1 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6 * config/aarch64/iterators.md (ABAL): New int iterator.
9 (sur): Add mappings for the above.
10 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12 (aarch64_<sur>abal<mode>_4): Likewise.
13 (aarch64_<sur>adalp<mode>_3): Likewise.
14 (<sur>sadv16qi): New define_expand.
16 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
18 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
19 (*movdf_internal): Ditto.
21 (*rsqrtsf2_sse): Ditto.
22 (*sqrt<mode>2_sse): Ditto.
24 2018-05-21 Tamar Christina <tamar.christina@arm.com>
26 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
28 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
29 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
30 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
31 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
33 * config/aarch64/arm_neon.h: Likewise.
34 * config/aarch64/iterators.md (VQ_I): New.
36 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
38 * config.gcc: Add arc/t-multilib-linux to tmake_file for
40 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
43 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
45 * config/nds32/constraints.md (S): New constraint.
46 * config/nds32/nds32.md (call_internal): Use constraint S.
47 (call_value_internal): Likewise.
48 (sibcall_internal): Likewise.
49 (sibcall_value_internal): Likewise.
51 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
52 Chung-Ju Wu <jasonwucj@gmail.com>
54 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
57 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
58 Chung-Ju Wu <jasonwucj@gmail.com>
60 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
61 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
62 (nds32_rtx_costs_impl): Simplify.
63 (nds32_address_cost_impl): Simplify.
64 (nds32_init_rtx_costs): New function.
65 (nds32_rtx_costs_speed_prefer): Likewise.
66 (nds32_rtx_costs_size_prefer): Likewise.
67 (nds32_address_cost_speed_prefer): Likewise.
68 (nds32_address_cost_speed_fwprop): Likewise.
69 (nds32_address_cost_size_prefer): Likewise.
70 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
71 * config/nds32/nds32.c (nds32_option_override): Use
72 nds32_init_rtx_costs function.
74 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
76 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
77 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
78 (TARGET_PIPELINE_N8): Likewise.
79 (TARGET_PIPELINE_N10): Likewise.
80 (TARGET_PIPELINE_N13): Likewise.
81 (TARGET_PIPELINE_GRAYWOLF): Likewise.
83 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
85 * config/nds32/nds32-fpu.md: Update copyright year.
87 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
89 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
91 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
93 * config/nds32/nds32.c
94 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
95 * config/nds32/nds32.opt (minline-asm-r15): New option.
97 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
99 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
101 * config/nds32/nds32.md (abssi2): New pattern.
103 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
105 * config/i386/i386.md (rex64namesuffix): New mode attribute.
106 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
107 Merge insn pattern from sse_cvtsi2ss<round_name> and
108 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
109 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
110 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
111 using SWI48 mode iterator.
112 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
113 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
114 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
115 pattern from sse_cvttss2si<round_saeonly_name>
116 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
117 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
118 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
119 using SWI48 mode iterator.
120 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
121 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
122 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
123 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
124 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
125 using SWI48 mode iterator.
126 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
127 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
128 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
129 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
130 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
132 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
133 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
134 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
135 pattern from sse_cvttsd2si<round_saeonly_name>
136 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
138 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
140 * config/nds32/nds32-md-auxiliary.c
141 (nds32_valid_smw_lwm_base_p): Refine.
142 (nds32_output_smw_single_word): Refine.
143 (nds32_output_smw_double_word): New.
144 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
146 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
148 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
149 (nds32_output_stack_pop): Refine.
150 (nds32_expand_unaligned_load): Refine.
151 (nds32_expand_unaligned_store): Refine.
153 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
154 Chung-Ju Wu <jasonwucj@gmail.com>
156 * config/nds32/constants.md: Add TP_REGNUM constant.
157 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
158 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
160 * config/nds32/nds32-doubleword.md: Consider flag_pic.
161 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
162 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
163 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
164 and PIC code generation.
165 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
167 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
169 * config/nds32/nds32.md: Support TLS and PIC.
170 * config/nds32/nds32.c: Support TLS and PIC.
171 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
172 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
175 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
177 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
180 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
181 Chung-Ju Wu <jasonwucj@gmail.com>
183 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
184 * config/nds32/nds32-md-auxiliary.c
185 (symbolic_reference_mentioned_p): New.
186 (nds32_legitimize_ict_address): New.
187 (nds32_expand_ict_move): New.
188 (nds32_indirect_call_referenced_p): New.
189 (nds32_symbol_binds_local_p): Delete.
190 (nds32_long_call_p): Modify.
191 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
192 * config/nds32/nds32-protos.h
193 (symbolic_reference_mentioned_p): Declare.
194 (nds32_legitimize_ict_address): Declare.
195 (nds32_expand_ict_move): Declare.
196 (nds32_indirect_call_referenced_p): Declare.
197 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
198 (nds32_relax_group): Use nds32_ict_const_p as condition.
199 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
200 (nds32_asm_file_start): Output ict_model directive in asm code.
201 (nds32_legitimate_address_p): Consider indirect call.
202 (nds32_print_operand): Consider indirect call.
203 (nds32_print_operand_address): Consider indirect call.
204 (nds32_insert_attributes): Handle "indirect_call" attribute.
205 (TARGET_LEGITIMATE_ADDRESS_P): Define.
206 (TARGET_LEGITIMATE_CONSTANT_P): Define.
207 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
208 (TARGET_DELEGITIMIZE_ADDRESS): Define.
209 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
210 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
211 (TARGET_ICT_MODEL_SMALL): Define.
212 (TARGET_ICT_MODEL_LARGE): Define.
213 * config/nds32/nds32.md (movsi): Consider ict model.
214 (call, call_value): Consider ict model.
215 (sibcall, sibcall_value): Consider ict model.
216 * config/nds32/nds32.opt (mict-model): New option.
217 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
220 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
221 Monk Chiang <sh.chiang04@gmail.com>
222 Jim Wilson <jimw@sifive.com>
224 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
225 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
226 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
227 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
228 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
229 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
230 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
231 compute save_libcall_adjustment properly.
232 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
233 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
234 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
235 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
236 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
237 (ABI_SPEC): Handle mabi=ilp32e.
238 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
240 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
241 <-march>: Add rv32e as an example.
243 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
246 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
247 (intra_create_variable_infos): Handle C++ constructors.
249 2018-05-18 Martin Liska <mliska@suse.cz>
251 * passes.def: Remove a redundant pass.
253 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
256 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
258 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
260 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
262 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
264 (armv2, armv2a, armv3, armv3m): Delete architectures.
265 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
266 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
267 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
269 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
270 (*mulsidi3adddi): Likewise.
271 (mulsidi3): Likewise.
272 (*mulsidi3_nov6): Likewise.
273 (umulsidi3): Likewise.
274 (umulsidi3_nov6): Likewise.
275 (umaddsidi4): Likewise.
276 (*umulsidi3adddi): Likewise.
277 (smulsi3_highpart): Likewise.
278 (*smulsi3_highpart_nov6): Likewise.
279 (umulsi3_highpart): Likewise.
280 (*umulsi3_highpart_nov6): Likewise.
281 * config/arm/arm.h (arm_arch3m): Delete.
282 * config/arm/arm.c (arm_arch3m): Delete.
283 (arm_option_override_internal): Update armv3-related comment.
284 (arm_configure_build_target): Delete use of isa_bit_mode32.
285 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
286 (arm_rtx_costs_internal): Delete check of arm_arch3m.
287 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
290 * config/arm/arm-protos.h (arm_arch3m): Delete.
291 * config/arm/arm-tables.opt: Regenerate.
292 * config/arm/arm-tune.md: Likewise.
293 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
294 deleted architectures.
296 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
298 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
299 (armv5t, armv5te): New features.
300 (ARMv5, ARMv5e): Delete fgroups.
301 (ARMv5t, ARMv5te): Adjust for above changes.
303 (armv5, armv5e): Delete arches.
304 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
306 (*call_reg_arm): Likewise.
307 (*call_value_reg_armv5): Likewise.
308 (*call_value_reg_arm): Likewise.
309 (*call_symbol): Likewise.
310 (*call_value_symbol): Likewise.
311 (*sibcall_insn): Likewise.
312 (*sibcall_value_insn): Likewise.
314 (prefetch): Likewise.
315 (define_split and define_peephole2 dependent on arm_arch5):
317 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
319 (TARGET_ARM_QBIT): Likewise.
320 (TARGET_DSP_MULTIPLY): Likewise.
321 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
322 (arm_arch5, arm_arch5e): Delete.
323 (arm_arch5t, arm_arch5te): Declare.
324 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
325 (arm_arch5t): Declare.
326 (arm_option_reconfigure_globals): Update for the above.
327 (arm_options_perform_arch_sanity_checks): Update comment, replace
328 use of arm_arch5 with arm_arch5t.
329 (use_return_insn): Likewise.
330 (arm_emit_call_insn): Likewise.
331 (output_return_instruction): Likewise.
332 (arm_final_prescan_insn): Likewise.
333 (arm_coproc_builtin_available): Likewise.
334 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
335 arm_arch5e with arm_arch5t and arm_arch5te.
336 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
337 (arm_arch5t, arm_arch5te): Declare.
338 * config/arm/arm-tables.opt: Regenerate.
339 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
340 * config/arm/t-multilib: Likewise.
341 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
342 instead of arm_arch5.
343 (*call_reg_thumb1): Likewise.
344 (*call_value_reg_thumb1_v5): Likewise.
345 (*call_value_reg_thumb1): Likewise.
346 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
348 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
350 2018-05-18 Martin Liska <mliska@suse.cz>
352 PR gcov-profile/84846
353 * doc/gcov.texi: Document -t option of gcov tool.
355 2018-05-18 Martin Liska <mliska@suse.cz>
357 PR gcov-profile/84846
358 * gcov.c (print_usage): Add new -t option.
359 (process_args): Handle the option.
360 (generate_results): Use stdout as output when requested by
363 2018-05-18 Martin Liska <mliska@suse.cz>
365 PR gcov-profile/84846
366 * coverage.c (coverage_init): Write PWD to .gcno file.
367 * doc/gcov.texi: Document how working directory is printed.
368 * gcov-dump.c (dump_gcov_file): Print PWD.
369 * gcov.c (output_intermediate_file): Likewise.
370 (read_graph_file): Read PWD string.
371 (output_lines): Print PWD.
373 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
376 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
377 for retval and return false if all args to phi are zero.
379 2018-05-18 Richard Biener <rguenther@suse.de>
381 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
383 (evrp_dom_walker::before_dom_children): Call it.
385 2018-05-18 Richard Biener <rguenther@suse.de>
387 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
388 results when processing array refs with variable index.
390 2018-05-18 Toon Moene <toon@moene.org>
392 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
393 directly after that of -floop-interchange. Indicate that both
394 options are enabled by default when specifying -O3.
396 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
398 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
399 iterator. Delete separate integer-mode vec_set<mode> expander.
400 (aarch64_simd_vec_setv2di): Delete.
401 (vec_setv2di): Delete.
402 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
403 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
404 the "w, r" alternative.
406 2018-05-18 Martin Liska <mliska@suse.cz>
408 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
409 * tree-pass.h (make_pass_lower_switch_O0): New function.
410 * tree-switch-conversion.c (node_has_low_bound): Remove.
411 (node_has_high_bound): Likewise.
412 (node_is_bounded): Likewise.
413 (class pass_lower_switch): Make it a template type and create
415 (pass_lower_switch::execute): Add template argument.
416 (make_pass_lower_switch): New function.
417 (make_pass_lower_switch_O0): New function.
418 (do_jump_if_equal): Remove.
419 (emit_case_nodes): Simplify to just handle all 3 cases and leave
420 all the hard work to tree optimization passes.
422 2018-05-18 Martin Liska <mliska@suse.cz>
424 * dbgcnt.c (limit_low): Renamed from limit.
425 (limit_high): New variable.
426 (dbg_cnt_is_enabled): Check for upper limit.
427 (dbg_cnt): Adjust dumping.
428 (dbg_cnt_set_limit_by_index): Add new argument for high
430 (dbg_cnt_set_limit_by_name): Likewise.
431 (dbg_cnt_process_single_pair): Parse new format.
432 (dbg_cnt_process_opt): Use strtok.
433 (dbg_cnt_list_all_counters): Remove 'value' and add
435 * doc/invoke.texi: Document changes.
437 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
439 * doc/sourcebuild.texi (scalar_all_fma): Document.
440 * tree.def (FMA_EXPR): Delete.
441 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
442 * internal-fn.c (ternary_direct): New macro.
443 (expand_ternary_optab_fn): Likewise.
444 (direct_ternary_optab_supported_p): Likewise.
445 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
446 * builtins.c (fold_builtin_fma): Delete.
447 (fold_builtin_3): Don't call it.
448 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
449 * expr.c (expand_expr_real_2): Likewise.
450 * fold-const.c (operand_equal_p): Likewise.
451 (fold_ternary_loc): Likewise.
452 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
453 * gimple.c (DEFTREECODE): Likewise.
454 * gimplify.c (gimplify_expr): Likewise.
455 * optabs-tree.c (optab_for_tree_code): Likewise.
456 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
457 * tree-eh.c (operation_could_trap_p): Likewise.
458 (stmt_could_throw_1_p): Likewise.
459 * tree-inline.c (estimate_operator_cost): Likewise.
460 * tree-pretty-print.c (dump_generic_node): Likewise.
461 (op_code_prio): Likewise.
462 * tree-ssa-loop-im.c (stmt_cost): Likewise.
463 * tree-ssa-operands.c (get_expr_operands): Likewise.
464 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
465 * fold-const-call.h (fold_fma): Delete.
466 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
467 CFN_FNMA and CFN_FNMS.
469 * genmatch.c (combined_fn): New enum.
470 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
471 (commutative_op): New function.
472 (commutate): Use it. Handle more than 2 operands.
473 (dt_operand::gen_gimple_expr): Use commutative_op.
474 (parser::parse_expr): Allow :c to be used with non-binary
475 operators if the commutative operand is known.
476 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
477 CFN_FMS, CFN_FNMA and CFN_FNMS.
478 (backprop::process_assign_use): Remove FMA_EXPR handling.
479 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
480 (gen_hsa_fma): New function.
481 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
482 IFN_FNMA and IFN_FNMS.
483 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
484 * gimple-fold.h (follow_all_ssa_edges): Declare.
485 * gimple-fold.c (follow_all_ssa_edges): New function.
486 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
487 gimple_build interface and use follow_all_ssa_edges to fold the result.
488 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
489 instead of checking for optabs directly.
490 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
491 rather than FMA_EXPRs.
492 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
493 call to IFN_FMA instead of an FMA_EXPR.
495 2018-05-17 Jim Wilson <jimw@sifive.com>
497 * expr.c (do_tablejump): When converting index to Pmode, if we have a
498 sign extended promoted subreg, and the range does not have the sign bit
499 set, then do a sign extend.
501 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
502 test, check for sign extended subreg and/or constant operands, and
503 do a sign extend in that case.
505 2018-05-17 Steve Ellcey <sellcey@cavium.com>
507 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
508 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
510 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
511 Change logics_shift_reg to logics_shift_imm.
512 (thunderx2t99_fp_loadpair_basic): Delete.
513 (thunderx2t99_fp_storepair_basic): Delete.
514 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
515 (thunderx2t99_asimd_polynomial): Delete.
516 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
517 and neon_fp_mul_d_scalar_q.
518 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
519 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
520 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
521 (thunderx2t99_asimd_lut): Add missing tbl types.
522 (thunderx2t99_asimd_ext): Delete.
523 (thunderx2t99_asimd_load1_1_mult): Delete.
524 (thunderx2t99_asimd_load1_2_mult): Delete.
525 (thunderx2t99_asimd_load1_ldp): New.
526 (thunderx2t99_asimd_load1): New.
527 (thunderx2t99_asimd_load2): Add missing *load2* types.
528 (thunderx2t99_asimd_load3): New.
529 (thunderx2t99_asimd_load4): New.
530 (thunderx2t99_asimd_store1_1_mult): Delete.
531 (thunderx2t99_asimd_store1_2_mult): Delete.
532 (thunderx2t99_asimd_store2_mult): Delete.
533 (thunderx2t99_asimd_store2_onelane): Delete.
534 (thunderx2t99_asimd_store_stp): New.
535 (thunderx2t99_asimd_store1): New.
536 (thunderx2t99_asimd_store2): New.
537 (thunderx2t99_asimd_store3): New.
538 (thunderx2t99_asimd_store4): New.
540 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
542 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
543 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
545 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
546 Segher Boessenkool <segher@kernel.crashing.org>
549 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
552 2018-05-17 Richard Biener <rguenther@suse.de>
554 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
555 for pruning loop and prune defs feeding only already visited PHIs.
