace element_width with a scalar_mode. Change the shift
12241 to unsigned int. Add constructors for scalar_float_mode and
12242 scalar_int_mode elements.
12243 (aarch64_vect_float_const_representable_p): Delete.
12244 (aarch64_can_const_movi_rtx_p)
12245 (aarch64_simd_scalar_immediate_valid_for_move)
12246 (aarch64_simd_make_constant): Update call to
12247 aarch64_simd_valid_immediate.
12248 (aarch64_advsimd_valid_immediate_hs): New function.
12249 (aarch64_advsimd_valid_immediate): Likewise.
12250 (aarch64_simd_valid_immediate): Remove mode and inverse
12251 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12252 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12253 and aarch64_float_const_representable_p on the result.
12254 (aarch64_output_simd_mov_immediate): Remove mode argument.
12255 Update call to aarch64_simd_valid_immediate and use of
12256 simd_immediate_info.
12257 (aarch64_output_scalar_simd_mov_immediate): Update call
12260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12261 Alan Hayward <alan.hayward@arm.com>
12262 David Sherwood <david.sherwood@arm.com>
12264 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12265 (mode_nunits): Likewise CONST_MODE_NUNITS.
12266 * machmode.def (ADJUST_NUNITS): Document.
12267 * genmodes.c (mode_data::need_nunits_adj): New field.
12268 (blank_mode): Update accordingly.
12269 (adj_nunits): New variable.
12270 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12272 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12273 listed in adj_nunits.
12274 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12275 listed in adj_nunits. Don't emit case statements for such modes.
12276 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12277 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12278 nothing if adj_nunits is nonnull.
12279 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12280 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12281 (emit_mode_fbit): Update use of print_maybe_const_decl.
12282 (emit_move_size): Likewise. Treat the array as non-const
12284 (emit_mode_adjustments): Handle adj_nunits.
12286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12288 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12289 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12290 (VECTOR_MODES): Use it.
12291 (make_vector_modes): Take the prefix as an argument.
12293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12294 Alan Hayward <alan.hayward@arm.com>
12295 David Sherwood <david.sherwood@arm.com>
12297 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12298 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12299 for MODE_VECTOR_BOOL.
12300 * machmode.def (VECTOR_BOOL_MODE): Document.
12301 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12302 (make_vector_bool_mode): New function.
12303 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12305 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12306 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12308 * stor-layout.c (int_mode_for_mode): Likewise.
12309 * tree.c (build_vector_type_for_mode): Likewise.
12310 * varasm.c (output_constant_pool_2): Likewise.
12311 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12312 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12313 for MODE_VECTOR_BOOL.
12314 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12315 of mode class checks.
12316 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12317 instead of a list of mode class checks.
12318 (expand_vector_scalar_condition): Likewise.
12319 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12322 Alan Hayward <alan.hayward@arm.com>
12323 David Sherwood <david.sherwood@arm.com>
12325 * machmode.h (mode_size): Change from unsigned short to
12327 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12328 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12329 or if measurement_type is not polynomial.
12330 (fixed_size_mode::includes_p): Check for constant-sized modes.
12331 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12332 return a poly_uint16 rather than an unsigned short.
12333 (emit_mode_size): Change the type of mode_size from unsigned short
12334 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12335 (emit_mode_adjustments): Cope with polynomial vector sizes.
12336 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12338 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12340 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12341 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12342 * caller-save.c (setup_save_areas): Likewise.
12343 (replace_reg_with_saved_mem): Likewise.
12344 * calls.c (emit_library_call_value_1): Likewise.
12345 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12346 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12347 (gen_lowpart_for_combine): Likewise.
12348 * convert.c (convert_to_integer_1): Likewise.
12349 * cse.c (equiv_constant, cse_insn): Likewise.
12350 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12351 (cselib_subst_to_values): Likewise.
12352 * dce.c (word_dce_process_block): Likewise.
12353 * df-problems.c (df_word_lr_mark_ref): Likewise.
12354 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12355 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12356 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12357 (rtl_for_decl_location): Likewise.
12358 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12359 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12360 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12361 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12362 (expand_expr_real_1): Likewise.
12363 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12364 (pad_below): Likewise.
12365 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12366 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12367 * ira.c (get_subreg_tracking_sizes): Likewise.
12368 * ira-build.c (ira_create_allocno_objects): Likewise.
12369 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12370 (ira_sort_regnos_for_alter_reg): Likewise.
12371 * ira-costs.c (record_operand_costs): Likewise.
12372 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12373 (resolve_simple_move): Likewise.
12374 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12375 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12376 (lra_constraints): Likewise.
12377 (CONST_POOL_OK_P): Reject variable-sized modes.
12378 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12379 (add_pseudo_to_slot, lra_spill): Likewise.
12380 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12381 * optabs-query.c (get_best_extraction_insn): Likewise.
12382 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12383 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12384 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12385 * recog.c (offsettable_address_addr_space_p): Likewise.
12386 * regcprop.c (maybe_mode_change): Likewise.
12387 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12388 * regrename.c (build_def_use): Likewise.
12389 * regstat.c (dump_reg_info): Likewise.
12390 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12391 (find_reloads, find_reloads_subreg_address): Likewise.
12392 * reload1.c (eliminate_regs_1): Likewise.
12393 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12394 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12395 (simplify_binary_operation_1, simplify_subreg): Likewise.
12396 * targhooks.c (default_function_arg_padding): Likewise.
12397 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12398 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12399 (verify_gimple_assign_ternary): Likewise.
12400 * tree-inline.c (estimate_move_cost): Likewise.
12401 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12402 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12403 (get_address_cost_ainc): Likewise.
12404 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12405 (vect_supportable_dr_alignment): Likewise.
12406 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12407 (vectorizable_reduction): Likewise.
12408 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12409 (vectorizable_operation, vectorizable_load): Likewise.
12410 * tree.c (build_same_sized_truth_vector_type): Likewise.
12411 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12412 * var-tracking.c (emit_note_insn_var_location): Likewise.
12413 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12414 (ADDR_VEC_ALIGN): Likewise.