557 2018-05-17 Richard Biener <rguenther@suse.de>
559 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
561 2018-05-17 Bin Cheng <bin.cheng@arm.com>
562 Richard Biener <rguenther@suse.de>
564 PR tree-optimization/85793
565 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
566 for VMAT_ELEMENTWISE.
568 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
570 * internal-fn.h (lookup_internal_fn): Declare
571 * internal-fn.c (lookup_internal_fn): New function.
572 * gimple.c (gimple_build_call_from_tree): Handle calls to
574 * gimple-pretty-print.c (dump_gimple_call): Print "." before
575 internal function names.
576 * tree-pretty-print.c (dump_generic_node): Likewise.
577 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
579 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
581 * gimple-fold.h (gimple_build): Make the function forms take
582 combined_fn rather than built_in_function.
583 (gimple_simplify): Likewise.
584 * gimple-match-head.c (gimple_simplify): Likewise.
585 * gimple-fold.c (gimple_build): Likewise.
586 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
587 rather than gimple_build_call_internal.
588 (get_initial_defs_for_reduction): Likewise.
589 (vect_create_epilog_for_reduction): Likewise.
590 (vectorizable_live_operation): Likewise.
592 2018-05-17 Martin Liska <mliska@suse.cz>
594 * gimple-ssa-sprintf.c (format_directive): Do not use
595 space in between 'G_' and '('.
597 2018-05-17 Jakub Jelinek <jakub@redhat.com>
600 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
601 even if the mask is not all ones.
604 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
606 (ix86_gimple_fold_builtin): Likewise.
609 * config/i386/i386.c: Include tree-vector-builder.h.
610 (ix86_vector_shift_count): New function.
611 (ix86_fold_builtin): Fold shift builtins by scalar count.
612 (ix86_gimple_fold_builtin): Likewise.
614 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
615 _mm512_setzero): New intrinsics.
617 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
618 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
620 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
621 code generation for cases where splatting a value is not useful.
622 * simplify-rtx.c (simplify_ternary_operation): Simplify
623 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
624 mode to a vec_concat.
626 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
628 * config.gcc: Support "goldmont-plus".
629 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
631 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
632 PROCESSOR_GOLDMONT_PLUS.
633 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
634 (processor_target_table): Add "goldmont-plus".
635 (PTA_GOLDMONT_PLUS): Define.
636 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
637 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
638 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
639 (fold_builtin_cpu): Add "goldmont-plus".
640 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
641 (ix86_option_override_internal): Add "goldmont-plus".
642 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
643 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
644 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
645 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
647 2018-05-17 Richard Biener <rguenther@suse.de>
649 PR tree-optimization/85757
650 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
651 remove defs that only feed that PHI from further processing.
653 2018-05-16 Jim Wilson <jimw@sifive.com>
655 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
657 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
659 2018-05-16 Mark Wielaard <mark@klomp.org>
661 * dwarf2out.c (count_index_strings): New function.
662 (output_indirect_strings): Call count_index_strings and generate
663 header for dwarf_version >= 5.
665 2018-05-16 Mark Wielaard <mark@klomp.org>
667 * dwarf2out.c (dwarf_FORM): New function.
668 (set_indirect_string): Use dwarf_FORM.
669 (reset_indirect_string): Likewise.
670 (size_of_die): Likewise.
671 (value_format): Likewise.
672 (output_die): Likewise.
673 (add_skeleton_AT_string): Likewise.
674 (output_macinfo_op): Likewise.
675 (index_string): Likewise.
676 (output_index_string_offset): Likewise.
677 (output_index_string): Likewise.
678 (count_index_strings): Likewise.
680 2018-05-16 Carl Love <cel@us.ibm.com>
682 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
683 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
685 2018-05-16 Martin Jambor <mjambor@suse.cz>
687 * ipa-prop.c (ipa_free_all_edge_args): Remove.
688 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
690 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
692 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
693 (fnma<mode>4): Likewise.
694 (fms<mode>4): Likewise.
695 (fnms<mode>4): Likewise.
696 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
697 (aarch64_fnma<mode>4): Likewise.
698 (aarch64_fms<mode>4): Likewise.
699 (aarch64_fnms<mode>4): Likewise.
700 (aarch64_fnmadd<mode>4): Likewise.
702 2018-05-16 Jason Merrill <jason@redhat.com>
704 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
706 2018-05-16 Richard Biener <rguenther@suse.de>
708 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
709 (dump_stmt_cost): Declare.
710 (add_stmt_cost): Dump cost we add.
711 (add_stmt_costs): New function.
712 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
714 (vect_analyze_stmt): Adjust prototype.
715 (vectorizable_condition): Likewise.
716 (vectorizable_live_operation): Likewise.
717 (vectorizable_reduction): Likewise.
718 (vectorizable_induction): Likewise.
719 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
720 cost vector to pass to vectorizable_ and record afterwards.
721 (vect_model_reduction_cost): Take cost vector argument and adjust.
722 (vect_model_induction_cost): Likewise.
723 (vectorizable_reduction): Likewise.
724 (vectorizable_induction): Likewise.
725 (vectorizable_live_operation): Likewise.
726 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
727 SLP_TREE_NUMBER_OF_VEC_STMTS.
728 (vect_analyze_slp_cost_1): Remove.
729 (vect_analyze_slp_cost): Likewise.
730 (vect_slp_analyze_node_operations): Take visited args and
731 a target cost vector. Avoid processing already visited stmt sets.
732 (vect_slp_analyze_operations): Use a local cost vector to gather
733 costs and register those of non-discarded instances.
734 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
735 (vect_schedule_slp_instance): Remove copying of
736 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
738 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
739 adding cost. Record cost entry location.
740 (vect_prologue_cost_for_slp_op): Function to compute cost of
741 a constant or invariant generated for SLP vect in the prologue,
742 split out from vect_analyze_slp_cost_1.
743 (vect_model_simple_cost): Make static. Adjust for SLP costing.
744 (vect_model_promotion_demotion_cost): Likewise.
745 (vect_model_store_cost): Likewise, make static.
746 (vect_model_load_cost): Likewise.
747 (vectorizable_bswap): Add cost vector arg and adjust.
748 (vectorizable_call): Likewise.
749 (vectorizable_simd_clone_call): Likewise.
750 (vectorizable_conversion): Likewise.
751 (vectorizable_assignment): Likewise.
752 (vectorizable_shift): Likewise.
753 (vectorizable_operation): Likewise.
754 (vectorizable_store): Likewise.
755 (vectorizable_load): Likewise.
756 (vectorizable_condition): Likewise.
757 (vectorizable_comparison): Likewise.
758 (can_vectorize_live_stmts): Likewise.
759 (vect_analyze_stmt): Likewise.
760 (vect_transform_stmt): Adjust calls to vectorizable_*.
761 * tree-vectorizer.c: Include gimple-pretty-print.h.
762 (dump_stmt_cost): New function.
764 2018-05-16 Richard Biener <rguenther@suse.de>
766 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
767 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
768 * tree-ssa-dse.c: Include tree-ssa-loop.h.
769 (check_name): New callback.
770 (dse_classify_store): Track cycles via a visited bitmap of PHI
771 defs and simplify handling of in-loop and across loop dead stores
772 and properly fail for loop-variant refs. Handle byte-tracking with
773 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
776 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
778 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
779 (vect_get_mask_type_for_stmt): Likewise.
780 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
782 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
783 to determine the statement's vector type and the vector type that
784 should be used for calculating nunits. Deal with cases in which
785 the type has to be deferred.
786 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
787 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
788 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
789 (vect_determine_vf_for_stmt): New functions, split out from...
790 (vect_determine_vectorization_factor): ...here.
791 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
792 (vect_get_mask_type_for_stmt): New functions, split out from
793 vect_determine_vectorization_factor.
795 2018-05-16 Richard Biener <rguenther@suse.de>
797 * tree-cfg.c (verify_gimple_assign_ternary): Properly
798 verify the [VEC_]COND_EXPR embedded comparison.
800 2018-05-15 Martin Sebor <msebor@redhat.com>
802 PR tree-optimization/85753
803 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
804 RECORD_TYPE in addition to ARRAY_TYPE.
806 2018-05-15 Martin Sebor <msebor@redhat.com>
809 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
811 2018-05-15 Richard Biener <rguenther@suse.de>
813 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
814 add by_clobber_p one. Change algorithm to collect all defs
815 representing uses we need to walk and try reducing them to
816 a single one before failing.
817 (dse_dom_walker::dse_optimize_stmt): Adjust.
819 2018-05-13 Mark Wielaard <mark@klomp.org>
821 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
822 (size_of_loc_descr): Likewise.
823 (output_loc_operands): Likewise.
824 (output_loc_operands_raw): Likewise.
825 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
826 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
827 (hash_loc_operands): Likewise.
828 (compare_loc_operands): Likewise.
830 2018-05-14 Mark Wielaard <mark@klomp.org>
832 * dwarf2out.c (count_index_addrs): New function.
833 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
835 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
837 PR tree-optimization/83648
838 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
839 return value as malloc candidate.
841 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
844 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
845 as true in call to suggest_attribute.
847 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
849 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
852 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
854 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
856 (aarch64_crypto_aesd_fused): Likewise.
858 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
860 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
861 (movsi_aarch64): Likewise.
862 (load_pairsi): Likewise.
863 (load_pairdi): Likewise.
864 (store_pairsi): Likewise.
865 (store_pairdi): Likewise.
866 (load_pairsf): Likewise.
867 (load_pairdf): Likewise.
868 (store_pairsf): Likewise.
869 (store_pairdf): Likewise.
870 (zero_extend): Likewise.
871 (trunc): Swap alternatives.
872 (fcvt_target): Add '?' to prefer w over r.
874 2018-05-14 Jakub Jelinek <jakub@redhat.com>
877 * config/i386/i386.md: Disallow non-commutative arithmetics in
878 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
879 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
880 in the peephole2 before it.
882 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
884 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
885 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
886 (ix86_handle_option): Handle -mcldemote.
887 * config.gcc: New header.
888 * config/i386/cldemoteintrin.h: New file.
889 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
890 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
892 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
893 OPTION_MASK_ISA_CLDEMOTE.
894 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
895 (ix86_valid_target_attribute_inner_p): Ditto.
896 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
897 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
898 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
899 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
900 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
902 * config/i386/i386.opt: Add -mcldemote.
903 * config/i386/x86intrin.h: New header.
904 * doc/invoke.texi: Add -mcldemote.
906 2018-05-14 Richard Biener <rguenther@suse.de>
908 * doc/match-and-simplify.texi: Adjust :s documentation.
910 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
912 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
913 intended memcpy size.
914 (REORDER_45): Likewise.
916 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
919 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
920 * vec.c (qsort_chk): Use gcc_qsort.
921 * Makefile.in (OBJS-libcommon): Add sort.o.
922 (build/sort.o): New target. Use it...
923 (BUILD_RTL): ... here, and...
924 (build/gencfn-macros): ... here, and...
925 (build/genmatch): ... here.
927 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
928 Chung-Ju Wu <jasonwucj@gmail.com>
930 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
931 * config/nds32/nds32-graywolf.md: New file.
932 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
933 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
935 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
936 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
937 * config/nds32/nds32.md (pipeline_model): Add graywolf.
938 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
939 * config/nds32/pipelines.md: Include n15 settings.
941 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
942 Chung-Ju Wu <jasonwucj@gmail.com>
944 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
945 * config/nds32/nds32-n13.md: New file.
946 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
947 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
949 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
950 * config/nds32/nds32.md (pipeline_model): Add n13.
951 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
952 * config/nds32/pipelines.md: Include n13 settings.
954 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
955 Chung-Ju Wu <jasonwucj@gmail.com>
957 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
958 * config/nds32/nds32-n10.md: New file.
959 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
960 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
962 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
963 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
964 * config/nds32/nds32.md (pipeline_model): Add n10.
965 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
966 * config/nds32/pipelines.md: Include n10 settings.
968 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
969 Kito Cheng <kito.cheng@gmail.com>
970 Chung-Ju Wu <jasonwucj@gmail.com>
972 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
973 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
974 Add enum values for DSP extension instructions.
975 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
977 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
978 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
980 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
981 * config/nds32/nds32-dspext.md: New file for DSP implementation.
982 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
983 * config/nds32/nds32-intrinsic.md: Likewise.
984 * config/nds32/nds32_intrinsic.h: Likewise.
985 * config/nds32/nds32-md-auxiliary.c: Likewise.
986 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
987 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
988 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
989 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
990 * config/nds32/nds32-protos.h: New declarations for DSP extension.
991 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
992 TYPE_DMAC in switch statement.
993 * config/nds32/nds32.c: New checking and implementation for DSP
994 extension instructions.
995 * config/nds32/nds32.h: Likewise.
996 * config/nds32/nds32.md: Likewise.
997 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
998 * config/nds32/predicates.md: Implement new predicates for DSP
1001 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1003 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1004 Reformat alternatives and attributes so it is easier to identify
1005 which constraints/attributes go with which instruction.
1006 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1007 (mov<mode>_softfloat32, FMOVE64): Likewise.
1008 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1009 (mov<mode>_softfloat64, FMOVE64): Likewise.
1011 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1013 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1015 (Basic PowerPC Built-in Functions): The new name of the
1016 subsection previously known as "PowerPC Built-in Functions".
1017 (Basic PowerPC Built-in Functions Available on all Configurations):
1019 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1020 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1021 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1022 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1024 2018-05-11 Martin Jambor <mjambor@suse.cz>
1027 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1030 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1033 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1035 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1037 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1038 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1039 (ix86_handle_option): Handle -mwaitpkg.
1040 * config.gcc: New header.
1041 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1042 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1043 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1045 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1046 OPTION_MASK_ISA_WAITPKG.
1047 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1048 (ix86_option_override_internal): Add PTA_WAITPKG.
1049 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1050 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1051 IX86_BUILTIN_TPAUSE.
1052 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1053 __builtin_ia32_umwait and __builtin_ia32_tpause.
1054 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1055 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1056 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1057 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1058 UNSPECV_TPAUSE): New.
1059 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1060 * config/i386/i386.opt: Add -mwaitpkg.
1061 * config/i386/waitpkgintrin.h: New file.
1062 * config/i386/x86intrin.h: New header.
1063 * doc/invoke.texi: Add -mwaitpkg.
1065 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1068 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1070 (cortex-m0): Use armv6s-m isa.
1071 (cortex-m0plus): Likewise.
1072 (cortex-m1): Likewise.
1073 (cortex-m0.small-multiply): Likewise.
1074 (cortex-m0plus.small-multiply): Likewise.
1075 (cortex-m1.small-multiply): Likewise.
1077 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1078 Jakub Jelinek <jakub@redhat.com>
1080 PR tree-optimization/85692
1081 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1082 source permute as well.
1084 2018-05-11 Martin Liska <mliska@suse.cz>
1087 * doc/extend.texi: Document LLVM style format for no_sanitize
1090 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1092 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1093 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1094 (mode_supports_vsx_dform_quad): Likewise.
1095 (mode_supports_vmx_dform): Move these functions to be next to the
1096 other mode_supports functions.
1097 (mode_supports_dq_form): Likewise.
1098 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1099 mode_supports_dq_form.
1100 (reg_offset_addressing_ok_p): Likewise.
1101 (offsettable_ok_by_alignment): Likewise.
1102 (rs6000_legitimate_offset_address_p): Likewise.
1103 (legitimate_lo_sum_address_p): Likewise.
1104 (rs6000_legitimize_address): Likewise.
1105 (rs6000_legitimize_reload_address): Likewise.
1106 (rs6000_secondary_reload_inner): Likewise.
1107 (rs6000_preferred_reload_class): Likewise.
1108 (rs6000_output_move_128bit): Likewise.
1110 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1112 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1113 Generate SImode target register for null target.
1114 <case IX86_BUILTIN_XGETBV>: Ditto.
1115 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1116 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1118 2018-05-10 Carl Love <cel@us.ibm.com>
1120 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1121 dcbtt and dcbtstt if operands[2] is 0.
1123 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1126 * config/i386/sse.md (usadv64qi): New expander.
1128 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1130 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1131 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1132 -maltivec=be support.
1133 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1134 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1135 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1136 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1137 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1138 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1139 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1140 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1141 altivec_vsumsws): Adjust.
1142 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1143 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1144 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1146 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1147 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1148 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1149 (altivec_lve<VI_char>x): Delete expand.
1150 (*altivec_lve<VI_char>x_internal): Rename to...
1151 (altivec_lve<VI_char>x): ... this.
1152 (altivec_lvxl_<mode>): Delete expand.
1153 (*altivec_lvxl_<mode>_internal): Rename to ...
1154 (altivec_lvxl_<mode>): ... this.
1155 (altivec_stvxl_<mode>): Delete expand.