12416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12417 Alan Hayward <alan.hayward@arm.com>
12418 David Sherwood <david.sherwood@arm.com>
12420 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12422 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12423 or if measurement_type is polynomial.
12424 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12425 * combine.c (make_extraction): Likewise.
12426 * dse.c (find_shift_sequence): Likewise.
12427 * dwarf2out.c (mem_loc_descriptor): Likewise.
12428 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12429 (extract_bit_field, extract_low_bits): Likewise.
12430 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12431 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12432 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12433 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12434 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12435 * reload.c (find_reloads): Likewise.
12436 * reload1.c (alter_reg): Likewise.
12437 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12438 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12439 * tree-if-conv.c (predicate_mem_writes): Likewise.
12440 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12441 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12442 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12443 * valtrack.c (dead_debug_insert_temp): Likewise.
12444 * varasm.c (mergeable_constant_section): Likewise.
12445 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12448 Alan Hayward <alan.hayward@arm.com>
12449 David Sherwood <david.sherwood@arm.com>
12451 * expr.c (expand_assignment): Cope with polynomial mode sizes
12452 when assigning to a CONCAT.
12454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12455 Alan Hayward <alan.hayward@arm.com>
12456 David Sherwood <david.sherwood@arm.com>
12458 * machmode.h (mode_precision): Change from unsigned short to
12460 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12462 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12463 or if measurement_type is not polynomial.
12464 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12465 in which the mode is already known to be a scalar_int_mode.
12466 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12467 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12469 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12470 for GET_MODE_PRECISION.
12471 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12472 for GET_MODE_PRECISION.
12473 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12475 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12476 (expand_field_assignment, make_extraction): Likewise.
12477 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12478 (get_last_value): Likewise.
12479 * convert.c (convert_to_integer_1): Likewise.
12480 * cse.c (cse_insn): Likewise.
12481 * expr.c (expand_expr_real_1): Likewise.
12482 * lra-constraints.c (simplify_operand_subreg): Likewise.
12483 * optabs-query.c (can_atomic_load_p): Likewise.
12484 * optabs.c (expand_atomic_load): Likewise.
12485 (expand_atomic_store): Likewise.
12486 * ree.c (combine_reaching_defs): Likewise.
12487 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12488 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12489 * tree.h (type_has_mode_precision_p): Likewise.
12490 * ubsan.c (instrument_si_overflow): Likewise.
12492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12493 Alan Hayward <alan.hayward@arm.com>
12494 David Sherwood <david.sherwood@arm.com>
12496 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12497 polynomial numbers of units.
12498 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12499 (valid_vector_subparts_p): New function.
12500 (build_vector_type): Remove temporary shim and take the number
12501 of units as a poly_uint64 rather than an int.
12502 (build_opaque_vector_type): Take the number of units as a
12503 poly_uint64 rather than an int.
12504 * tree.c (build_vector_from_ctor): Handle polynomial
12505 TYPE_VECTOR_SUBPARTS.
12506 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12507 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12508 (build_vector_from_val): If the number of units is variable,
12509 use build_vec_duplicate_cst for constant operands and
12510 VEC_DUPLICATE_EXPR otherwise.
12511 (make_vector_type): Remove temporary is_constant ().
12512 (build_vector_type, build_opaque_vector_type): Take the number of
12513 units as a poly_uint64 rather than an int.
12514 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12516 * cfgexpand.c (expand_debug_expr): Likewise.
12517 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12518 (store_constructor, expand_expr_real_1): Likewise.
12519 (const_scalar_mask_from_tree): Likewise.
12520 * fold-const-call.c (fold_const_reduction): Likewise.
12521 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12522 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12523 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12524 (fold_relational_const): Likewise.
12525 (native_interpret_vector): Likewise. Change the size from an
12526 int to an unsigned int.
12527 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12528 TYPE_VECTOR_SUBPARTS.
12529 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12530 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12531 duplicating a non-constant operand into a variable-length vector.
12532 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12533 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12534 * ipa-icf.c (sem_variable::equals): Likewise.
12535 * match.pd: Likewise.
12536 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12537 * print-tree.c (print_node): Likewise.
12538 * stor-layout.c (layout_type): Likewise.
12539 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12540 * tree-cfg.c (verify_gimple_comparison): Likewise.
12541 (verify_gimple_assign_binary): Likewise.
12542 (verify_gimple_assign_ternary): Likewise.
12543 (verify_gimple_assign_single): Likewise.
12544 * tree-pretty-print.c (dump_generic_node): Likewise.
12545 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12546 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12547 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12548 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12549 (vect_shift_permute_load_chain): Likewise.
12550 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12551 (expand_vector_condition, optimize_vector_constructor): Likewise.
12552 (lower_vec_perm, get_compute_type): Likewise.
12553 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12554 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12555 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12556 (vect_recog_mask_conversion_pattern): Likewise.
12557 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12558 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12559 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12560 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12561 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12562 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12563 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12564 (supportable_widening_operation): Likewise.
12565 (supportable_narrowing_operation): Likewise.
12566 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12568 * varasm.c (output_constant): Likewise.
12570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12571 Alan Hayward <alan.hayward@arm.com>
12572 David Sherwood <david.sherwood@arm.com>
12574 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12575 so that both the length == 3 and length != 3 cases set up their
12576 own permute vectors. Add comments explaining why we know the
12577 number of elements is constant.
12578 (vect_permute_load_chain): Likewise.
12580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12581 Alan Hayward <alan.hayward@arm.com>
12582 David Sherwood <david.sherwood@arm.com>
12584 * machmode.h (mode_nunits): Change from unsigned char to
12586 (ONLY_FIXED_SIZE_MODES): New macro.
12587 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12588 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12589 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12591 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12592 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12593 or if measurement_type is not polynomial.
12594 * genmodes.c (ZERO_COEFFS): New macro.
12595 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12597 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12598 Use ZERO_COEFFS when emitting initializers.
12599 * data-streamer.h (bp_pack_poly_value): New function.
12600 (bp_unpack_poly_value): Likewise.
12601 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12602 for GET_MODE_NUNITS.
12603 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12604 for GET_MODE_NUNITS.
12605 * tree.c (make_vector_type): Remove temporary shim and make
12606 the real function take the number of units as a poly_uint64
12607 rather than an int.
12608 (build_vector_type_for_mode): Handle polynomial nunits.
12609 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12610 * emit-rtl.c (const_vec_series_p_1): Likewise.
12611 (gen_rtx_CONST_VECTOR): Likewise.
12612 * fold-const.c (test_vec_duplicate_folding): Likewise.
12613 * genrecog.c (validate_pattern): Likewise.
12614 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12615 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12616 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12617 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12618 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12619 * rtlanal.c (subreg_get_info): Likewise.
12620 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12621 (vect_grouped_load_supported): Likewise.
12622 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12623 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12624 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12625 (simplify_const_unary_operation, simplify_binary_operation_1)
12626 (simplify_const_binary_operation, simplify_ternary_operation)
12627 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12628 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12629 instead of CONST_VECTOR_NUNITS.
12630 * varasm.c (output_constant_pool_2): Likewise.
12631 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12632 explicit-encoded elements in the XVEC for variable-length vectors.
12634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12636 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12639 Alan Hayward <alan.hayward@arm.com>
12640 David Sherwood <david.sherwood@arm.com>
12642 * coretypes.h (fixed_size_mode): Declare.
12643 (fixed_size_mode_pod): New typedef.
12644 * builtins.h (target_builtins::x_apply_args_mode)
12645 (target_builtins::x_apply_result_mode): Change type to
12646 fixed_size_mode_pod.
12647 * builtins.c (apply_args_size, apply_result_size, result_vector)
12648 (expand_builtin_apply_args_1, expand_builtin_apply)
12649 (expand_builtin_return): Update accordingly.
12651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12653 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12654 * cselib.c (cselib_hash_rtx): Likewise.
12655 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12656 CONST_VECTOR encoding.
12658 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12659 Jeff Law <law@redhat.com>
12662 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12663 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12664 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12665 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12668 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12669 explicitly probe *sp in a noreturn function if there were any callee
12670 register saves or frame pointer is needed.
12672 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12675 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12676 BLKmode for ternary, binary or unary expressions.
12679 * var-tracking.c (delete_vta_debug_insn): New inline function.
12680 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12681 insns from get_insns () to NULL instead of each bb separately.
12682 Use delete_vta_debug_insn. No longer static.
12683 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12684 delete_vta_debug_insns callers.
12685 * rtl.h (delete_vta_debug_insns): Declare.
12686 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12687 instead of variable_tracking_main.
12689 2018-01-03 Martin Sebor <msebor@redhat.com>
12691 PR tree-optimization/83603
12692 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12693 arguments past the endof the argument list in functions declared
12694 without a prototype.
12695 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12696 Avoid checking when arguments are null.
12698 2018-01-03 Martin Sebor <msebor@redhat.com>
12701 * doc/extend.texi (attribute const): Fix a typo.
12702 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12703 issuing -Wsuggest-attribute for void functions.
12705 2018-01-03 Martin Sebor <msebor@redhat.com>
12707 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12708 offset_int::from instead of wide_int::to_shwi.
12709 (maybe_diag_overlap): Remove assertion.
12710 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12711 * gimple-ssa-sprintf.c (format_directive): Same.
12712 (parse_directive): Same.
12713 (sprintf_dom_walker::compute_format_length): Same.
12714 (try_substitute_return_value): Same.
12716 2018-01-03 Jeff Law <law@redhat.com>
12718 PR middle-end/83654
12719 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12720 non-constant residual for zero at runtime and avoid probing in
12721 that case. Reorganize code for trailing problem to mirror handling
12724 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12726 PR tree-optimization/83501
12727 * tree-ssa-strlen.c (get_string_cst): New.
12728 (handle_char_store): Call get_string_cst.
12730 2018-01-03 Martin Liska <mliska@suse.cz>
12732 PR tree-optimization/83593
12733 * tree-ssa-strlen.c: Include tree-cfg.h.
12734 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12735 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12736 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12738 (strlen_dom_walker::before_dom_children): Call
12739 gimple_purge_dead_eh_edges. Dump tranformation with details
12741 (strlen_dom_walker::before_dom_children): Update call by adding
12742 new argument cleanup_eh.
12743 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12745 2018-01-03 Martin Liska <mliska@suse.cz>
12748 * cif-code.def (VARIADIC_THUNK): New enum value.
12749 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12752 2018-01-03 Jan Beulich <jbeulich@suse.com>
12754 * sse.md (mov<mode>_internal): Tighten condition for when to use
12755 vmovdqu<ssescalarsize> for TI and OI modes.
12757 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12759 Update copyright years.
12761 2018-01-03 Martin Liska <mliska@suse.cz>
12764 * ipa-visibility.c (function_and_variable_visibility): Skip
12765 functions with noipa attribure.
12767 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12769 * gcc.c (process_command): Update copyright notice dates.
12770 * gcov-dump.c (print_version): Ditto.
12771 * gcov.c (print_version): Ditto.
12772 * gcov-tool.c (print_version): Ditto.
12773 * gengtype.c (create_file): Ditto.
12774 * doc/cpp.texi: Bump @copying's copyright year.
12775 * doc/cppinternals.texi: Ditto.
12776 * doc/gcc.texi: Ditto.
12777 * doc/gccint.texi: Ditto.
12778 * doc/gcov.texi: Ditto.
12779 * doc/install.texi: Ditto.
12780 * doc/invoke.texi: Ditto.
12782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12784 * vector-builder.h (vector_builder::m_full_nelts): Change from
12785 unsigned int to poly_uint64.
12786 (vector_builder::full_nelts): Update prototype accordingly.
12787 (vector_builder::new_vector): Likewise.
12788 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12789 (vector_builder::operator ==): Likewise.
12790 (vector_builder::finalize): Likewise.
12791 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12792 Take the number of elements as a poly_uint64 rather than an
12794 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12795 from unsigned int to poly_uint64.