1156 (*altivec_stvxl_<mode>_internal): Rename to ...
1157 (altivec_stvxl_<mode>): ... this.
1158 (altivec_stve<VI_char>x): Delete expand.
1159 (*altivec_stve<VI_char>x_internal): Rename to ...
1160 (altivec_stve<VI_char>x): ... this.
1161 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1162 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1163 reduc_plus_scal_<mode>): Adjust.
1164 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1166 (rs6000_cpu_cpp_builtins): Adjust.
1167 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1168 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1169 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1170 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1171 -maltivec=be support.
1172 (rs6000_split_vec_extract_var): Adjust.
1173 (rs6000_split_v4si_init): Adjust.
1174 (swap_selector_for_mode): Delete.
1175 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1176 altivec_expand_stvex_be): Delete.
1177 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1178 -maltivec=be support.
1179 (rs6000_gimple_fold_builtin): Ditto.
1180 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1182 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1183 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1184 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1185 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1186 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1187 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1188 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1189 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1190 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1191 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1192 anonymous split): Adjust.
1193 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1194 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1196 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1198 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1199 when --with-gxx-include-dir is also specified.
1200 * configure: Regenerate.
1202 2018-05-09 Jim Wilson <jimw@sifive.com>
1205 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1206 * config/riscv/t-withmultilib: New.
1207 * config/riscv/withmultilib.h: New.
1208 * doc/install.texi: Document RISC-V --with-multilib-list support.
1210 2018-05-09 Richard Biener <rguenther@suse.de>
1212 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1214 (vect_bb_vectorization_profitable_p): Adjust. Compute
1215 actual scalar cost using the cost vector and the add_stmt_cost
1218 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1220 PR rtl-optimization/85645
1221 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1222 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1224 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1226 PR rtl-optimization/85645
1227 * shrink-wrap.c (spread_components): Return a boolean saying if
1228 anything was changed.
1229 (try_shrink_wrapping_separate): Iterate spread_components until
1230 nothing changes anymore.
1232 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1234 PR rtl-optimization/85645
1235 * regrename.c (build_def_use): Also kill the chains that include the
1236 destination of a REG_CFA_REGISTER note.
1238 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1240 PR rtl-optimization/85645
1241 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1242 insn that has a REG_CFA_REGISTER note.
1244 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1246 * cfgexpand.c (expand_clobber): New function.
1247 (expand_gimple_stmt_1): Use it.
1248 * tree-vect-stmts.c (vect_clobber_variable): New function,
1250 (vectorizable_simd_clone_call): ...here.
1251 (vectorizable_store): Emit a clobber either side of an
1252 IFN_STORE_LANES sequence.
1253 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1255 2018-05-09 Tom de Vries <tom@codesourcery.com>
1258 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1259 (define_insn "trap_if_false"): Add exit after trap.
1261 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1263 PR rtl-optimization/85638
1264 * bb-reorder.c: Include common/common-target.h.
1265 (create_forwarder_block): New function extracted from...
1266 (fix_up_crossing_landing_pad): ...here. Rename into...
1267 (dw2_fix_up_crossing_landing_pad): ...this.
1268 (sjlj_fix_up_crossing_landing_pad): New function.
1269 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1270 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1271 from both partitions and exit the loop after one iteration.
1273 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1276 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1278 (Basic PowerPC Built-in Functions): The new name of the
1279 subsection previously known as "PowerPC Built-in Functions".
1280 (Basic PowerPC Built-in Functions Available on all Configurations):
1282 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1284 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1286 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1288 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1291 2018-05-08 Jim Wilson <jimw@sifive.com>
1293 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1294 (LD_EMUL_SUFFIX): New.
1295 (LINK_SPEC): Use it.
1297 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1299 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1301 (Basic PowerPC Built-in Functions): The new name of the
1302 subsection previously known as "PowerPC Built-in Functions".
1303 (Basic PowerPC Built-in Functions Available on all Configurations):
1305 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1307 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1309 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1311 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1314 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1317 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1318 after cmpelim optimization.
1320 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1322 * config.gcc: Support "goldmont".
1323 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1324 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1326 * config/i386/i386.c (m_GOLDMONT): Define.
1327 (processor_target_table): Add "goldmont".
1328 (PTA_GOLDMONT): Define.
1329 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1330 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1331 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1332 (fold_builtin_cpu): Add "goldmont".
1333 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1334 (ix86_option_override_internal): Add "goldmont".
1335 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1336 (processor_type): Add PROCESSOR_GOLDMONT.
1337 * config/i386/i386.md: Add CPU "glm".
1338 * config/i386/glm.md: New file.
1339 * config/i386/x86-tune.def: Add m_GOLDMONT.
1340 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1342 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1345 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1347 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1348 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1349 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1352 * config/i386/i386.c (ix86_fold_builtin): Handle
1353 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1356 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1357 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1359 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1362 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1363 (check_arch): Likewise.
1364 (check_fpu): Return the result rather than printing it.
1365 (end arch): Fix operator precedence.
1366 (end cpu): Likewise.
1367 (END): Print the result from check_fpu.
1369 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1370 Alan Hayward <alan.hayward@arm.com>
1371 David Sherwood <david.sherwood@arm.com>
1373 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1374 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1375 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1376 (*fcmuo<mode>_and): New patterns.
1378 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1380 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1381 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1382 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1383 (cmp_op, sve_imm_con): New code attributes.
1384 (SVE_COND_INT_CMP, imm_con): Delete.
1385 (cmp_op): Remove above unspecs from int attribute.
1386 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1388 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1389 comparison-specific unspecs.
1390 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1391 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1392 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1393 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1394 (*vec_fcm<cmp_op><mode>): Rename to...
1395 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1396 (*vec_fcmuo<mode>): Rename to...
1397 (*fcmuo<mode>): ...this and adjust likewise.
1398 (*pred_fcm<cmp_op><mode>): New pattern.
1399 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1400 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1402 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1404 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1405 (aarch64_emit_sve_predicated_cond): New function.
1406 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1407 (aarch64_emit_unspec_cond_or): Replace with...
1408 (aarch64_emit_sve_or_conds): ...this new function. Use
1409 aarch64_emit_sve_ptrue_op for the individual comparisons and
1410 aarch64_emit_binop to OR them together.
1411 (aarch64_emit_inverted_unspec_cond): Replace with...
1412 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1413 aarch64_emit_sve_ptrue_op for the comparison and
1414 aarch64_emit_unop to invert the result.
1415 (aarch64_expand_sve_vec_cmp_float): Update after the above
1416 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1418 2018-05-07 Nathan Sidwell <nathan@acm.org>
1420 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1421 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1422 (Backwards Compatibility): Likewise.
1424 2018-05-07 Luis Machado <luis.machado@linaro.org>
1428 2018-05-07 Luis Machado <luis.machado@linaro.org>
1430 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1431 <prefetch_dynamic_strides>: New const bool field.
1432 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1433 prefetch_dynamic_strides.
1434 (exynosm1_prefetch_tune): Likewise.
1435 (thunderxt88_prefetch_tune): Likewise.
1436 (thunderx_prefetch_tune): Likewise.
1437 (thunderx2t99_prefetch_tune): Likewise.
1438 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1439 (aarch64_override_options_internal): Update to set
1440 PARAM_PREFETCH_DYNAMIC_STRIDES.
1441 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1442 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1443 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1444 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1445 prefetch-dynamic-strides setting.
1447 2018-05-07 Luis Machado <luis.machado@linaro.org>
1449 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1450 <minimum_stride>: New const int field.
1451 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1452 minimum_stride field.
1453 (exynosm1_prefetch_tune): Likewise.
1454 (thunderxt88_prefetch_tune): Likewise.
1455 (thunderx_prefetch_tune): Likewise.
1456 (thunderx2t99_prefetch_tune): Likewise.
1457 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1458 (aarch64_override_options_internal): Update to set
1459 PARAM_PREFETCH_MINIMUM_STRIDE.
1460 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1461 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1462 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1463 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1464 stride is constant and is below the minimum stride threshold.
1466 2018-05-07 Luis Machado <luis.machado@linaro.org>
1468 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1471 2018-05-07 Luis Machado <luis.machado@linaro.org>
1473 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1474 <prefetch_dynamic_strides>: New const bool field.
1475 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1476 prefetch_dynamic_strides.
1477 (exynosm1_prefetch_tune): Likewise.
1478 (thunderxt88_prefetch_tune): Likewise.
1479 (thunderx_prefetch_tune): Likewise.
1480 (thunderx2t99_prefetch_tune): Likewise.
1481 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1482 (aarch64_override_options_internal): Update to set
1483 PARAM_PREFETCH_DYNAMIC_STRIDES.
1484 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1485 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1486 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1487 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1488 prefetch-dynamic-strides setting.
1490 2018-05-07 Luis Machado <luis.machado@linaro.org>
1492 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1493 <minimum_stride>: New const int field.
1494 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1495 minimum_stride field.
1496 (exynosm1_prefetch_tune): Likewise.
1497 (thunderxt88_prefetch_tune): Likewise.
1498 (thunderx_prefetch_tune): Likewise.
1499 (thunderx2t99_prefetch_tune): Likewise.
1500 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1501 (aarch64_override_options_internal): Update to set
1502 PARAM_PREFETCH_MINIMUM_STRIDE.
1503 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1504 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1505 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1506 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1507 stride is constant and is below the minimum stride threshold.
1509 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1512 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1513 the type is addressable. Don't force op into register if it has
1516 2018-05-05 Roland McGrath <mcgrathr@google.com>
1519 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1520 any section for which we don't know a specific type it should have,
1521 regardless of name. Previously this was done only for the exact
1522 names ".init_array", ".fini_array", and ".preinit_array".
1523 (default_elf_asm_named_section): Add comment about
1524 relationship with default_section_type_flags and SECTION_NOTYPE.
1525 (get_section): Don't consider it a type conflict if one side has
1526 SECTION_NOTYPE and the other doesn't, as long as neither has the
1527 SECTION_BSS et al used in the default_section_type_flags logic.
1529 2018-05-05 Tom de Vries <tom@codesourcery.com>
1532 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1533 (workaround_barsyncs): New function.
1534 (nvptx_reorg): Use workaround_barsyncs.
1535 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1536 (define_expand "nvptx_membar_cta"): New define_expand.
1537 (define_insn "*nvptx_membar_cta"): New insn.
1539 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1541 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1542 To improve optimization opportunities.
1543 * builtin-types.def: The new needed builtin types for the above.
1545 2018-05-04 Richard Biener <rguenther@suse.de>
1547 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1548 * gimple-ssa-store-merging.c
1549 (imm_store_chain_info::output_merged_store): Remove redundant create,
1550 release split_store vector contents on failure.
1551 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1552 scalar stmt vector on cache hit.
1554 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1556 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1558 * config.gcc (powerpc-xilinx-eabi*): Remove.
1559 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1561 (fusion_addis_mem_combo_load): Ditto.
1562 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1564 (rs6000_cpu_cpp_builtins): Ditto.
1565 * config/rs6000/rs6000-linux.c
1566 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1567 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1568 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1570 (rs6000_setup_reg_addr_masks): Ditto.
1571 (rs6000_init_hard_regno_mode_ok): Ditto.
1572 (rs6000_option_override_internal): Ditto.
1573 (legitimate_lo_sum_address_p): Ditto.
1574 (rs6000_legitimize_address): Ditto.
1575 (rs6000_legitimize_reload_address): Ditto.
1576 (rs6000_legitimate_address_p): Ditto.
1577 (abi_v4_pass_in_fpr): Ditto.
1578 (setup_incoming_varargs): Ditto.
1579 (rs6000_gimplify_va_arg): Ditto.
1580 (rs6000_split_multireg_move): Ditto.
1581 (rs6000_savres_strategy): Ditto.
1582 (rs6000_emit_prologue_components): Ditto.
1583 (rs6000_emit_epilogue_components): Ditto.
1584 (rs6000_emit_prologue): Ditto.
1585 (rs6000_emit_epilogue): Ditto.
1586 (rs6000_elf_file_end): Ditto.
1587 (rs6000_function_value): Ditto.
1588 (rs6000_libcall_value): Ditto.
1589 * config/rs6000/rs6000.h: Ditto.
1590 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1591 (TARGET_MINMAX): ... this. New.
1592 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1593 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1594 (*movsi_internal1_single): Delete.
1595 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1596 mfpu=, mxilinx-fpu): Delete.
1597 * config/rs6000/singlefp.h: Delete.
1598 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1599 * config/rs6000/t-rs6000: Ditto.
1600 * config/rs6000/t-xilinx: Delete.
1601 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1602 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1603 (VStype_simple): Delete.
1604 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1605 * config/rs6000/xfpu.h: Delete.
1606 * config/rs6000/xfpu.md: Delete.
1607 * config/rs6000/xilinx.h: Delete.
1608 * config/rs6000/xilinx.opt: Delete.
1609 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1610 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1612 2018-05-04 Tom de Vries <tom@codesourcery.com>
1615 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1618 2018-05-04 Richard Biener <rguenther@suse.de>
1621 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1622 (expand_complex_div_wide): Likewise.
1623 (expand_complex_operations_1): Likewise.
1624 (expand_complex_libcall): Preserve EH info of the original stmt.
1625 (tree_lower_complex): Handle removed blocks.
1626 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1627 on complex multiplication and division libcall builtins.
1629 2018-05-04 Richard Biener <rguenther@suse.de>
1632 * fold-const.c (negate_expr_p): Restrict negation of operand
1633 zero of a division to when we know that can happen without
1635 (fold_negate_expr_1): Likewise.
1637 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1640 * real.h (real_nextafter): Declare.
1641 * real.c (real_nextafter): New function.
1642 * fold-const-call.c (fold_const_nextafter): New function.
1643 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1644 CASE_CFN_NEXTTOWARD.
1645 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1646 even when arg1_mode is different from arg0_mode.
1648 2018-05-03 Nathan Sidwell <nathan@acm.org>
1650 * doc/extend.texi (Deprecated Features): Remove
1652 (Backwards Compatibility): Likewise.
1653 * doc/invoke.texi (C++ Language Options): Likewise.
1654 (C++ Dialect Options): Likewise.
1656 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1659 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1660 _mm512_mask_mullox_epi64): New intrinsics.
1662 2018-05-03 Tom de Vries <tom@codesourcery.com>
1665 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1666 dump files): Add offload-tree.
1668 2018-05-03 Richard Biener <rguenther@suse.de>
1670 PR tree-optimization/85615
1671 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1672 to loops not nested in BBs loop father to avoid creating multi-entry
1675 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1677 PR tree-optimization/70291
1678 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1679 arguments. Change return type to tree. Emit libcall as a new
1680 statement rather than replacing existing one when inplace_p is true.
1681 (expand_complex_multiplication_components): New function.
1682 (expand_complex_multiplication): Expand floating-point complex
1683 multiplication using the above.
1684 (expand_complex_division): Rename inner_type parameter to type.
1685 Update expand_complex_libcall call-site.
1686 (expand_complex_operations_1): Update expand_complex_multiplication
1687 and expand_complex_division call-sites.
1689 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1692 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1693 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1694 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1695 the highest significant bit of the shift count mask is clear. In
1696 check whether and[sq]i3 is needed verify that all significant bits
1697 of the shift count other than the highest are set.
1699 2018-05-02 Tom de Vries <tom@codesourcery.com>
1702 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1703 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1704 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1705 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1706 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1707 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1708 __builtin_goacc_parlevel_size.
1710 2018-05-02 Richard Biener <rguenther@suse.de>
1712 PR tree-optimization/85597
1713 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1714 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1717 2018-05-02 Tom de Vries <tom@codesourcery.com>
1720 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1721 dump files): Add ltrans-tree.
1723 2018-05-02 Tom de Vries <tom@codesourcery.com>
1726 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1727 dump files): Add wpa-ipa.
1729 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1731 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1732 powerpc*-*-linux*paired* target.
1733 * config/rs6000/750cl.h: Delete.
1734 * config/rs6000/paired.h: Delete.
1735 * config/rs6000/paired.md: Delete.
1736 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1738 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1739 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1740 comment. Remove paired float support.
1741 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1742 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1744 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1745 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1747 * config/rs6000/rs6000.c: Remove paired float support.
1748 (paired_expand_vector_init, paired_expand_vector_move,
1749 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1750 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1751 paired_expand_builtin, paired_expand_predicate_builtin,
1752 paired_init_builtins): Delete.
1753 * config/rs6000/rs6000.h: Remove paired float support.
1754 * config/rs6000/rs6000.md: Remove paired float support.
1755 (move_from_CR_ov_bit): Delete.
1756 * config/rs6000/rs6000.opt (mpaired): Delete.
1757 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1758 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1760 2018-05-02 Richard Biener <rguenther@suse.de>
1763 * gimplify.c (gimplify_save_expr): When in SSA form allow
1764 SAVE_EXPRs to compute to SSA vars.
1766 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1769 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1770 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1771 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1772 clobber operands[2], instead use a new pseudo. Formatting fixes.
1774 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1776 PR tree-optimization/85586
1777 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1778 exit early for statements in the same group if the accesses are
1781 2018-05-02 Tom de Vries <tom@codesourcery.com>
1784 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1787 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1789 PR tree-optimization/85143
1790 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1792 2018-05-01 Tom de Vries <tom@codesourcery.com>
1795 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1796 not found" error message.
1798 2018-05-01 Tom de Vries <tom@codesourcery.com>
1801 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1802 * vec.c (test_ordered_remove_if): New function.
1803 (vec_c_tests): Call test_ordered_remove_if.
1804 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1805 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1806 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1807 VEC_ORDERED_REMOVE_IF.
1809 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1811 PR tree-optimization/82665
1812 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1813 pointer subtraction where arguments come from a memchr call.
1815 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1817 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1818 --push-state --as-needed and --pop-state instead of --as-needed and
1819 --no-as-needed if ld supports it.
1820 * configure: Regenerated.
1823 * doc/install.texi2html: Replace _002d with - and _002a with * in
1824 generated html files using sed.
1826 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1829 * gcc-rich-location.c (blank_line_before_p): New function.
1830 (use_new_line): New function.
1831 (gcc_rich_location::add_fixit_insert_formatted): New function.
1832 * gcc-rich-location.h
1833 (gcc_rich_location::add_fixit_insert_formatted): New function.
1835 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1837 * selftest.c (assert_streq): Rename "expected" and "actual" to
1838 "val1" and "val2". Extend NULL-handling to cover both inputs
1839 symmetrically, while still requiring both to be non-NULL for a pass.
1840 * selftest.h (assert_streq): Rename "expected" and "actual" to
1842 (ASSERT_EQ): Likewise.
1843 (ASSERT_EQ_AT): Likewise.
1844 (ASSERT_KNOWN_EQ): Likewise.
1845 (ASSERT_KNOWN_EQ_AT): Likewise.
1846 (ASSERT_NE): Likewise.
1847 (ASSERT_MAYBE_NE): Likewise.
1848 (ASSERT_MAYBE_NE_AT): Likewise.
1849 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1850 the assertion to pass.
1851 (ASSERT_STREQ_AT): Likewise.
1853 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1855 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1856 interaction with -pie.
1858 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1860 * selftest.h: Fix alphabetization of per-source-file selftest
1863 2018-04-30 Jason Merrill <jason@redhat.com>
1865 PR c++/61982 - dead stores to destroyed objects.
1866 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1869 2018-04-30 Jason Merrill <jason@redhat.com>
1871 * tree.c (build_clobber): New.
1872 * tree.h: Declare it.
1873 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1875 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1877 * diagnostic-show-locus.c (layout::layout): Update for
1878 location_get_source_line returning a char_span.
1879 (struct char_span): Move to input.h.
1880 (struct correction): Update for fields in char_span becoming
1882 (struct source_line): Update for location_get_source_line
1883 returning a char_span.
1884 (layout::print_line): Likewise.
1885 * edit-context.c (edited_file::print_content): Likewise.
1886 (edited_file::print_diff_hunk): Likewise.
1887 (edited_file::print_run_of_changed_lines): Likewise.
1888 (edited_file::get_num_lines): Likewise.
1889 (edited_line::edited_line): Likewise.
1890 * final.c (asm_show_source): Likewise.
1891 * input.c (location_get_source_line): Convert return type
1892 from const char * to char_span, losing the final "line_len"
1894 (dump_location_info): Update for the above.
1895 (get_substring_ranges_for_loc): Likewise. Use a char_span
1896 when handling the literal within the line.
1897 (test_reading_source_line): Update for location_get_source_line
1898 returning a char_span.
1899 * input.h (class char_span): Move here from
1900 diagnostic-show-locus.c, converting from a struct to a class.
1901 Make data members private.
1902 (char_span::operator bool): New.
1903 (char_span::length): New.
1904 (char_span::get_buffer): New.
1905 (char_span::operator[]): New.
1906 (char_span::subspan): Make const.
1907 (char_span::xstrdup): New.
1908 (location_get_source_line): Convert return type from const char *
1909 to char_span, losing the final "line_size" param.
1911 2018-04-30 Jan Hubicka <jh@suse.cz>
1913 * lto-wrapper.c (ltrans_priorities): New static var.
1914 (cmp_priority): New.
1915 (run_gcc): Read priorities and if doing parallel build order
1916 the Makefile by them.
1918 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1920 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1922 2018-04-30 Richard Biener <rguenther@suse.de>
1924 * tree-cfg.c (verify_address): Remove base argument, add
1925 flag whether to check TREE_ADDRESSABLE and do that.
1926 (verify_expr): Remove.
1927 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1928 (verify_gimple_assign_single): Likewise.
1929 (verify_gimple_switch): Likewise.
1930 (verify_expr_location_1): Dereference tp once. Add (disabled)
1931 piece from verify_expr.
1932 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1934 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1936 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1938 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1940 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1941 (small_data_pattern): Likewise.
1942 (arc_rewrite_small_data): Likewise.
1943 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1944 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1945 (get_symbol_alignment): New function.
1946 (legitimate_small_data_address_p): Likewise.
1947 (legitimate_scaled_address): Update, call
1948 legitimate_small_data_address_p.
1949 (output_sdata): New static variable.
1950 (arc_print_operand): Update how we handle small data operands.
1951 (arc_print_operand_address): Likewise.
1952 (arc_legitimate_address_p): Update, use
1953 legitimate_small_data_address_p.
1954 (arc_rewrite_small_data_p): Remove.
1955 (arc_rewrite_small_data_1): Likewise.
1956 (arc_rewrite_small_data): Likewise.
1957 (small_data_pattern): Likewise.
1958 (compact_sda_memory_operand): Update to use
1959 legitimate_small_data_address_p and get_symbol_alignment.
1960 (prepare_move_operands): Don't rewite sdata pattern.
1961 (prepare_extend_operands): Remove.
1962 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1964 (zero_extendqisi2): Likewise.
1965 (zero_extendhisi2): Likewise.
1966 (extendqihi2): Likewise.
1967 (extendqisi2): Likewise.
1968 (extendhisi2): Likewise.
1972 * config/arc/constraints.md (Usd): Change it to memory constraint.
1974 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1976 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1977 as source of std instructions.
1978 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1979 allow 6-bit constants as source for store instructions.
1980 (movdi_insn): Update instruction pattern to allow 6-bit constants
1981 as source for store instructions.
1983 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1985 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1987 2018-04-30 Nathan Sidwell <nathan@acm.org>
1988 Sandra Loosemore <sandra@codesourcery.com>
1990 * dumpfile.c (dump_open): Allow '-' for stdout.
1991 * doc/invoke.texi (Developer Options): Document dump filename
1992 determination early. Document stdin/stdout selection.
1994 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1996 Microblaze Target: PIC data text relative
1998 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1999 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2001 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2002 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2003 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2004 New addressing mode for data-text relative position indepenedent code.
2005 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2006 'ADDRESS_SYMBOLIC_TXT_REL'.
2007 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2008 (microblaze_legitimate_pic_operand): Exclude function calls from
2009 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2010 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2012 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2013 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2014 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2015 for 'address + offset'.
2016 (microblaze_expand_prologue): Add new function prologue call for
2018 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2019 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2020 table in case of TARGET_PIC_DATA_TEXT_REL.
2021 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2022 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2023 Add new macros 'UNSPEC_TEXT',
2024 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2025 + exclude function calls from 'UNSPEC_PLT' in case of data text
2027 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2028 new target hook for generating address diff vector tables in case of
2030 * doc/tm.texi : Regenerate.
2031 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2032 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2033 of addr diff vector generation.
2034 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2035 target hook definition.
2036 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2037 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2038 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2039 option for data text relative.
2041 2018-04-30 Richard Biener <rguenther@suse.de>
2043 * tree-chrec.h (evolution_function_is_constant_p): Remove
2045 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2047 2018-04-30 Richard Biener <rguenther@suse.de>
2050 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2052 2018-04-30 Richard Biener <rguenther@suse.de>
2054 PR tree-optimization/28364
2055 PR tree-optimization/85275
2056 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2057 copying first exit test.
2059 2018-04-28 Mark Wielaard <mark@klomp.org>
2061 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2063 (dwarf_AT): Handle DW_AT_addr_base.
2064 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2066 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2069 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2070 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2071 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2072 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2074 2018-04-28 Richard Biener <rguenther@suse.de>
2076 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2077 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2078 to reflect use. Only add interesting stmts.
2080 2018-04-27 Martin Jambor <mjambor@suse.cz>
2083 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2084 the jump function allows for passing through aggregate values.
2086 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2088 * input.h (in_system_header_at): Convert from macro to inline
2090 (from_macro_expansion_at): Likewise.
2091 (from_macro_definition_at): Likewise.
2093 2018-04-27 Jeff Law <law@redhat.com>
2095 * config.gcc: Mark tile* targets as deprecated/obsolete.
2097 2018-04-27 Richard Biener <rguenther@suse.de>
2099 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2102 2018-04-27 Richard Biener <rguenther@suse.de>
2104 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2106 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2108 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2109 with Yd constraint. Set "preferred_for_speed" attribute from
2110 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2112 (*movdi_internal): Ditto.
2113 (movti_interunit splitters): Remove
2114 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2115 (movdi_interunit splitters): Ditto.
2116 * config/i386/constraints.md (Ye): Remove.
2117 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2119 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2122 * config/aarch64/constraints.md (Usg): Limit to 31.
2125 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2127 PR tree-optimization/85529
2128 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2129 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2130 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2131 zero extension or masking of the MSB bit.
2132 (optimize_range_tests): Add FIRST_BB argument, pass it through
2133 to optimize_range_tests_var_bound.
2134 (maybe_optimize_range_tests, reassociate_bb): Adjust
2135 optimize_range_tests callers.
2137 2018-04-26 Richard Biener <rguenther@suse.de>
2138 Jakub Jelinek <jakub@redhat.com>
2140 * cgraph.h (symbol_table): Just declare debug method here.
2141 * symtab.c (symbol_table::debug): Define.
2143 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2145 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2147 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2149 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2150 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2151 (*movdi_internal): Substitute Yi and Yj constraint with x
2152 and Ym and Yn constraint with y constraint. Update "isa"
2153 attribute and set "preferred_for_speed" attribute from
2154 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2155 (*movsi_internal): Ditto.
2156 (*movdf_internal): Ditto.
2157 (*movsf_internal): Ditto.
2158 (*zero_extendsidi2): Ditto.
2159 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2160 (sse2_loadld): Ditto.
2161 (*vec_extract<ssevecmodelower>_0): Ditto.
2162 (*vec_extractv4si_0_zext_sse4): Ditto.
2163 (vec_concatv2di): Ditto.
2164 (*vec_dup<mode>): Ditto.
2165 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2166 * config/i386/constraints.md (Yi): Remove.
2171 2018-04-26 Nathan Sidwell <nathan@acm.org>
2173 * dumpfile.c (dump_open): New.
2174 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2175 (dump_finish): Detect stdio/stderr by value not name.
2177 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2179 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2181 2018-04-26 Tom de Vries <tom@codesourcery.com>
2184 * config/nvptx/nvptx.c (verify_neutering_jumps)
2185 (verify_neutering_labels): New function
2186 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2188 2018-04-26 Tom de Vries <tom@codesourcery.com>
2191 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2192 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2195 2018-04-26 Richard Biener <rguenther@suse.de>
2196 Tom de Vries <tom@codesourcery.com>
2199 * lto-streamer-out.c (output_function): Fixup loops if required to match
2200 discovery done in the reader.
2202 2018-04-26 Richard Biener <rguenther@suse.de>
2204 PR tree-optimization/85116
2205 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2206 have a loop exit from the single latch predecessor. Remove
2207 case of header with just condition.
2208 (ch_base::copy_headers): Exclude infinite loops from any
2210 (pass_ch::execute): Record exits.
2212 2018-04-26 Richard Biener <rguenther@suse.de>
2214 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2215 prologue cost vector and pass it to vect_get_load_cost.
2216 (vect_get_peeling_costs_all_drs): Likewise.
2217 (vect_peeling_hash_get_lowest_cost): Likewise.
2218 (vect_enhance_data_refs_alignment): Likewise.
2220 2018-04-26 Richard Biener <rguenther@suse.de>
2223 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2224 checking of integer<->pointer conversions.
2225 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2226 sign-/zero-extending pointer types.
2227 (expand_omp_for_static_chunk): Likewise.
2229 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2230 Jean Lee <xiaoyur347@gmail.com>
2232 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2233 (TARGET_ASAN_SHADOW_OFFSET): Define.
2234 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2235 true for -fsanitize=address.
2237 2018-04-25 Mark Wielaard <mark@klomp.org>
2239 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2242 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2244 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2245 than "alu", remove explicit "memory" and "imm_disp" attributes.
2246 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2249 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2250 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2251 gen_lowpart_no_emit.
2253 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2256 * config/i386/i386.c (ix86_expand_builtin): Change memory
2257 operand to XI, extend p0 to Pmode.
2258 * config/i386/i386.md: Change unspec volatile and operand
2259 1 mode to XI, change operand 0 mode to P.
2261 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2263 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2264 GET_MODE_MASK before any checking.
2265 (nds32_can_use_bset_p): Likewise.
2266 (nds32_can_use_btgl_p): Likewise.
2268 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2270 * config/nds32/nds32-doubleword.md: New define_split pattern for
2271 illegal register number.
2273 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2275 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2277 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2279 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2281 2018-04-25 Richard Biener <rguenther@suse.de>
2283 * lto-streamer.h (LTO_major_version): Bump to 8.
2285 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2287 * BASE-VER: Set to 9.0.0.
2289 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2291 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2292 in __abskf2 and __powikf2.
2294 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2297 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2298 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2299 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2300 Use the above on operand 2. Reindent.
2301 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2303 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2306 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2308 * config/i386/i386.opt (mcet): Removed.
2309 * doc/install.texi: Remove -mcet documentation.
2310 * doc/invoke.texi: Likewise.
2312 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2315 * doc/install.texi: Remove -mcet from bootstrap-cet.
2317 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2320 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2321 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2325 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2326 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2327 containing a CONST_VECTOR.
2329 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2331 * doc/install.texi: Update newlib dependency for nvptx.
2333 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2336 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2337 instead of INTVAL when shifting x left.
2339 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2341 PR tree-optimization/85478
2342 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2343 vect_grouped_store_supported for single element vectors.
2345 2018-04-24 Richard Biener <rguenther@suse.de>
2348 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2349 load cost increase to the case of non-constant step.
2351 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2354 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2355 destination if any_malformed_asm.
2357 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2360 * expr.c (store_field): In the bitfield case, if the value comes from
2361 a function call and is returned in registers by means of a PARALLEL,
2362 do not change the mode of the temporary unless BLKmode and VOIDmode.
2364 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2366 PR rtl-optimization/85423
2367 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2368 dependencies to debug insns when the previous insn is non-debug.
2370 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2372 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2373 enums into a single definition.
2374 (fls): Fix predicates and printing.
2377 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2379 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2380 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2381 and short u6 immediate.
2382 (check_if_valid_sleep_operand): Remove.
2383 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2385 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2387 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2388 flag_always_save_lp condition.
2389 * config/nds32/nds32.opt (malways-save-lp): New option.
2391 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2393 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2394 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2395 * config/nds32/nds32.h
2396 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2397 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2399 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2401 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2402 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2404 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2405 Chung-Ju Wu <jasonwucj@gmail.com>
2407 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2409 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2410 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2412 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2414 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2416 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2418 * config/nds32/nds32-protos.h (nds32_data_alignment,
2419 nds32_local_alignment): Declare.
2420 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2421 nds32_local_alignment): New functions.
2422 (TARGET_CONSTANT_ALIGNMENT): Define.
2423 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2425 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2427 * config/nds32/nds32.c
2428 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2429 (TARGET_MODES_TIEABLE_P): Likewise.
2431 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2433 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2436 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2437 Chung-Ju Wu <jasonwucj@gmail.com>
2439 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2440 for unaligned access.
2441 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2443 * config/nds32/nds32-intrinsic.md: Likewise.
2444 * config/nds32/nds32_intrinsic.h: Likewise.
2445 * config/nds32/nds32.h (nds32_builtins): Likewise.
2446 * config/nds32/nds32.opt (munaligned-access): New option.
2447 * config/nds32/nds32.c (nds32_asm_file_start): Display
2448 flag_unaligned_access status.