12796 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12797 (vec_perm_indices::new_vector): Likewise.
12798 (vec_perm_indices::length): Likewise.
12799 (vec_perm_indices::nelts_per_input): Likewise.
12800 (vec_perm_indices::input_nelts): Likewise.
12801 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12802 number of elements per input as a poly_uint64 rather than an
12803 unsigned int. Use the original encoding for variable-length
12804 vectors, rather than clamping each individual element.
12805 For the second and subsequent elements in each pattern,
12806 clamp the step and base before clamping their sum.
12807 (vec_perm_indices::series_p): Handle polynomial element counts.
12808 (vec_perm_indices::all_in_range_p): Likewise.
12809 (vec_perm_indices_to_tree): Likewise.
12810 (vec_perm_indices_to_rtx): Likewise.
12811 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12812 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12813 (tree_vector_builder::new_binary_operation): Handle polynomial
12814 element counts. Return false if we need to know the number
12815 of elements at compile time.
12816 * fold-const.c (fold_vec_perm): Punt if the number of elements
12817 isn't known at compile time.
12819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12821 * vec-perm-indices.h (vec_perm_builder): Change element type
12822 from HOST_WIDE_INT to poly_int64.
12823 (vec_perm_indices::element_type): Update accordingly.
12824 (vec_perm_indices::clamp): Handle polynomial element_types.
12825 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12826 (vec_perm_indices::all_in_range_p): Likewise.
12827 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12829 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12830 polynomial vec_perm_indices element types.
12831 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12832 * fold-const.c (fold_vec_perm): Likewise.
12833 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12834 * tree-vect-generic.c (lower_vec_perm): Likewise.
12835 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12836 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12837 element type to HOST_WIDE_INT.
12839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12840 Alan Hayward <alan.hayward@arm.com>
12841 David Sherwood <david.sherwood@arm.com>
12843 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12844 rather than an int. Use plus_constant.
12845 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12846 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12849 Alan Hayward <alan.hayward@arm.com>
12850 David Sherwood <david.sherwood@arm.com>
12852 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12853 a HOST_WIDE_INT to a poly_int64.
12855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12856 Alan Hayward <alan.hayward@arm.com>
12857 David Sherwood <david.sherwood@arm.com>
12859 * calls.c (load_register_parameters): Cope with polynomial
12860 mode sizes. Require a constant size for BLKmode parameters
12861 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12862 forces a parameter to be padded at the lsb end in order to
12863 fill a complete number of words, require the parameter size
12864 to be ordered wrt UNITS_PER_WORD.
12866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12867 Alan Hayward <alan.hayward@arm.com>
12868 David Sherwood <david.sherwood@arm.com>
12870 * reload1.c (spill_stack_slot_width): Change element type
12871 from unsigned int to poly_uint64_pod.
12872 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
12879 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
12886 operands whose sizes aren't ordered.
12887 (match_reload): Refer to this check here.
12889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12890 Alan Hayward <alan.hayward@arm.com>
12891 David Sherwood <david.sherwood@arm.com>
12893 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12894 that the mode size is in the set {1, 2, 4, 8, 16}.
12896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12897 Alan Hayward <alan.hayward@arm.com>
12898 David Sherwood <david.sherwood@arm.com>
12900 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12901 Use plus_constant instead of gen_rtx_PLUS.
12903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12904 Alan Hayward <alan.hayward@arm.com>
12905 David Sherwood <david.sherwood@arm.com>
12907 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12908 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12909 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12910 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12911 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12912 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12913 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12914 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12915 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12916 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12918 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12919 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12920 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12921 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12922 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12923 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12924 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12925 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12926 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12927 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12929 * expr.c (emit_move_resolve_push): Treat the input and result
12930 of PUSH_ROUNDING as a poly_int64.
12931 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12932 (emit_push_insn): Likewise.
12933 * lra-eliminations.c (mark_not_eliminable): Likewise.
12934 * recog.c (push_operand): Likewise.
12935 * reload1.c (elimination_effects): Likewise.
12936 * rtlanal.c (nonzero_bits1): Likewise.
12937 * calls.c (store_one_arg): Likewise. Require the padding to be
12938 known at compile time.
12940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12941 Alan Hayward <alan.hayward@arm.com>
12942 David Sherwood <david.sherwood@arm.com>
12944 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12945 Use plus_constant instead of gen_rtx_PLUS.
12947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12948 Alan Hayward <alan.hayward@arm.com>
12949 David Sherwood <david.sherwood@arm.com>
12951 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12952 rather than an int.
12954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12955 Alan Hayward <alan.hayward@arm.com>
12956 David Sherwood <david.sherwood@arm.com>
12958 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12959 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12960 via stack temporaries. Treat the mode size as polynomial too.
12962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12963 Alan Hayward <alan.hayward@arm.com>
12964 David Sherwood <david.sherwood@arm.com>
12966 * expr.c (expand_expr_real_2): When handling conversions involving
12967 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12968 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12969 as a poly_uint64 too.
12971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12972 Alan Hayward <alan.hayward@arm.com>
12973 David Sherwood <david.sherwood@arm.com>
12975 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12978 Alan Hayward <alan.hayward@arm.com>
12979 David Sherwood <david.sherwood@arm.com>
12981 * combine.c (can_change_dest_mode): Handle polynomial
12982 REGMODE_NATURAL_SIZE.
12983 * expmed.c (store_bit_field_1): Likewise.
12984 * expr.c (store_constructor): Likewise.
12985 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12986 and polynomial REGMODE_NATURAL_SIZE.
12987 (gen_lowpart_common): Likewise.
12988 * reginfo.c (record_subregs_of_mode): Likewise.
12989 * rtlanal.c (read_modify_subreg_p): Likewise.
12991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12992 Alan Hayward <alan.hayward@arm.com>
12993 David Sherwood <david.sherwood@arm.com>
12995 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12996 numbers of elements.
12998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12999 Alan Hayward <alan.hayward@arm.com>
13000 David Sherwood <david.sherwood@arm.com>
13002 * match.pd: Cope with polynomial numbers of vector elements.