2450 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2452 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2453 -mno-relax is present.
2454 * config/riscv/linux.h (LINK_SPEC): Ditto.
2456 2018-04-20 Martin Sebor <msebor@redhat.com>
2459 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2461 (gimple_fold_builtin_stxcpy_chk): Same.
2462 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2464 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2467 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2468 __powikf2 when long double is IEEE 128-bit.
2470 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2472 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2473 step to make sure stack always aligned.
2475 2018-04-20 Carl Love <cel@us.ibm.com>
2478 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2479 size check for arg0.
2481 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2482 Tom de Vries <tom@codesourcery.com>
2485 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2486 Emit insns for calls too.
2487 (nvptx_find_par): Always look for worker-level predecessor insn.
2488 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2490 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2491 (nvptx_process_pars): Propagate frames for calls.
2493 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2496 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2498 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2499 (ix86_handle_option): Don't handle OPT_mibt.
2500 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2502 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2504 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2505 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2506 (ix86_target_macros): Define __CET__ with flag_cf_protection
2507 for -fcf-protection.
2508 * config/i386/i386.c (isa2_opts): Remove -mibt.
2509 * config/i386/i386.h (TARGET_IBT): Removed.
2510 (TARGET_IBT_P): Likewise.
2511 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2512 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2513 * config/i386/i386.opt (mcet): Update help message.
2516 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2517 -mcet as an alias for -mshstk.
2519 2018-04-20 Richard Biener <rguenther@suse.de>
2522 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2523 complexity by forcing a single use of the multiply operand.
2525 2018-04-20 Martin Jambor <mjambor@suse.cz>
2528 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2529 recursion dependency to only apply to non-clones.
2531 2018-04-20 Martin Jambor <mjambor@suse.cz>
2534 * ipa-cp.c (create_specialized_node): Check that clones of
2535 self-recursive edges exist during IPA-CP.
2537 2018-04-19 Toon Moene <toon@moene.org>
2539 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2542 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2544 PR tree-optimization/85467
2545 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2546 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2547 VECTOR_CST element to type.
2549 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2552 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2553 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2554 (builtin_longjmp): Likewise.
2555 (save_stack_nonlocal): New pattern.
2556 (restore_stack_nonlocal): Likewise.
2558 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2561 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2562 Replace ASM_OUTPUT_LABEL with fprintf.
2564 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2567 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2568 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2569 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2570 define __IBT__ and __SHSTK__ for -fcf-protection.
2571 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2573 (ix86_trampoline_init): Likewise.
2574 (x86_output_mi_thunk): Likewise.
2575 (ix86_notrack_prefixed_insn_p): Likewise.
2576 (ix86_option_override_internal): Don't disallow -fcf-protection.
2577 * config/i386/i386.md (rdssp<mode>): Also enable for
2579 (incssp<mode>): Likewise.
2580 (nop_endbr): Likewise.
2581 * config/i386/i386.opt (mcet): Change help message to built-in
2585 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2586 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2587 enable CET built-in functions.
2589 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2591 * common/config/i386/i386-common.c
2592 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2593 OPTION_MASK_ISA_MOVDIRI_UNSET,
2594 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2595 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2596 * config.gcc (movdirintrin.h): New header.
2597 * config/i386/cpuid.h (bit_MOVDIRI,
2598 bit_MOVDIR64B): New bits.
2599 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2601 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2602 (VOID, PVOID, PCVOID)): New function types.
2603 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2604 __builtin_ia32_directstoreu_u64,
2605 __builtin_ia32_movdir64b): New builtins.
2606 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2607 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2609 (ix86_valid_target_attribute_inner_p): Ditto.
2610 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2611 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2612 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2613 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2614 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2615 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2616 (movdiri<mode>, movdir64b_<mode>): New.
2617 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2618 * config/i386/immintrin.h: Include movdirintrin.h.
2619 * config/i386/movdirintrin.h: New file.
2620 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2622 2018-04-19 Richard Biener <rguenther@suse.de>
2625 * cfg.c (clear_bb_flags): When loop state says we have
2626 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2628 2018-04-19 Richard Biener <rguenther@suse.de>
2630 PR tree-optimization/84737
2631 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2632 copying restrict info.
2633 (vect_setup_realignment): Use it.
2634 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2635 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2636 the first DR to all generated stores.
2637 (vectorizable_load): Likewise for loads.
2639 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2641 PR tree-optimization/85446
2642 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2643 the integral and pointer types to have the same precision.
2645 * doc/install.texi: Document --disable-cet being the default and
2648 2018-04-18 Martin Liska <mliska@suse.cz>
2650 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2653 2018-04-18 Martin Liska <mliska@suse.cz>
2656 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2659 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2660 arguments if they are comparable.
2662 2018-04-18 Martin Liska <mliska@suse.cz>
2665 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2668 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2671 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2674 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2675 ENDBR after calling __morestack.
2677 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2680 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2681 by using gcc_base_ver to generate a gcc_driver_version, and use
2682 it when generating GCC_DRIVER_NAME.
2683 * configure: Regenerate.
2685 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2688 * config.gcc: Obsolete powerpc*-*-*spe*.
2690 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2693 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2694 (stabstr_D): Change type of unum from unsigned int to
2695 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2698 2018-04-17 Jim Wilson <jimw@sifive.com>
2701 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2702 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2703 Set arg_pointer_offset after using pretend_args_size.
2705 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2707 PR rtl-optimization/85431
2708 * dse.c (record_store): Ignore zero width stores.
2711 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2712 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2713 __builtin_stack_restore rather than after it.
2714 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2715 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2716 argument instead of virtual_dynamic_stack_rtx.
2718 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2720 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2722 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2723 Add note to error message to explain internal mapping of overloaded
2724 built-in function name to non-overloaded built-in function name.
2725 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2728 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2731 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2732 where the inputs overlap with the output.
2734 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2736 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2737 (=v, v) alternative and explicit "memory" attribute.
2738 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2739 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2741 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2742 "sselog1" type instead of "sselog".
2743 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2744 "sselog". Remove explicit "memory" attribute.
2745 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2746 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2748 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2749 "sselog1" type instead of "sselog". Remove explicit "memory"
2751 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2752 use "sselog1" type instead of "sselog". Remove explicit "memory"
2754 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2755 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2757 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2758 "sselog1" type instead of "sselog". Remove explicit "memory"
2760 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2761 use "sselog1" type instead of "sselog". Remove explicit "memory"
2765 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2768 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2771 2018-04-17 Martin Jambor <mjambor@suse.cz>
2774 * ipa-cp.c (create_specialized_node): Call
2775 expand_all_artificial_thunks if necessary.
2777 2018-04-17 Martin Liska <mliska@suse.cz>
2780 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2781 in message, remote space in between '_G' and '('.
2783 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2786 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2787 avx512f_vmcmp<mode>3<round_saeonly_name>,
2788 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2789 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2790 avx512f_rndscale<mode><round_saeonly_name>,
2791 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2792 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2793 Use %<iptr>2 instead of %2 for -masm=intel.
2794 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2795 avx512f_vcvttss2usi<round_saeonly_name>,
2796 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2798 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2799 avx512f_vcvttsd2usi<round_saeonly_name>,
2800 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2801 Use %q1 instead of %1 for -masm=intel.
2802 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2803 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2804 of %3 for -masm=intel.
2805 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2807 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2809 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2811 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2812 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2814 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2816 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2817 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2818 %g1 and one with %0 and %1.
2819 (avx512er_vmrcp28<mode><round_saeonly_name>,
2820 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2822 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2823 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2824 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2825 of %0 and %{%4%} for -masm=intel.
2826 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2827 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2828 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2829 order of %0 and %{%5%}%{z%} for -masm=intel.
2831 2018-04-17 Jan Hubicka <jh@suse.cz>
2834 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2836 2018-04-17 Martin Liska <mliska@suse.cz>
2839 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2840 for target_clone error message. Make default implementation
2841 clone to be a local declaration.
2842 (separate_attrs): Add new argument and check for an empty
2844 (expand_target_clones): Handle it.
2845 (ipa_target_clone): Make redirection just for target_clones
2848 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2849 Tom de Vries <tom@codesourcery.com>
2852 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2853 tiled basic blocks without omp continue statements.
2855 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2858 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2859 vec_extract expression as having side effects to make sure it gets
2862 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2865 * config/i386/i386.c (get_builtin_code_for_version): Check
2868 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2871 * gcc/config.gcc: Support "skylake".
2872 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2874 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2875 (processor_target_table): Add "skylake".
2876 (ix86_option_override_internal): Add "skylake".
2877 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2878 PROCESSOR_CANNONLAKE.
2879 (get_builtin_code_for_version): Fix priority for
2880 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2881 PROCESSOR_SKYLAKE-AVX512.
2882 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2883 (processor_type): Add PROCESSOR_SKYLAKE.
2885 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2886 Jason Merrill <jason@redhat.com>
2889 * convert.c (convert_to_integer_1): Use direct recursion for
2890 enumeral types and types with a precision less than the number
2891 of bits in their mode.
2893 2018-04-16 Julia Koval <julia.koval@intel.com>
2896 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2897 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2899 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2902 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2903 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2904 and -mno-direct-move.
2906 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2909 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2910 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2911 Ensure negative shifts result in {0}.
2913 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2915 PR rtl-optimization/79916
2916 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2917 regs (if any) to define how to gnerate SD moves when LRA is in
2920 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2922 PR rtl-optimization/85393
2923 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2924 * except.c (expand_dw2_landing_pad_for_region): Make static.
2925 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2926 a label and unconditional jump to old_bb, rather than
2927 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2930 PR rtl-optimization/85376
2931 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2932 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2933 instead of a specific value.
2935 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2936 Bin Cheng <bin.cheng@arm.com>
2938 PR tree-optimization/82965
2939 PR tree-optimization/83991
2940 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2941 by_profile_only parameter.
2942 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2943 information if the loop was predicted to iterate too many times.
2944 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2946 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2949 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2952 2018-04-13 Martin Liska <mliska@suse.cz>
2953 Jakub Jelinek <jakub@redhat.com>
2956 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2957 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2958 * builtins.c (expand_builtin_memory_copy_args): Use
2959 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2960 handle dest_addr == pc_rtx.
2962 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2965 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2966 asked to not generate direct moves.
2967 (fix_trunc<mode>si2_stfiwx): Similar.
2968 (fix_trunc<mode>si2_internal): Similar.
2970 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2973 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2974 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2975 lookup if dest in some wider mode is known to be const0_rtx and
2976 if so, record permanent equivalence for it to be ZERO_EXTEND of
2977 the narrower mode destination.
2979 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2981 * lto-streamer-out.c (output_function): Revert 259346.
2982 * omp-expand.c (expand_oacc_for): Likewise.
2984 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2986 PR rtl-optimization/85354
2987 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2988 * sel-sched.c (sel_global_init): ... here.
2990 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2993 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2994 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2995 mode for PE-COFF targets.
2996 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2997 (i386_pe_asm_lto_end): Likewise.
2998 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2999 (TARGET_ASM_LTO_END): Likewise.
3000 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3001 (i386_pe_asm_lto_start): New function.
3002 (i386_pe_asm_lto_end): Likewise.
3004 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3005 Richard Biener <rguenther@suse.de>
3008 * lto-streamer-out.c (output_function): Fix CFG loop state before
3010 * omp-expand.c (expand_oacc_for): Handle calls to internal
3011 functions like regular functions.
3013 2018-04-12 Richard Biener <rguenther@suse.de>
3016 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3017 for the early LTO debug to properly generate references to it
3018 during DIE emission. Do not re-use that for the skeleton for
3020 (dwarf2out_early_finish): Likewise.
3022 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3025 * config/i386/sse.md
3026 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3027 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3028 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3029 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3030 and output is a reg, avoid creating invalid lowpart subreg, but
3031 instead split into a 512-bit move. Don't split if not AVX512VL,
3032 input is xmm16+ reg and output is a mem.
3033 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3034 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3035 xmm16+ reg and output is a mem.
3037 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3039 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3040 also for flag_dwarf2_cfi_asm.
3042 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3044 PR rtl-optimization/85342
3045 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3046 a bool scalar var inside of the loop instead. Don't try to update
3047 recog_data.operand after failed apply_change_group.
3049 2018-04-12 Tom de Vries <tom@codesourcery.com>
3052 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3053 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3054 array with flexible array member as array without given dimension.
3055 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3056 argument for undefined param to true.
3058 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3061 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3062 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3063 from PowerPC section.
3064 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3065 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3066 help text that is too long.
3067 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3068 help text that is too long.
3069 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3070 help text that is too long.
3072 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3074 * config/alpha/alpha.md (stack_probe_internal): Rename
3075 from "probe_stack". Update all callers.
3077 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3079 PR rtl-optimization/84566
3080 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3081 sched_macro_fuse_insns.
3083 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3086 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3087 (compute_block_dependences): ... from here.
3089 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3091 PR tree-optimization/85331
3092 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3093 from int to HOST_WIDE_INT.
3095 2018-04-11 Martin Jambor <mjambor@suse.cz>
3098 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3099 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3100 not the same as the source val.
3101 (cgraph_edge_brings_value_p): New parameter.
3102 (gather_edges_for_value): Pass destination value to
3103 cgraph_edge_brings_value_p.
3104 (perhaps_add_new_callers): Likewise.
3105 (get_info_about_necessary_edges): Likewise and exclude values brought
3106 only by self-recursive edges.
3107 (create_specialized_node): Redirect only clones of self-calling edges.
3108 (+self_recursive_pass_through_p): New function.
3109 (find_more_scalar_values_for_callers_subset): Use it.
3110 (find_aggregate_values_for_callers_subset): Likewise.
3111 (known_aggs_to_agg_replacement_list): Removed.
3112 (decide_whether_version_node): Re-calculate known constants for all
3113 remaining context clones.
3115 2018-04-11 Richard Biener <rguenther@suse.de>
3118 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3119 from early DWARF output.
3120 (dwarf2out_early_finish): Output line info unconditionally into
3121 early DWARF and add reference to it.
3123 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3126 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3127 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3128 other than V2DFmode using iptr mode attribute.
3129 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3131 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3133 PR rtl-optimization/84659
3134 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3136 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3139 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3141 (output_loc_list): Pass address of a dummy size variable even in the
3142 locview handling loop.
3143 (index_location_lists): Add comment on why skip_loc_list_entry can't
3146 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3149 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3152 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3155 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3156 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3157 and -mstring-compare-inline-limit.
3159 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3162 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3163 for stack clash protection in a register whenever we need it to be in
3166 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3168 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3169 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3171 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3174 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3176 (mlong-double-): Ditto.
3177 * config/rs6000/sysv4.opt (msdata=): Ditto.
3178 (mtls-size=): Ditto.
3180 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3182 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3183 erroneous entries for
3184 "vector int vec_ldl (int, long int *)", and
3185 "vector unsigned int vec_ldl (int, unsigned long int *)".
3186 Add comments and entries for
3187 "vector bool char vec_ldl (int, bool char *)",
3188 "vector bool short vec_ldl (int, bool short *)",
3189 "vector bool int vec_ldl (int, bool int *)",
3190 "vector bool long long vec_ldl (int, bool long long *)",
3191 "vector pixel vec_ldl (int, pixel *)",
3192 "vector long long vec_ldl (int, long long *)",
3193 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3194 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3195 type tree bool_long_long_type_node and correct definition of
3196 bool_V2DI_type_node to make reference to this new type tree.
3197 (rs6000_mangle_type): Replace erroneous reference to
3198 bool_long_type_node with bool_long_long_type_node.
3199 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3200 comments to emphasize sign distinctions for char and int types and
3201 replace RS6000_BTI_bool_long constant with
3202 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3203 use of RS6000_BTI_pixel.
3204 (bool_long_type_node): Remove this macro definition.
3205 (bool_long_long_type_node): New macro definition
3207 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3209 PR rtl-optimization/85300
3210 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3211 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3212 simplify_unary_operation fails.
3214 2018-04-10 Martin Liska <mliska@suse.cz>
3216 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3217 cgraph_edge and ipa_ref.
3219 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3223 * config/i386/sse.md
3224 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3225 computation of the VEC_MERGE selector from mask.
3226 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3227 Fix decoding of the VEC_MERGE selector into mask.
3229 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3231 PR tree-optimization/85286
3232 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3234 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3236 * final.c (final_1): Set insn_last_address as well as
3237 insn_current_address.
3239 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3242 * explow.c (emit_stack_probe): Call validize_mem on memory location
3243 before passing it to gen_probe_stack. Create address operand and
3244 legitimize it for the probe_stack_address case.
3246 2018-04-09 Jan Hubicka <jh@suse.cz>
3249 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3250 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3251 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3253 (free_lang_data): Rebuild type inheritance graph.
3255 2018-04-09 Martin Sebor <msebor@redhat.com>
3257 * invoke.texi (-finline-small-functions): Mention other optimization
3259 (-findirect-inlining, -fpartial-inlining): Same.
3260 (-finline-functions-called-once): Same.
3261 (-freorder-blocks-and-partition): Same.
3263 2018-04-09 Jan Hubicka <jh@suse.cz>
3266 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3267 jumps; choose last target that matches the criteria (i.e.
3268 no partition changes for non-crossing jumps).
3269 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3270 support for redirecting crossing jumps to non-crossing.
3272 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3274 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3275 also for naked functions.
3277 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3279 * config/arc/arc.md (add_shift): New pattern.
3280 (add_shift2): Likewise.
3281 (sub_shift): Likewise.
3282 (sub_shift_cmp0_noout): Likewise.
3283 (compare_si_ashiftsi): Likewise.
3284 (xbfu_cmp0_noout): New combine pattern.
3285 (xbfu_cmp0"): Likewise.
3286 (movsi_set_cc_insn): Place the predicable variant first.
3287 (commutative_binary_cmp0_noout): Remove clobber.
3288 (commutative_binary_cmp0): New pattern.
3289 (noncommutative_binary_cmp0): Likewise.
3290 (noncommutative_binary_cmp0_noout): Likewise.
3291 (noncommutative_binary_comparison_result_used): Removed.
3292 (rsub_cmp0): New pattern.
3293 (rsub_cmp0_noout): Likewise.
3294 (extzvsi): Changed, keep only meaningful variants.
3295 (SQH, SEZ): New iterators.
3296 (SQH_postfix): New mode attribute.
3297 (SEZ_prefix): New code attribute.
3298 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3299 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3300 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3302 (noncommutative_operator): Check the availability of barrel
3305 2018-04-09 Richard Biener <rguenther@suse.de>
3307 PR tree-optimization/85284
3308 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3309 Only use the niter constraining form of simple_iv when the exit
3312 2018-04-09 Tom de Vries <tom@codesourcery.com>
3315 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3316 (define_expand "*memory_barrier"): New define_expand.
3317 (define_insn "memory_barrier"): New insn.
3319 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3321 PR rtl-optimization/80463
3322 PR rtl-optimization/83972
3323 PR rtl-optimization/83480
3325 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3326 correct producer for the insn.
3327 (tidy_control_flow): Fixup seqnos in case of debug insns.
3329 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3331 PR rtl-optimization/83913
3333 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3334 different sched-times when merging exprs.
3336 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3338 PR rtl-optimization/83962
3340 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3341 tidy_fallthru_edge and tidy_control_flow.
3343 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3345 PR rtl-optimization/83530
3347 * sel-sched.c (force_next_insn): New global variable.
3348 (remove_insn_for_debug): When force_next_insn is true, also leave only
3349 next insn in the ready list.
3350 (sel_sched_region): When the region wasn't scheduled, make another pass
3351 over it with force_next_insn set to 1.
3353 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3355 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3357 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3358 for interrupt control.
3359 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3360 functions for interrupt control.
3361 * config/nds32/nds32-intrinsic.md: Likewise.
3362 * config/nds32/nds32_intrinsic.h: Likewise.
3363 * config/nds32/nds32.h (nds32_builtins): Likewise.
3365 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3367 * config/nds32/nds32.c (nds32_init_machine_status,
3368 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3369 strict_aligned_p field.
3370 (nds32_expand_to_rtl_hook): New function.
3371 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3372 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3374 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3375 Chung-Ju Wu <jasonwucj@gmail.com>
3377 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3378 * config/nds32/nds32-n7.md: New file.
3379 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3380 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3382 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3383 * config/nds32/nds32.md (pipeline_model): Add n7.
3384 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3385 * config/nds32/pipelines.md: Include n7 settings.
3387 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3388 Chung-Ju Wu <jasonwucj@gmail.com>
3390 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3391 * config/nds32/nds32-e8.md: New file.
3392 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3393 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3395 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3396 * config/nds32/nds32.md (pipeline_model): Add e8.
3397 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3398 * config/nds32/pipelines.md: Include e8 settings.
3400 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3401 Chung-Ju Wu <jasonwucj@gmail.com>
3403 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3404 * config/nds32/nds32-n8.md: New file.
3405 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3406 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3408 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3409 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3410 * config/nds32/nds32.md (pipeline_model): Add n8.
3411 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3412 * config/nds32/pipelines.md: Include n8 settings.
3414 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3415 Chung-Ju Wu <jasonwucj@gmail.com>
3417 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3418 * config/nds32/nds32-n9-2r1w.md: New file.
3419 * config/nds32/nds32-n9-3r2w.md: New file.
3420 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3421 nds32_register_ports): New or modify for cpu n9.
3422 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3424 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3425 * config/nds32/nds32-utils.c: New file.
3426 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3427 TARGET_MUL_SLOW): Define.
3428 * config/nds32/nds32.md (pipeline_model): New attribute.
3429 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3430 New options that support cpu n9.
3431 * config/nds32/pipelines.md: Include n9 settings.
3432 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3434 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3436 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3437 information if necessary.
3438 (output_cond_branch_compare_zero): Likewise.
3439 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3440 (nds32_target_alignment): Refine for alignment.
3441 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3442 (FUNCTION_BOUNDARY): Modify.
3443 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3445 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3447 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3449 * config/nds32/constants.md (unspec_volatile_element): Add values for
3450 TLB operation and data prefetch.
3451 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3452 functions for TLB operation and data prefetch.
3453 * config/nds32/nds32-intrinsic.md: Likewise.
3454 * config/nds32/nds32_intrinsic.h: Likewise.
3455 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3456 (nds32_print_operand): Likewise.
3457 * config/nds32/nds32.h (nds32_builtins): Likewise.
3459 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3460 Andrew Pinski <pinsika@gcc.gnu.org>
3463 * match.pd: Use constant_boolean_node of correct type instead of
3464 boolean_true_node or boolean_false_node for simplifying
3465 pointer comparisons to zero.
3467 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3469 PR tree-optimization/80021
3470 * tree.c (verify_type_variant): Make error call in verify_variant_match
3471 translatable and remove final full stop.
3473 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3475 * config/nds32/constants.md (unspec_volatile_element): Add
3476 UNSPEC_VOLATILE_EH_RETURN.
3477 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3478 nds32_output_stack_pop): Support dwarf exception handling process.
3479 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3480 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3481 exception handling process.
3482 (nds32_compute_stack_frame): Likewise.
3483 (nds32_return_addr_rtx): Likewise.
3484 (nds32_initial_elimination_offset): Likewise.
3485 (nds32_expand_prologue): Likewise.
3486 (nds32_expand_epilogue): Likewise.
3487 (nds32_dynamic_chain_address): New function.
3488 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3490 (DYNAMIC_CHAIN_ADDRESS): Define.
3491 (EH_RETURN_DATA_REGNO): Define.
3492 (EH_RETURN_STACKADJ_RTX): Define.
3493 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3494 patterns for dwarf exception handling.
3496 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3498 * config/nds32/nds32.h: Clean up obsolete macros.
3500 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3502 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3503 Add enum values for particular instructions.
3504 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3505 particular intrinsic functions.
3506 * config/nds32/nds32-intrinsic.md: Likewise.
3507 * config/nds32/nds32_intrinsic.h: Likewise.
3508 * config/nds32/nds32.h (nds32_builtins): Likewise.
3509 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3510 (btst, ave): New patterns for particular instructions.
3512 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3514 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3515 Add enum values for atomic load/store and memory sync.
3516 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3518 * config/nds32/nds32-intrinsic.md: Likewise.
3519 * config/nds32/nds32_intrinsic.h: Likewise.
3520 * config/nds32/nds32.h (nds32_builtins): Likewise.
3522 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3524 PR tree-optimization/85257
3525 * fold-const.c (native_encode_vector): If not all elts could fit
3526 and off is -1, return 0 rather than offset.
3527 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3528 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3529 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3530 adjust buffer in native_interpret_expr call.
3532 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3534 * config/nds32/constants.md (unspec_volatile_element): Add cache
3535 control enum values.
3536 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3537 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3538 * config/nds32/nds32.c (nds32_cctl_names): New.
3539 (nds32_print_operand): Handle cache control register names.
3540 * config/nds32/nds32.h (nds32_builtins): New enum values.
3541 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3543 * config/nds32/nds32.md (type): Add mmu.
3544 * config/nds32/pipelines.md (simple_insn): Add mmu.
3546 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3548 * config/nds32/nds32.md (type): Remove call.
3549 * config/nds32/pipelines.md (simple_insn): Likewise.
3551 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3553 * config/nds32/constants.md (unspec_volatile_element): Add
3554 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3555 UNSPEC_VOLATILE_FMFCFG.
3556 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3557 description for fmfcfg and fmfcsr.
3558 (bdesc_1arg): Add fmtcsr.
3559 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3560 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3561 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3562 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3563 unspec_fmfcfg): New patterns.
3564 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3565 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3566 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3567 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3568 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3569 __nds32__fmfcfg): Define.
3571 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3573 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3574 intrinsic register names.
3575 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3576 intrinsic register enum values and macros.
3578 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3580 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3581 for load/store addressing form.
3582 (nds32_print_operand_address): Likewise.
3584 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3587 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3588 based on LABEL_REF. Remove useless assertion.
3589 (pic_address_needs_scratch): Fix formatting.
3590 (sparc_legitimize_pic_address): Minor tweaks.
3591 (sparc_delegitimize_address): Adjust assertion accordingly.
3592 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3593 into symbolic_operand.
3594 (movsi_high_pic_label_ref): Likewise.
3595 (movsi_lo_sum_pic_label_ref): Likewise.
3596 (movdi_pic_label_ref): Likewise.
3597 (movdi_high_pic_label_ref): Likewise.
3598 (movdi_lo_sum_pic_label_ref): Likewise.
3600 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3602 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3603 custom LIB_SPEC setup.
3605 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3606 Kito Cheng <kito.cheng@gmail.com>
3608 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3609 * config/riscv/freebsd.h: New.
3611 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3613 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3614 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3617 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3618 Kito Cheng <kito.cheng@gmail.com>
3620 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3621 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3622 * config/nds32/nds32-protos.h (nds32_output_call,
3623 nds32_output_return): Declare.
3624 * config/nds32/nds32.md: Refine all the call and return patterns.
3626 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3629 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3630 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3632 PR rtl-optimization/84872
3633 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3634 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3637 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3639 * expr.c (copy_blkmode_to_reg): Revert 254862.
3640 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3642 2018-04-06 Richard Biener <rguenther@suse.de>
3645 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3646 after seeing a component reference with an adjacent field. Treat
3647 refs to arrays at struct end of external decls similar to
3648 refs to unconstrained commons.
3650 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3653 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3654 look through SAVE_EXPRs with non-side-effects argument. Adjust
3656 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3659 2018-04-06 Richard Biener <rguenther@suse.de>
3662 * alias.c (find_base_term): New wrapper around find_base_term
3663 unwinding CSELIB_VAL_PTR changes.
3664 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3667 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3669 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3671 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3672 constant definitions.
3673 ("nop"): lr 0,0 -> nopr r0
3674 ("nop_lr0", "nop_lr1"): New insn definitions.
3676 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3678 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3679 NDS32_V3PUSH_AVAILABLE_P macro.
3681 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3682 Chung-Ju Wu <jasonwucj@gmail.com>
3684 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3685 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3686 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3687 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3688 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3689 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3690 * config/nds32/constraints.md: New constraints and checking for hard
3691 float configuration.
3692 * config/nds32/iterators.md: New mode iterator and attribute for hard
3693 float configuration.
3694 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3696 * config/nds32/nds32-fpu.md: New file.
3697 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3698 deal with hard float code generation.
3699 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3701 (abi_type, float_reg_number): New enum type.
3702 * config/nds32/nds32-predicates.c: New predicates for hard float.
3703 * config/nds32/nds32-protos.h: Declare functions for hard float.
3704 * config/nds32/nds32.c: Implementation for hard float configuration.
3705 * config/nds32/nds32.h: Definitions for hard float configuration.
3706 * config/nds32/nds32.md: Include hard float machine description and
3707 modify patterns for hard float configuration.
3708 * config/nds32/nds32.opt: New options for hard float configuration.
3709 * config/nds32/predicates.md: New predicates for hard float
3712 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3714 * common/config/nds32/nds32-common.c
3715 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3717 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3720 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3721 CONSTRUCTOR_ELT (ctor, ...)->value.
3723 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3726 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3728 2018-04-05 Tom de Vries <tom@codesourcery.com>
3731 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3734 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3735 Kito Cheng <kito.cheng@gmail.com>
3737 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3738 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3739 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3740 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3742 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3743 Kito Cheng <kito.cheng@gmail.com>
3745 * config/nds32/constraints.md (Ufe): New memory constraint.
3746 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3747 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3748 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3750 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3751 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3753 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3755 * config/nds32/nds32.md: Use optimize_size in the condition for
3756 alu-shift instructions.
3758 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3760 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3762 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3764 * config/nds32/nds32.md (negsi2): Refine pattern.
3766 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3767 Chung-Ju Wu <jasonwucj@gmail.com>
3769 * config/nds32/iterators.md (shift_rotate): New code iterator.
3770 (shift): New code attribute.
3771 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3772 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3773 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3774 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3775 bit-wise operations.
3776 (andsi3, *andsi3): Ditto.
3777 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3778 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3779 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3780 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3781 nds32_ior_operand, nds32_xor_operand): New predicates.
3783 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3785 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3786 (addsi3, subsi3): ... this.
3788 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3790 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3792 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3794 * config/nds32/nds32.md: Adjust indention.
3796 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3798 * config/nds32/nds32.md (feature): New attribute.
3800 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3802 * config/nds32/nds32.md (subtype): New attribute.
3804 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3807 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3808 expansion to perform a bitwise AND of the argument followed by a
3809 boolean negation of the result.
3811 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3813 PR rtl-optimization/84878
3814 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3815 the basic block. Assert the use reference is not artificial and that
3816 it has an associated insn.
3818 2018-04-04 Michael Matz <matz@suse.de>
3820 * builtins.c (compute_objsize): Pass correct operand
3821 to array_at_struct_end_p.
3823 2018-04-04 Richard Biener <rguenther@suse.de>
3826 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3827 from contexts for DINFO_LEVEL_TERSE and below.
3829 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3831 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3832 resiter_operand condition.
3833 * config/nds32/nds32.md (*move<mode>): Ditto.
3835 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3836 Monk Chiang <sh.chiang04@gmail.com>
3838 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3840 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3842 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3844 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3845 Kito Cheng <kito.cheng@gmail.com>
3847 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3848 nds32_cond_code_str, output_cond_branch,
3849 output_cond_branch_compare_zero, nds32_expand_cbranch,
3850 nds32_expand_cstore, nds32_expand_movcc,
3851 nds32_output_cbranchsi4_equality_zero,
3852 nds32_output_cbranchsi4_equality_reg,
3853 nds32_output_cbranchsi4_equality_reg_or_const_int,
3854 nds32_output_cbranchsi4_greater_less_zero: New functions.
3855 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3856 nds32_expand_cstore, nds32_expand_movcc,
3857 nds32_output_cbranchsi4_equality_zero,
3858 nds32_output_cbranchsi4_equality_reg,
3859 nds32_output_cbranchsi4_equality_reg_or_const_int,
3860 nds32_output_cbranchsi4_greater_less_zero): Declare.
3861 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3862 nds32_rimm11s_operand): New predicates.
3863 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3864 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3867 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3869 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3870 * config/nds32/nds32.md: Ditto.
3871 * config/nds32/pipelines.md: Ditto.
3873 2018-04-04 Richard Biener <rguenther@suse.de>
3875 PR tree-optimization/85168
3876 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3877 propagating abnormals.
3879 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3881 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3883 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3884 Kito Cheng <kito.cheng@gmail.com>
3886 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3887 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3888 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3889 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3890 * config/nds32/nds32.md (sibcall_internal): New.
3891 (sibcall_register): Remove.
3892 (sibcall_immediate): Remove.
3893 (sibcall_value_internal): New.
3894 (sibcall_value_register): Remove.
3895 (sibcall_value_immediate): Remove.
3896 * config/nds32/predicates.md (nds32_general_register_operand): New.
3897 (nds32_call_address_operand): New.
3899 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3901 PR rtl-optimization/85167
3902 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3903 bb_defs if *split_p, instead preinitialize it to NULL.
3905 PR tree-optimization/85156
3906 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3907 evaluating the argument multiple times.
3909 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3911 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3913 (_mm_cvtpd_ps): Likewise.
3914 (_mm_cvttpd_epi32): Likewise.