13004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13005 Alan Hayward <alan.hayward@arm.com>
13006 David Sherwood <david.sherwood@arm.com>
13008 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13009 in a POINTER_PLUS_EXPR.
13011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13012 Alan Hayward <alan.hayward@arm.com>
13013 David Sherwood <david.sherwood@arm.com>
13015 * omp-simd-clone.c (simd_clone_subparts): New function.
13016 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13017 (ipa_simd_modify_function_body): Likewise.
13019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13020 Alan Hayward <alan.hayward@arm.com>
13021 David Sherwood <david.sherwood@arm.com>
13023 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13024 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13025 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13026 (expand_vector_condition, vector_element): Likewise.
13027 (subparts_gt): New function.
13028 (get_compute_type): Use subparts_gt.
13029 (count_type_subparts): Delete.
13030 (expand_vector_operations_1): Use subparts_gt instead of
13031 count_type_subparts.
13033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13034 Alan Hayward <alan.hayward@arm.com>
13035 David Sherwood <david.sherwood@arm.com>
13037 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13038 (vect_compile_time_alias): ...this new function. Do the calculation
13039 on poly_ints rather than trees.
13040 (vect_prune_runtime_alias_test_list): Update call accordingly.
13042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13043 Alan Hayward <alan.hayward@arm.com>
13044 David Sherwood <david.sherwood@arm.com>
13046 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13048 (vect_schedule_slp_instance): Likewise.
13050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13051 Alan Hayward <alan.hayward@arm.com>
13052 David Sherwood <david.sherwood@arm.com>
13054 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13055 constant and extern definitions for variable-length vectors.
13056 (vect_get_constant_vectors): Note that the number of units
13057 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
13064 of units as polynomial. Choose between WIDE and NARROW based
13067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13068 Alan Hayward <alan.hayward@arm.com>
13069 David Sherwood <david.sherwood@arm.com>
13071 * tree-vect-stmts.c (simd_clone_subparts): New function.
13072 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13075 Alan Hayward <alan.hayward@arm.com>
13076 David Sherwood <david.sherwood@arm.com>
13078 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13079 vectors as polynomial. Use build_index_vector for
13080 IFN_GOMP_SIMD_LANE.
13082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13083 Alan Hayward <alan.hayward@arm.com>
13084 David Sherwood <david.sherwood@arm.com>
13086 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13087 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13088 for variable-length vectors.
13089 (vectorizable_mask_load_store): Treat the number of units as
13090 polynomial, asserting that it is constant if the condition has
13091 already been enforced.
13092 (vectorizable_store, vectorizable_load): Likewise.
13094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13095 Alan Hayward <alan.hayward@arm.com>
13096 David Sherwood <david.sherwood@arm.com>
13098 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13099 of units as polynomial. Punt if we can't tell at compile time
13100 which vector contains the final result.
13102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13103 Alan Hayward <alan.hayward@arm.com>
13104 David Sherwood <david.sherwood@arm.com>
13106 * tree-vect-loop.c (vectorizable_induction): Treat the number
13107 of units as polynomial. Punt on SLP inductions. Use an integer
13108 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13109 cast of such a series for variable-length floating-point
13112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13113 Alan Hayward <alan.hayward@arm.com>
13114 David Sherwood <david.sherwood@arm.com>
13116 * tree.h (build_index_vector): Declare.
13117 * tree.c (build_index_vector): New function.
13118 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13119 of units as polynomial, forcibly converting it to a constant if
13120 vectorizable_reduction has already enforced the condition.
13121 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13122 to create a {1,2,3,...} vector.
13123 (vectorizable_reduction): Treat the number of units as polynomial.
13124 Choose vectype_in based on the largest scalar element size rather
13125 than the smallest number of units. Enforce the restrictions
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-data-refs.c (vector_alignment_reachable_p): Treat the
13133 number of units as polynomial.
13135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13136 Alan Hayward <alan.hayward@arm.com>
13137 David Sherwood <david.sherwood@arm.com>
13139 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13140 * target.def (autovectorize_vector_sizes): Return the vector sizes
13141 by pointer, using vector_sizes rather than a bitmask.
13142 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13143 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13144 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13146 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13147 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13148 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13149 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13150 * omp-general.c (omp_max_vf): Likewise.
13151 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13152 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13153 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13154 * tree-vect-slp.c (vect_slp_bb): Likewise.
13155 * doc/tm.texi: Regenerate.
13156 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13158 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13159 the vector size as a poly_uint64 rather than an unsigned int.
13160 (current_vector_size): Change from an unsigned int to a poly_uint64.
13161 (get_vectype_for_scalar_type): Update accordingly.
13162 * tree.h (build_truth_vector_type): Take the size and number of
13163 units as a poly_uint64 rather than an unsigned int.
13164 (build_vector_type): Add a temporary overload that takes
13165 the number of units as a poly_uint64 rather than an unsigned int.
13166 * tree.c (make_vector_type): Likewise.
13167 (build_truth_vector_type): Take the number of units as a poly_uint64
13168 rather than an unsigned int.
13170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13171 Alan Hayward <alan.hayward@arm.com>
13172 David Sherwood <david.sherwood@arm.com>
13174 * target.def (get_mask_mode): Take the number of units and length
13175 as poly_uint64s rather than unsigned ints.
13176 * targhooks.h (default_get_mask_mode): Update accordingly.
13177 * targhooks.c (default_get_mask_mode): Likewise.
13178 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13179 * doc/tm.texi: Regenerate.
13181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13182 Alan Hayward <alan.hayward@arm.com>
13183 David Sherwood <david.sherwood@arm.com>
13185 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13186 * omp-general.c (omp_max_vf): Likewise.
13187 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13188 (expand_omp_simd): Handle polynomial safelen.
13189 * omp-low.c (omplow_simd_context): Add a default constructor.
13190 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13191 (lower_rec_simd_input_clauses): Update accordingly.
13192 (lower_rec_input_clauses): Likewise.
13194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13195 Alan Hayward <alan.hayward@arm.com>
13196 David Sherwood <david.sherwood@arm.com>
13198 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13199 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13200 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13201 (vect_analyze_slp_cost): Likewise.
13202 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13203 (vect_model_load_cost): Likewise.
13205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13206 Alan Hayward <alan.hayward@arm.com>
13207 David Sherwood <david.sherwood@arm.com>
13209 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13210 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13211 from an unsigned int * to a poly_uint64_pod *.
13212 (calculate_unrolling_factor): New function.
13213 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13216 Alan Hayward <alan.hayward@arm.com>
13217 David Sherwood <david.sherwood@arm.com>
13219 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13220 from an unsigned int to a poly_uint64.
13221 (_loop_vec_info::slp_unrolling_factor): Likewise.
13222 (_loop_vec_info::vectorization_factor): Change from an int
13224 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13225 (vect_get_num_vectors): New function.
13226 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13227 (vect_get_num_copies): Use vect_get_num_vectors.
13228 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13229 to an unsigned int *.
13230 (vect_analyze_data_refs): Change min_vf from an int * to a
13232 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13233 than an unsigned HOST_WIDE_INT.
13234 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13235 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13236 to an unsigned int *.
13237 (vect_analyze_data_ref_dependences): Likewise.
13238 (vect_compute_data_ref_alignment): Handle polynomial vf.
13239 (vect_enhance_data_refs_alignment): Likewise.
13240 (vect_prune_runtime_alias_test_list): Likewise.
13241 (vect_shift_permute_load_chain): Likewise.
13242 (vect_supportable_dr_alignment): Likewise.
13243 (dependence_distance_ge_vf): Take the vectorization factor as a
13244 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13245 (vect_analyze_data_refs): Change min_vf from an int * to a
13247 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13248 vfm1 as a poly_uint64 rather than an int. Make the same change
13249 for the returned bound_scalar.
13250 (vect_gen_vector_loop_niters): Handle polynomial vf.
13251 (vect_do_peeling): Likewise. Update call to
13252 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13253 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13255 * tree-vect-loop.c (vect_determine_vectorization_factor)
13256 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13257 (vect_get_known_peeling_cost): Likewise.
13258 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13259 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13260 (vect_transform_loop): Likewise. Use the lowest possible VF when
13261 updating the upper bounds of the loop.
13262 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13263 rather than an int.
13264 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13265 polynomial unroll factors.
13266 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13267 (vect_make_slp_decision): Likewise.
13268 (vect_supported_load_permutation_p): Likewise, and polynomial
13270 (vect_analyze_slp_cost): Handle polynomial vf.
13271 (vect_slp_analyze_node_operations): Likewise.
13272 (vect_slp_analyze_bb_1): Likewise.
13273 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13274 than an unsigned HOST_WIDE_INT.
13275 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13276 (vectorizable_load): Handle polynomial vf.
13277 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13279 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13282 Alan Hayward <alan.hayward@arm.com>
13283 David Sherwood <david.sherwood@arm.com>
13285 * match.pd: Handle bit operations involving three constants
13286 and try to fold one pair.
13288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13290 * tree-vect-loop-manip.c: Include gimple-fold.h.
13291 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13292 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13293 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13294 Add a path that uses a step of VF instead of 1, but disable it
13296 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13297 and niters_no_overflow parameters. Update calls to
13298 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13299 Create a new SSA name if the latter choses to use a ste other
13300 than zero, and return it via niters_vector_mult_vf_var.
13301 * tree-vect-loop.c (vect_transform_loop): Update calls to
13302 vect_do_peeling, vect_gen_vector_loop_niters and
13303 slpeel_make_loop_iterate_ntimes.
13304 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13305 (vect_gen_vector_loop_niters): Update declarations after above changes.
13307 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13309 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13310 128-bit round to integer instructions.
13311 (ceil<mode>2): Likewise.
13312 (btrunc<mode>2): Likewise.
13313 (round<mode>2): Likewise.
13315 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13317 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13318 unaligned VSX load/store on P8/P9.
13319 (expand_block_clear): Allow the use of unaligned VSX
13320 load/store on P8/P9.
13322 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13324 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13326 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13327 swap associated with both a load and a store.
13329 2018-01-02 Andrew Waterman <andrew@sifive.com>
13331 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13332 * config/riscv/riscv.md (clear_cache): Use it.
13334 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13336 * web.c: Remove out-of-date comment.
13338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13340 * expr.c (fixup_args_size_notes): Check that any existing
13341 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13342 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13343 (emit_single_push_insn): ...here.
13345 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13347 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13348 (const_vector_encoded_nelts): New function.
13349 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13350 (const_vector_int_elt, const_vector_elt): Declare.
13351 * emit-rtl.c (const_vector_int_elt_1): New function.
13352 (const_vector_elt): Likewise.
13353 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13354 of CONST_VECTOR_ELT.
13356 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13358 * expr.c: Include rtx-vector-builder.h.
13359 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13360 directly on the tree encoding.
13361 (const_vector_from_tree): Likewise.
13362 * optabs.c: Include rtx-vector-builder.h.
13363 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13364 sequence of "u" values.
13365 * vec-perm-indices.c: Include rtx-vector-builder.h.
13366 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13367 directly on the vec_perm_indices encoding.
13369 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13371 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13372 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13373 * rtx-vector-builder.h: New file.
13374 * rtx-vector-builder.c: Likewise.
13375 * rtl.h (rtx_def::u2): Add a const_vector field.
13376 (CONST_VECTOR_NPATTERNS): New macro.
13377 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13378 (CONST_VECTOR_DUPLICATE_P): Likewise.
13379 (CONST_VECTOR_STEPPED_P): Likewise.
13380 (CONST_VECTOR_ENCODED_ELT): Likewise.
13381 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13382 (unwrap_const_vec_duplicate): Likewise.
13383 (const_vec_series_p): Check for a non-duplicated vector encoding.
13384 Say that the function only returns true for integer vectors.
13385 * emit-rtl.c: Include rtx-vector-builder.h.
13386 (gen_const_vec_duplicate_1): Delete.
13387 (gen_const_vector): Call gen_const_vec_duplicate instead of
13388 gen_const_vec_duplicate_1.