3915 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3916 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3917 vector, pixel, and bool following altivec.h include.
3919 2018-04-03 Martin Sebor <msebor@redhat.com>
3921 * doc/extend.texi (Common Function Attributes): Clarify.
3922 (const attribute): Likewise.
3923 (pure attribute): Likewise.
3925 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3928 * config/i386/i386.c (ix86_expand_vector_set): Use
3929 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3931 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3933 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3934 instructions when changing rounding bits to preserve precision bits
3935 in the x87 control word.
3937 2018-04-03 Martin Liska <mliska@suse.cz>
3939 PR tree-optimization/82491
3940 * rtl.h (strip_offset_and_add): Replace += suboffset with
3941 poly_uint64 () + suboffset.
3943 2018-03-29 Martin Liska <mliska@suse.cz>
3944 Martin Jambor <mjambor@suse.cz>
3947 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3948 param_type is not an integral or pointer type.
3950 2018-04-03 Richard Biener <rguenther@suse.de>
3952 * sese.h (recompute_all_dominators): Remove.
3954 2018-04-02 Martin Sebor <msebor@redhat.com>
3956 * doc/invoke.texi (-Wrestrict): Fix typos.
3958 2018-04-02 Jim Wilson <jimw@sifive.com>
3960 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3961 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3962 (<optab>di3, <optab>si3_extend): Likewise.
3963 (<optab>si3_mask, <optab>si3_mask_1): New.
3964 (<optab>di3_mask, <optab>di3_mask_1): New.
3965 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3966 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3967 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3969 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3971 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3974 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3976 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3977 (nds32_canonicalize_comparison): New function.
3979 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3980 Kito Cheng <kito.cheng@gmail.com>
3981 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3983 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3984 * config/nds32/constants.md (unspec_volatile_element): Add
3985 UNSPEC_VOLATILE_RELAX_GROUP.
3986 * config/nds32/nds32-relax-opt.c: New file.
3987 * config/nds32/nds32-predicates.c
3988 (nds32_symbol_load_store_p): New function.
3989 * config/nds32/nds32-protos.h
3990 (nds32_symbol_load_store_p): Declare function.
3991 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3992 * config/nds32/nds32.c
3993 (nds32_register_pass): New function to register pass.
3994 (nds32_register_passes): New function to register passes.
3995 * config/nds32/nds32.md (relax_group): New pattern.
3996 * config/nds32/nds32.opt (mrelax-hint): New option.
3997 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3999 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4001 * config/nds32/t-nds32: Modify files dependency.
4003 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4005 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4006 (PROFILE_HOOK): Define its implementation.
4008 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4010 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4011 type and 32-bit size.
4013 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4016 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4017 (V_128_256): New mode iterator.
4018 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4019 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4020 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4022 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4023 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4025 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4028 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4029 NaN inputs correctly.
4031 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4034 * config/rs6000/vsx.md (??r): New mode attribute.
4035 (*vsx_mov<mode>_64bit): Use it.
4036 (*vsx_mov<mode>_32bit): Likewise.
4038 2018-03-30 Martin Sebor <msebor@redhat.com>
4040 PR tree-optimization/84818
4041 * builtins.c (check_access): Use warning_n.
4043 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4046 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4048 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4051 2018-03-30 Julia Koval <julia.koval@intel.com>
4054 * x86-tune.def (movx, partial_reg_dependency): Enable for
4057 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4060 * lra-constraints.c (process_alt_operands): Move setting
4061 this_alternative_matches below.
4063 2018-03-29 Martin Liska <mliska@suse.cz>
4066 * doc/invoke.texi: Document how LTO works with debug info.
4067 Describe auto-load support of binutils. Mention 'x86-64'
4068 as valid option value of -march option.
4070 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4072 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4075 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4076 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4077 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4080 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4083 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4084 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4085 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4086 for RS6000_BTM_POWERPC64.
4087 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4088 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4089 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4094 2018-03-28 Carl Love <cel@us.ibm.com>
4097 2017-09-27 Carl Love <cel@us.ibm.com>
4099 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4100 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4101 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4104 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4106 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4107 instead of __vector bool.
4108 (_mm_max_pu8): Likewise.
4109 (_mm_min_pi16): Likewise.
4111 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4114 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4115 (DIVWEUO): Likewise.
4117 (DIVDEUO): Likewise.
4118 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4119 DIVWEUO and DIVDEUO.
4120 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4121 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4122 (div_extend): Likewise.
4123 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4125 (__builtin_divweuo): Likewise.
4126 (__builtin_divdeo): Likewise.
4127 (__builtin_divdeuo): Likewise.
4129 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4132 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4133 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4135 PR tree-optimization/82004
4136 * gimple-match-head.c (optimize_pow_to_exp): New function.
4137 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4138 Don't fold to exp if optimize_pow_to_exp is false.
4140 2018-03-28 Martin Liska <mliska@suse.cz>
4143 * calls.c (initialize_argument_information): Fix trailing space.
4144 * common.opt: Fix typo and provide better explanation for
4145 -fsanitize-coverage option.
4146 * config/i386/i386.opt: Fix typo.
4148 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4149 Martin Liska <mliska@suse.cz>
4152 * gimplify.c (asan_poison_variable): Don't do the check for
4153 gimplify_omp_ctxp here.
4154 (gimplify_decl_expr): Do it here.
4155 (gimplify_target_expr): Likewise.
4157 2018-03-28 Martin Liska <mliska@suse.cz>
4160 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4161 chkp_type_bounds_count if MPX is not enabled.
4163 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4165 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4167 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4170 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4171 function to create the function decl for complex long double
4172 multiply and divide for -mabi=ieeelongdouble.
4173 (init_float128_ieee): Call it.
4175 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4178 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4179 -fcf-protection=branch -mibt.
4180 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4182 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4185 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4187 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4190 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4191 extern array declarations.
4193 2018-03-27 Richard Biener <rguenther@suse.de>
4196 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4197 explicit single_use checks.
4199 2018-03-27 Richard Biener <rguenther@suse.de>
4201 PR tree-optimization/85082
4202 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4205 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4207 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4208 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4209 Turn on fasynchronous-unwind-tables and funwind-tables.
4211 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4214 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4215 (*bmi_blsr_<mode>_ccz): Ditto.
4217 2018-03-26 Tom de Vries <tom@codesourcery.com>
4219 PR tree-optimization/85063
4220 * omp-general.c (offloading_function_p): New function. Factor out
4222 * omp-offload.c (pass_omp_target_link::gate): ... here.
4223 * omp-general.h (offloading_function_p): Declare.
4224 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4225 with attribute omp declare target for offloading functions.
4227 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4229 PR tree-optimization/84005
4230 * tree-data-ref.h (get_base_for_alignment): Declare.
4231 * tree-data-ref.c (get_base_for_alignment_1): New function.
4232 (get_base_for_alignment): Likewise.
4233 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4234 get_base_for_alignment to find a suitable base object, instead
4235 of always using drb->base_address.
4237 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4240 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4241 known size by default.
4243 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4246 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4247 and non BLKmode operands.
4249 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4252 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4253 Clean up attributes.
4255 2018-03-23 Richard Biener <rguenther@suse.de>
4258 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4259 we are going to emit early debug for LTO.
4261 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4264 * function.c (match_asm_constraints_1): Don't optimize if input
4265 doesn't satisfy general_operand predicate for output's mode.
4268 * alias.c (write_dependence_p): Don't require for x_canonicalized
4269 non-VOIDmode if x has VOIDmode.
4272 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4273 just don't try to optimize it rather than assert it never happens.
4275 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4277 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4278 macro expansions for definition of ST_INTERNAL_<mode> and
4279 LD_INTERNAL_<mode> builtins.
4280 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4282 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4284 (altivec_expand_st_builtin): Likewise.
4285 (altivec_expand_builtin): Remove calls to deleted functions.
4286 (rs6000_address_for_altivec): Delete this function.
4287 * config/rs6000/vector.md: Remove expands for
4288 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4290 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4293 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4294 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4295 re-computing once computed.
4296 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4297 (arm_init_machine_status): Initialize
4298 machine->static_chain_stack_bytes.
4300 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4303 * doc/extend.texi: Add four new prototypes for vec_ld.
4304 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4305 definitions for more logical presentation.
4306 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4307 entries for V1TI variants of __builtin_altivec_ld builtin.
4308 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4309 handling of V1TI variant of LVX icode pattern.
4310 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4311 (rs6000_gimple_fold_builtin): Likewise.
4312 (altivec_init_builtins): Add code to define
4313 __builtin_altivec_lvx_v1ti function.
4315 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4318 * function.c (match_asm_constraints_1): Don't do the optimization
4319 if input isn't a REG, SUBREG, MEM or constant.
4321 2018-03-22 Tom de Vries <tom@codesourcery.com>
4323 PR tree-optimization/84956
4324 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4325 bb_has_abnormal_pred.
4327 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4330 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4331 DECL_INITIAL (decl) to decl at the end.
4332 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4335 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4337 * doc/extend.texi (__builtin_tgmath): Document when complex
4338 integer types are treated as _Complex _Float64.
4340 2018-03-21 Tom de Vries <tom@codesourcery.com>
4342 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4344 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4346 PR tree-optimization/84960
4347 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4348 if it is ENTRY block, move them into single succ of ENTRY in that case.
4350 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4352 PR tree-optimization/84811
4353 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4354 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4355 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4356 (known_subrange_p): Update accordingly. Cast each value involved
4357 in the size comparison, rather than casting the result of the
4360 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4362 PR tree-optimization/84982
4363 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4364 by flipping the least significant bit rather than all bits from
4365 bitpos to bitpos + bitsize - 1.
4367 2018-03-21 Nathan Sidwell <nathan@acm.org>
4369 * doc/extend.texi (Deprecated Features): Remove mention of
4370 long-deleted deprecations.
4372 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4375 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4376 * configure: Regenerate.
4378 2018-03-21 Tom de Vries <tom@codesourcery.com>
4380 PR tree-optimization/83126
4381 * tree-parloops.c (num_phis): New function.
4382 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4384 2018-03-21 Nathan Sidwell <nathan@acm.org>
4386 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4387 mention anon-struct/union members and trailing attributes.
4389 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4391 PR tree-optimization/84969
4392 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4393 builtin memset partitions if they set different rhs values.
4395 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4397 PR rtl-optimization/84989
4398 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4399 VEC_DUPLICATE with scalar result mode.
4401 2018-03-21 Martin Liska <mliska@suse.cz>
4404 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4405 not intended return statement.
4407 2018-03-21 Martin Liska <mliska@suse.cz>
4410 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4411 (chkp_find_bound_slots_1): Limit number of iterations.
4413 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4416 * Minor grammar fixes for x86 options.
4418 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4421 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4422 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4424 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4427 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4428 (altivec_lvx_<mode>_1op): Likewise.
4429 (altivec_stvx_<mode>_2op): Likewise.
4430 (altivec_stvx_<mode>_1op): Likewise.
4431 (altivec_lvx_<VM2:mode>): New define_expand.
4432 (altivec_stvx_<VM2:mode>): Likewise.
4433 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4434 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4435 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4436 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4437 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4438 (rs6000_gen_lvx): Likewise.
4439 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4440 (altivec_expand_stv_builtin): Likewise.
4441 (altivec_expand_builtin): Likewise.
4442 * config/rs6000/vector.md: Likewise.
4444 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4447 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4450 2018-03-20 Richard Biener <rguenther@suse.de>
4453 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4454 sign-conversions as zero, fall back to standard scalar_stmt
4457 2018-03-20 Martin Liska <mliska@suse.cz>
4460 * predict.c (rebuild_frequencies): Handle case when we have
4461 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4463 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4466 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4467 flag_section_anchors.
4468 * varasm.c (use_blocks_for_decl_p): Remove hack for
4469 dw2_force_const_mem.
4472 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4474 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4475 be created, use lowpart_subreg of operands[0] rather than operands[0]
4477 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4478 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4479 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4480 and n constraint instead of aarch64_shift_imm_di and Usd.
4481 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4482 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4484 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4487 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4488 to favor GPR over NEON registers.
4489 (<shift>di3_neon): Likewise.
4491 2018-03-20 Tom de Vries <tom@codesourcery.com>
4494 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4495 (nvptx_process_pars): Emit bar.sync asap and alap.
4497 2018-03-20 Tom de Vries <tom@codesourcery.com>
4500 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4501 seen_label if seen_label is already set.
4503 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4506 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4507 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4508 Use 1U instead of 1. Formatting fixes.
4511 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4512 instead of TREE_TYPE (s1) for the return value.
4514 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4516 PR tree-optimization/84946
4517 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4518 bitsize + bitsize in poly_uint64 rather than poly_int64.
4521 * dwarf2asm.c: Include fold-const.c.
4522 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4523 of decl rather than decl itself.
4525 PR rtl-optimization/84643
4526 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4528 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4531 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4532 calling assemble_variable.
4534 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4537 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4538 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4540 2018-03-19 Jim Wilson <jimw@sifive.com>
4543 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4544 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4545 (riscv_first_stack_step): Likewise.
4546 (riscv_option_override): Use STACK_BOUNDARY instead of
4548 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4550 (BIGGEST_ALIGNMENT): Set to 128.
4551 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4552 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4555 2018-03-19 Richard Biener <rguenther@suse.de>
4557 PR tree-optimization/84933
4558 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4559 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4561 2018-03-19 Richard Biener <rguenther@suse.de>
4563 PR tree-optimization/84859
4564 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4565 (cond_if_else_store_replacement): Perform sinking operation on
4566 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4567 Generalize what a BB with a single eligible store is.
4569 2018-03-19 Richard Biener <rguenther@suse.de>
4571 PR tree-optimization/84929
4572 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4573 chrec_is_positive against non-chrec arg.
4575 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4578 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4580 2018-03-18 Martin Liska <mliska@suse.cz>
4582 PR rtl-optimization/84635
4583 * regrename.c (build_def_use): Use matches_mode only when
4586 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4588 PR tree-optimization/84913
4589 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4590 vectorize chains of COND_EXPRs.
4592 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4594 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4596 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4598 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4600 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4602 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4604 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4605 Kito Cheng <kito.cheng@gmail.com>
4607 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4608 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4609 (nds32_adjust_reg_alloc_order): New function.
4610 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4612 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4614 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4615 nds32_print_operand, nds32_print_operand_address): Use
4616 HOST_WIDE_INT_PRINT_DEC instead.
4618 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4620 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4622 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4625 * config/i386/i386.c (initial_ix86_tune_features,
4626 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4628 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4629 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4630 rather than 1u << ix86_tune. Formatting fix.
4631 (ix86_option_override_internal): Change ix86_arch_mask from
4632 unsigned int to unsigned HOST_WIDE_INT, initialize to
4633 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4634 (ix86_function_specific_restore): Likewise.
4636 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4639 * postreload.c (reload_combine_recognize_pattern): Perform
4640 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4641 truncate_int_for_mode the result for the destination's mode.
4644 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4645 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4647 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4650 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4653 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4654 regno_allocno_class_array and sorted_pseudos.
4655 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4656 insns where regno is used.
4658 2018-03-16 Martin Liska <mliska@suse.cz>
4661 * multiple_target.c (create_dispatcher_calls): Redirect
4662 reference in the symbol table.
4664 2018-03-16 Martin Liska <mliska@suse.cz>
4667 * multiple_target.c (create_dispatcher_calls): Redirect also
4670 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4674 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4675 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4676 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4678 2018-03-16 Julia Koval <julia.koval@intel.com>
4680 * doc/invoke.texi (Skylake Server): Add CLWB.
4681 Cannonlake): Remove CLWB.
4683 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4685 PR tree-optimization/84841
4686 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4688 (FLOAT_ONE_CONST_TYPE): Define.
4689 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4690 (sort_by_operand_rank): Put entries with higher constant_type last
4691 rather than first to match comments.
4693 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4695 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4698 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4701 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4702 check and use address of target always.
4704 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4707 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4708 (indirect_thunk_bnd_needed): Likewise.
4709 (indirect_thunks_used): Likewise.
4710 (indirect_thunks_bnd_used): Likewise.
4711 (indirect_return_needed): New.
4712 (indirect_return_bnd_needed): Likewise.
4713 (output_indirect_thunk_function): Add a bool argument for
4715 (output_indirect_thunk_function): Don't generate alias for
4716 function return thunk.
4717 (ix86_code_end): Call output_indirect_thunk_function to generate
4718 function return thunks.
4719 (ix86_output_function_return): Set indirect_return_bnd_needed
4720 and indirect_return_needed instead of indirect_thunk_bnd_needed
4721 and indirect_thunk_needed.
4723 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4725 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4726 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4727 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4729 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4730 Paul Hua <paul.hua.gm@gmail.com>
4733 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4735 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4737 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4738 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4741 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4744 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4745 instead of GET_MODE_SIZE when comparing Units.