13389 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13390 (gen_const_vec_duplicate): Use rtx_vector_builder.
13391 (gen_const_vec_series): Likewise.
13392 (gen_rtx_CONST_VECTOR): Likewise.
13393 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13394 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13395 Build a new vector rather than modifying a CONST_VECTOR in-place.
13396 (handle_special_swappables): Update call accordingly.
13397 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13398 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13399 Build a new vector rather than modifying a CONST_VECTOR in-place.
13400 (handle_special_swappables): Update call accordingly.
13402 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13404 * simplify-rtx.c (simplify_const_binary_operation): Use
13405 CONST_VECTOR_ELT instead of XVECEXP.
13407 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13409 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13410 the selector elements to be different from the data elements
13411 if the selector is a VECTOR_CST.
13412 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13413 ssizetype for the selector.
13415 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13417 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13418 before testing each element individually.
13419 * tree-vect-generic.c (lower_vec_perm): Likewise.
13421 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13423 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13424 * selftest-run-tests.c (selftest::run_tests): Call it.
13425 * vector-builder.h (vector_builder::operator ==): New function.
13426 (vector_builder::operator !=): Likewise.
13427 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13428 (vec_perm_indices::all_from_input_p): New function.
13429 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13430 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13431 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13432 instead of reading the VECTOR_CST directly. Detect whether both
13433 vector inputs are the same before constructing the vec_perm_indices,
13434 and update the number of inputs argument accordingly. Use the
13435 utility functions added above. Only construct sel2 if we need to.
13437 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13439 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13440 the broadcast of the low byte.
13441 (expand_mult_highpart): Use an explicit encoding for the permutes.
13442 * optabs-query.c (can_mult_highpart_p): Likewise.
13443 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13444 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13445 (vectorizable_bswap): Likewise.
13446 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13447 explicit encoding for the power-of-2 permutes.
13448 (vect_permute_store_chain): Likewise.
13449 (vect_grouped_load_supported): Likewise.
13450 (vect_permute_load_chain): Likewise.
13452 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13454 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13455 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13456 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13457 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13458 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13459 (vect_gen_perm_mask_any): Likewise.
13461 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13463 * int-vector-builder.h: New file.
13464 * vec-perm-indices.h: Include int-vector-builder.h.
13465 (vec_perm_indices): Redefine as an int_vector_builder.
13466 (auto_vec_perm_indices): Delete.
13467 (vec_perm_builder): Redefine as a stand-alone class.
13468 (vec_perm_indices::vec_perm_indices): New function.
13469 (vec_perm_indices::clamp): Likewise.
13470 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13471 (vec_perm_indices::new_vector): New function.
13472 (vec_perm_indices::new_expanded_vector): Update for new
13473 vec_perm_indices class.
13474 (vec_perm_indices::rotate_inputs): New function.
13475 (vec_perm_indices::all_in_range_p): Operate directly on the
13476 encoded form, without computing elided elements.
13477 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13478 encoding. Update for new vec_perm_indices class.
13479 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13480 the given vec_perm_builder.
13481 (expand_vec_perm_var): Update vec_perm_builder constructor.
13482 (expand_mult_highpart): Use vec_perm_builder instead of
13483 auto_vec_perm_indices.
13484 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13485 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13486 or double series encoding as appropriate.
13487 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13488 vec_perm_indices instead of auto_vec_perm_indices.
13489 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13490 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13491 (vect_permute_store_chain): Likewise.
13492 (vect_grouped_load_supported): Likewise.
13493 (vect_permute_load_chain): Likewise.
13494 (vect_shift_permute_load_chain): Likewise.
13495 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13496 (vect_transform_slp_perm_load): Likewise.
13497 (vect_schedule_slp_instance): Likewise.
13498 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13499 (vectorizable_mask_load_store): Likewise.
13500 (vectorizable_bswap): Likewise.
13501 (vectorizable_store): Likewise.
13502 (vectorizable_load): Likewise.
13503 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13504 vec_perm_indices instead of auto_vec_perm_indices. Use
13505 tree_to_vec_perm_builder to read the vector from a tree.
13506 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13507 vec_perm_builder instead of a vec_perm_indices.
13508 (have_whole_vector_shift): Use vec_perm_builder and
13509 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13510 truncation to calc_vec_perm_mask_for_shift.
13511 (vect_create_epilog_for_reduction): Likewise.
13512 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13513 from auto_vec_perm_indices to vec_perm_indices.
13514 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13515 instead of changing individual elements.
13516 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13517 the vector in d.perm.
13518 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13519 from auto_vec_perm_indices to vec_perm_indices.
13520 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13521 instead of changing individual elements.
13522 (arm_vectorize_vec_perm_const): Use new_vector to install
13523 the vector in d.perm.
13524 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13525 Update vec_perm_builder constructor.
13526 (rs6000_expand_interleave): Likewise.
13527 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13528 (rs6000_expand_interleave): Likewise.
13530 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13532 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13533 to qimode could truncate the indices.
13534 * optabs.c (expand_vec_perm_var): Likewise.
13536 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13538 * Makefile.in (OBJS): Add vec-perm-indices.o.
13539 * vec-perm-indices.h: New file.
13540 * vec-perm-indices.c: Likewise.
13541 * target.h (vec_perm_indices): Replace with a forward class
13543 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13544 * optabs.h: Include vec-perm-indices.h.
13545 (expand_vec_perm): Delete.
13546 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13547 (expand_vec_perm_const): Declare.
13548 * target.def (vec_perm_const_ok): Replace with...
13549 (vec_perm_const): ...this new hook.
13550 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13551 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13552 * doc/tm.texi: Regenerate.
13553 * optabs.def (vec_perm_const): Delete.
13554 * doc/md.texi (vec_perm_const): Likewise.
13555 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13556 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13557 expand_vec_perm for constant permutation vectors. Assert that
13558 the mode of variable permutation vectors is the integer equivalent
13559 of the mode that is being permuted.
13560 * optabs-query.h (selector_fits_mode_p): Declare.
13561 * optabs-query.c: Include vec-perm-indices.h.
13562 (selector_fits_mode_p): New function.
13563 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13564 is defined, instead of checking whether the vec_perm_const_optab
13565 exists. Use targetm.vectorize.vec_perm_const instead of
13566 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13567 fit in the vector mode before using a variable permute.
13568 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13569 vec_perm_indices instead of an rtx.
13570 (expand_vec_perm): Replace with...
13571 (expand_vec_perm_const): ...this new function. Take the selector
13572 as a vec_perm_indices rather than an rtx. Also take the mode of
13573 the selector. Update call to shift_amt_for_vec_perm_mask.
13574 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13575 Use vec_perm_indices::new_expanded_vector to expand the original
13576 selector into bytes. Check whether the indices fit in the vector
13577 mode before using a variable permute.
13578 (expand_vec_perm_var): Make global.
13579 (expand_mult_highpart): Use expand_vec_perm_const.
13580 * fold-const.c: Includes vec-perm-indices.h.
13581 * tree-ssa-forwprop.c: Likewise.
13582 * tree-vect-data-refs.c: Likewise.
13583 * tree-vect-generic.c: Likewise.
13584 * tree-vect-loop.c: Likewise.
13585 * tree-vect-slp.c: Likewise.
13586 * tree-vect-stmts.c: Likewise.
13587 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13589 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13590 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13591 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13592 (aarch64_vectorize_vec_perm_const): ...this new function.
13593 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13594 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13595 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13596 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13597 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13598 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13599 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13601 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13602 check for NEON modes.
13603 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13604 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13605 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13606 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13608 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13609 the old VEC_PERM_CONST conditions.
13610 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13611 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13612 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13613 (ia64_vectorize_vec_perm_const_ok): Merge into...
13614 (ia64_vectorize_vec_perm_const): ...this new function.
13615 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13616 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13617 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13618 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13619 * config/mips/mips.c (mips_expand_vec_perm_const)
13620 (mips_vectorize_vec_perm_const_ok): Merge into...
13621 (mips_vectorize_vec_perm_const): ...this new function.
13622 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13623 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13624 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13625 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13626 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13627 (rs6000_expand_vec_perm_const): Delete.
13628 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13630 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13631 (altivec_expand_vec_perm_const_le): Take each operand individually.
13632 Operate on constant selectors rather than rtxes.
13633 (altivec_expand_vec_perm_const): Likewise. Update call to
13634 altivec_expand_vec_perm_const_le.
13635 (rs6000_expand_vec_perm_const): Delete.
13636 (rs6000_vectorize_vec_perm_const_ok): Delete.
13637 (rs6000_vectorize_vec_perm_const): New function.
13638 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13639 an element count and rtx array.
13640 (rs6000_expand_extract_even): Update call accordingly.
13641 (rs6000_expand_interleave): Likewise.
13642 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13643 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13644 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13645 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13646 (rs6000_expand_vec_perm_const): Delete.
13647 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13648 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13649 (altivec_expand_vec_perm_const_le): Take each operand individually.
13650 Operate on constant selectors rather than rtxes.
13651 (altivec_expand_vec_perm_const): Likewise. Update call to
13652 altivec_expand_vec_perm_const_le.
13653 (rs6000_expand_vec_perm_const): Delete.
13654 (rs6000_vectorize_vec_perm_const_ok): Delete.
13655 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13656 reference to the SPE evmerge intructions.
13657 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13658 an element count and rtx array.
13659 (rs6000_expand_extract_even): Update call accordingly.
13660 (rs6000_expand_interleave): Likewise.
13661 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13662 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13664 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13666 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13668 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13669 vector mode and that that mode matches the mode of the data
13671 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13672 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13673 directly using expand_vec_perm_1 when forcing selectors into
13675 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13677 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13679 * optabs-query.h (can_vec_perm_p): Delete.
13680 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13681 * optabs-query.c (can_vec_perm_p): Split into...
13682 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13683 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13684 particular selector is valid.
13685 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13686 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13687 (vect_grouped_load_supported): Likewise.
13688 (vect_shift_permute_load_chain): Likewise.
13689 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13690 (vect_transform_slp_perm_load): Likewise.
13691 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13692 (vectorizable_bswap): Likewise.
13693 (vect_gen_perm_mask_checked): Likewise.
13694 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13695 implementations of variable permutation vectors into account
13696 when deciding which selector to use.
13697 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13698 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13699 with a false third argument.
13700 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13701 to test whether the constant selector is valid and can_vec_perm_var_p
13702 to test whether a variable selector is valid.
13704 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13706 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13707 * optabs-query.c (can_vec_perm_p): Likewise.
13708 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13709 instead of vec_perm_indices.
13710 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13711 (vect_gen_perm_mask_checked): Likewise,
13712 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13713 (vect_gen_perm_mask_checked): Likewise,
13715 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13717 * optabs-query.h (qimode_for_vec_perm): Declare.
13718 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13719 (qimode_for_vec_perm): ...this new function.
13720 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13722 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13724 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13725 does not have a conditional at the top.
13727 2018-01-02 Richard Biener <rguenther@suse.de>
13729 * ipa-inline.c (big_speedup_p): Fix expression.
13731 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13734 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13737 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13741 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13742 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13743 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13744 cond_taken_branch_cost 3->4.
13746 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13748 PR tree-optimization/83581
13749 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13750 TODO_cleanup_cfg if any changes have been made.
13752 PR middle-end/83608
13753 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13754 convert_modes if target mode has the right side, but different mode
13757 PR middle-end/83609
13758 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13759 last argument when extracting from CONCAT. If either from_real or
13760 from_imag is NULL, use expansion through memory. If result is not
13761 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13762 the parts directly to inner mode, if even that fails, use expansion
13765 PR middle-end/83623
13766 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13767 check for bswap in mode rather than HImode and use that in expand_unop
13770 Copyright (C) 2018 Free Software Foundation, Inc.
13772 Copying and distribution of this file, with or without modification,
13773 are permitted in any medium without royalty provided the copyright
13774 notice and this notice are preserved.