4747 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4750 * varasm.c (hash_section): Return an unchangeble hash value
4751 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4752 Return !aarch64_can_use_per_function_literal_pools_p ().
4754 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4757 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4758 rather than address of cmode as last argument to prepare_cmp_insn.
4760 2018-03-15 Julia Koval <julia.koval@intel.com>
4762 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4763 F_AVX512VNNI, F_AVX512BITALG): New.
4765 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4768 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4769 insn for floating-point loads and stores.
4771 2018-03-14 Carl Love <cel@us.ibm.com>
4773 * config/rs6000/rs6000-c.c: Add macro definitions for
4774 ALTIVEC_BUILTIN_VEC_PERMXOR.
4775 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4776 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4777 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4778 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4780 * config/doc/extend.texi: Add prototypes for vec_permxor.
4782 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4785 * diagnostic-show-locus.c (class layout_point): Convert m_line
4786 from int to linenum_type.
4787 (line_span::comparator): Use linenum "compare" function when
4788 comparing line numbers.
4789 (test_line_span): New function.
4790 (layout_range::contains_point): Convert param "row" from int to
4792 (layout_range::intersects_line_p): Likewise.
4793 (layout::will_show_line_p): Likewise.
4794 (layout::print_source_line): Likewise.
4795 (layout::should_print_annotation_line_p): Likewise.
4796 (layout::print_annotation_line): Likewise.
4797 (layout::print_leading_fixits): Likewise.
4798 (layout::annotation_line_showed_range_p): Likewise.
4799 (struct line_corrections): Likewise for field m_row.
4800 (line_corrections::line_corrections): Likewise for param "row".
4801 (layout::print_trailing_fixits): Likewise.
4802 (layout::get_state_at_point): Likewise.
4803 (layout::get_x_bound_for_row): Likewise.
4804 (layout::print_line): Likewise.
4805 (diagnostic_show_locus): Likewise for locals "last_line" and
4807 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4808 * input.c (selftest::test_linenum_comparisons): New function.
4809 (selftest::input_c_tests): Call it.
4810 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4811 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4812 * selftest.h (ASSERT_GT): New macro.
4813 (ASSERT_GT_AT): New macro.
4814 (ASSERT_LT): New macro.
4815 (ASSERT_LT_AT): New macro.
4817 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4819 PR rtl-optimization/84780
4820 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4822 2018-03-14 Martin Liska <mliska@suse.cz>
4824 * tree.c (record_node_allocation_statistics): Use
4825 get_stats_node_kind.
4826 (get_stats_node_kind): New function extracted from
4827 record_node_allocation_statistics.
4828 (free_node): Use get_stats_node_kind.
4830 2018-03-14 Richard Biener <rguenther@suse.de>
4832 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4833 that the value-set of ANTIC_IN doesn't grow.
4836 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4838 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4839 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4841 2018-03-14 Julia Koval <julia.koval@intel.com>
4843 * config.gcc (icelake-client, icelake-server): New.
4845 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4846 (initial_ix86_arch_features): Ditto.
4847 (PTA_SKYLAKE): Add SGX.
4848 (PTA_ICELAKE): Remove.
4849 (PTA_ICELAKE_CLIENT): New.
4850 (PTA_ICELAKE_SERVER): New.
4851 (ix86_option_override_internal): Split up icelake on icelake client and
4853 (get_builtin_code_for_version): Ditto.
4854 (fold_builtin_cpu): Ditto.
4855 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4856 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4857 * config/i386/i386.h (processor_type): Ditto.
4858 * doc/invoke.texi: Ditto.
4860 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4863 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4864 INTEGER_CST offset, add it together with bitpos / 8 and
4865 sign extend based on POINTER_SIZE.
4869 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4872 * config/i386/constraints.md (Yc): New register constraint.
4873 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4874 Use Yc constraint for alternative 2 of operand 0. Remove
4875 preferred_for_speed attribute.
4877 2018-03-14 Richard Biener <rguenther@suse.de>
4879 PR tree-optimization/84830
4880 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4881 with the old one to avoid oscillations.
4883 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4886 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4888 (assign_by_spills): Return a flag of reload assignment failure.
4889 Do not process the reload assignment failures. Do not spill other
4890 reload pseudos if they has the same reg class. Update n if
4892 (lra_assign): Add a return arg. Set up from the result of
4893 assign_by_spills call.
4894 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4895 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4896 usage_insns if it is not NULL.
4897 (spill_hard_reg_in_range): New function.
4898 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4899 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4900 function prototypes.
4901 (lra_assign): Change prototype.
4902 * lra.c (lra): Add code to deal with fails by splitting hard reg
4905 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4907 * config/riscv/riscv.opt (mrelax): New option.
4908 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4909 "norelax" when riscv_mrelax is disabled.
4910 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4912 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4915 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4916 reassociation for int modes.
4918 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4920 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4921 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4923 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4924 * config/aarch64/aarch64-sve.md
4925 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4926 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4927 (*extend<mode><Vwide>2): Rename to...
4928 (aarch64_sve_extend<mode><Vwide>2): ...this.
4929 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4930 renaming the old pattern to...
4931 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4933 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4934 define_expand, renaming the old pattern to...
4935 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4936 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4937 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4938 account when deciding which SVE instruction the optab should use.
4939 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4941 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4943 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4944 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4945 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4946 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4947 (tlsdesc_small_<mode>): Turn a define_expand and use
4948 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4949 (tlsdesc_small_advsimd_<mode>): ...this.
4950 (tlsdesc_small_sve_<mode>): New pattern.
4952 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4954 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4955 (UNSPEC_UMUL_HIGHPART): New constants.
4956 (MUL_HIGHPART): New int iteraor.
4957 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4958 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4960 (*<su>mul<mode>3_highpart): New define_insn.
4962 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4965 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4968 2018-03-13 Martin Liska <mliska@suse.cz>
4971 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4973 (sem_item_optimizer::~sem_item_optimizer): Release it.
4974 (sem_item_optimizer::merge_classes): Register variable aliases.
4975 (sem_item_optimizer::fixup_pt_set): New function.
4976 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4977 * ipa-icf.h: Declare new variables and functions.
4979 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4982 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4983 integer_pow2p@2 and test integer_pow2p in condition.
4984 (A < 0 ? C : 0): Similarly for @1.
4987 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4988 characters starting at p contain '\0' character, don't look beyond
4992 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4993 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4996 * reg-stack.c (change_stack): Change update_end var from int to
4997 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4998 also call set_block_for_insn on the newly added insns and rescan.
5001 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5002 on the last operand.
5005 * tree.c (stabilize_reference_1): Return save_expr (e) for
5006 STATEMENT_LIST even if it doesn't have side-effects.
5008 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5010 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5012 2018-03-12 Renlin Li <renlin.li@arm.com>
5014 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5015 aarch64_output_scalar_simd_mov_immediate.
5017 2018-03-12 Martin Sebor <msebor@redhat.com>
5019 PR tree-optimization/83456
5020 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5021 for perfectly overlapping calls to memcpy.
5022 (gimple_fold_builtin_memory_chk): Same.
5023 (gimple_fold_builtin_strcpy): Handle no-warning.
5024 (gimple_fold_builtin_stxcpy_chk): Same.
5025 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5027 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5029 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5030 parameter. Use it for SFmode.
5031 (rs6000_function_arg_advance_1): Adjust.
5032 (rs6000_function_arg): Adjust.
5033 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5035 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5037 PR rtl-optimization/84169
5038 PR rtl-optimization/84780
5039 * combine.c (can_combine_p): Check for a 2-insn combination whether
5040 the destination register is used between the two insns, too.
5042 2018-03-12 Richard Biener <rguenther@suse.de>
5044 PR tree-optimization/84803
5045 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5046 for refs DR analysis didn't process.
5048 2018-03-12 Richard Biener <rguenther@suse.de>
5050 PR tree-optimization/84777
5051 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5052 force-vectorize loops ignore whether we are optimizing for size.
5054 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5056 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5057 (TARGET_MD_ASM_ADJUST): Define.
5059 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5060 Kito Cheng <kito.cheng@gmail.com>
5061 Chung-Ju Wu <jasonwucj@gmail.com>
5063 * config/nds32/nds32.c (nds32_compute_stack_frame,
5064 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5065 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5066 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5067 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5068 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5069 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5070 * config/nds32/nds32.md (prologue, epilogue): Use macro
5071 NDS32_V3PUSH_AVAILABLE_P to do checking.
5073 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5076 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5077 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5078 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5079 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5080 addition of most attributes on !orig_type_die or the attribute not
5081 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5083 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5084 Chung-Ju Wu <jasonwucj@gmail.com>
5086 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5088 * config/nds32/nds32.opt (mvh): New option.
5090 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5091 Chung-Ju Wu <jasonwucj@gmail.com>
5093 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5095 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5096 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5099 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5100 Chung-Ju Wu <jasonwucj@gmail.com>
5102 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5104 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5105 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5107 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5108 Kito Cheng <kito.cheng@gmail.com>
5109 Chung-Ju Wu <jasonwucj@gmail.com>
5111 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5112 UNSPEC_FFMISM and UNSPEC_FLMISM.
5113 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5114 for ffb, ffmism and flmism.
5115 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5116 (unspec_ffmism): Ditto.
5117 (unspec_flmism): Ditto.
5118 (nds32_expand_builtin_impl): Check if string extension is available.
5119 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5120 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5122 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5125 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5128 * lra-assigns.c (assign_by_spills): Return a flag of reload
5129 assignment failure. Do not process the reload assignment
5130 failures. Do not spill other reload pseudos if they has the same
5132 (lra_assign): Add a return arg. Set up from the result of
5133 assign_by_spills call.
5134 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5135 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5136 usage_insns if it is not NULL.
5137 (spill_hard_reg_in_range): New function.
5138 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5139 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5140 function prototypes.
5141 (lra_assign): Change prototype.
5142 * lra.c (lra): Add code to deal with fails by splitting hard reg
5145 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5148 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5150 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5153 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5154 (dw_val_node): Add val_symbolic_view.
5155 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5156 (symview_upper_bound): New.
5157 (new_line_info_table): Initialize symviews_since_reset.
5158 (dwarf2out_source_line): Count symviews_since_reset and set
5159 symview_upper_bound.
5160 (dw_val_equal_p): Handle symview.
5161 (add_AT_symview): New.
5162 (print_dw_val): Handle symview.
5163 (attr_checksum, attr_checksum_ordered): Likewise.
5164 (same_dw_val_p, size_of_die): Likewise.
5165 (value_format, output_die): Likewise.
5166 (add_high_low_attributes): Use add_AT_symview for entry_view.
5167 (dwarf2out_finish): Reset symview_upper_bound, clear
5170 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5173 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5174 Add strict argument and use it.
5175 (rs6000_split_multireg_move): Update for new strict argument.
5176 (mem_operand_gpr): Disallow all non-offsettable addresses.
5177 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5179 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5182 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5183 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5184 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5187 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5188 decl, use remap_type if we want to use the type.
5190 2018-03-09 Martin Sebor <msebor@redhat.com>
5192 PR tree-optimization/84526
5193 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5195 (builtin_access::generic_overlap): Be prepared to handle non-array
5198 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5200 PR rtl-optimization/84682
5201 * lra-constraints.c (process_address_1): Check is_address flag
5202 for address constraints.
5203 (process_alt_operands): Likewise.
5204 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5205 preprocess_constraints.
5206 * recog.h (preprocess_constraints): Add oploc parameter.
5208 * recog.c (preprocess_constraints): Test address_operand for
5209 CT_ADDRESS constraints.
5211 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5214 * lra-assigns.c (assign_by_spills): Return a flag of reload
5215 assignment failure. Do not process the reload assignment
5216 failures. Do not spill other reload pseudos if they has the same
5218 (lra_assign): Add a return arg. Set up from the result of
5219 assign_by_spills call.
5220 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5221 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5222 usage_insns if it is not NULL.
5223 (spill_hard_reg_in_range): New function.
5224 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5225 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5226 function prototypes.
5227 (lra_assign): Change prototype.
5228 * lra.c (lra): Add code to deal with fails by splitting hard reg
5231 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5234 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5235 Accept complain bool parameter. Only emit errors if it is true.
5236 (arm_parse_cpu_option_name): Likewise.
5237 (arm_target_thumb_only): Adjust callers of the above.
5238 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5239 prototype to take a default true bool parameter.
5240 (arm_parse_arch_option_name): Likewise.
5242 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5243 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5247 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5248 * configure: Regenerate.
5249 * configure.ac ("linker --version-script option"): New.
5250 ("linker soname option"): New.
5252 2018-03-09 Richard Biener <rguenther@suse.de>
5254 PR tree-optimization/84775
5255 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5256 immediate uses of predicate stmts and mark them modified.
5259 PR tree-optimization/84178
5260 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5262 (version_loop_for_if_conversion): Delay update_ssa call.
5263 (tree_if_conversion): Delay update_ssa until after predicate
5266 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5269 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5270 when the function accesses prior frames.
5272 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5275 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5276 gen_llsym, otherwise call maybe_gen_llsym.
5279 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5280 has ',' character inside of it.
5282 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5285 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5286 as clobbering CC_REGNUM.
5288 2018-03-08 Richard Biener <rguenther@suse.de>
5291 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5292 (follow_copies_to_constant): Do not follow SSA names registered
5295 2018-03-08 Richard Biener <rguenther@suse.de>
5297 PR tree-optimization/84178
5298 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5300 (version_loop_for_if_conversion): Delay update_ssa call.
5301 (tree_if_conversion): Delay update_ssa until after predicate
5304 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5306 PR tree-optimization/84178
5307 * tree-if-conv.c (release_bb_predicate): Remove the
5308 the assertion that the stmts have NULL use_ops.
5309 Discard the statements, asserting that they haven't
5310 yet been added to a BB.
5312 2018-03-08 Richard Biener <rguenther@suse.de>
5314 PR tree-optimization/84746
5315 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5316 (phi_translate): Pass in destination ANTIC_OUT set.
5317 (phi_translate_1): Likewise. For a simplified result lookup
5318 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5319 (phi_translate_set): Adjust.
5320 (do_pre_regular_insertion): Likewise.
5321 (do_pre_partial_partial_insertion): Likewise.
5323 2018-03-08 Martin Liska <mliska@suse.cz>
5325 PR gcov-profile/84735
5326 * doc/gcov.texi: Document usage of profile files.
5327 * gcov-io.h: Document changes in the format.
5329 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5333 * dwarf2out.c (struct dw_line_info_table): Update comments for
5335 (FORCE_RESET_NEXT_VIEW): New.
5336 (FORCE_RESETTING_VIEW_P): New.
5337 (RESETTING_VIEW_P): Check for -1 too.
5338 (ZERO_VIEW_P): Likewise.
5339 (new_line_info_table): Force-reset next view.
5340 (dwarf2out_begin_function): Likewise.
5341 (dwarf2out_source_line): Simplify zero_view_p initialization.
5342 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5343 view directly. Omit view when omitting .loc at line 0.
5345 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5347 PR tree-optimization/84740
5348 * tree-switch-conversion.c (process_switch): Call build_constructors
5349 only if info.phi_count is non-zero.
5351 PR tree-optimization/84739
5352 * tree-tailcall.c (find_tail_calls): Check call arguments against
5353 DECL_ARGUMENTS (current_function_decl) rather than
5354 DECL_ARGUMENTS (func) when checking for tail recursion.
5356 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5358 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5359 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5360 Volker Reichelt's entry and add entries for people that perform
5361 GCC fuzzy testing and report numerous bugs.
5363 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5366 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5367 readonly data in sdata, if that is disabled.
5368 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5369 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5370 -mreadonly-in-sdata option.
5372 2018-03-07 Martin Sebor <msebor@redhat.com>
5374 PR tree-optimization/84468
5375 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5376 basic block when looking for nul assignment.
5378 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5381 * except.h (output_function_exception_table): Adjust prototype.
5382 * except.c (output_function_exception_table): Remove FNNAME parameter
5383 and add SECTION parameter. Ouput one part of the table at a time.
5384 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5385 the first part of the exception table and emit unwind directives.
5386 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5387 (i386_pe_seh_cold_init): Likewise.
5388 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5389 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5390 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5391 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5392 * config/i386/winnt.c: Include except.h.
5393 (struct seh_frame_state): Add reg_offset, after_prologue and
5394 in_cold_section fields.
5395 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5396 (i386_pe_seh_cold_init): New function.
5397 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5398 to seh->in_cold_section.
5399 (seh_emit_push): Record the offset of the push.
5400 (seh_emit_save): Record the offet of the save.
5401 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5402 Test seh->after_prologue to disregard the epilogue.
5403 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5404 (i386_pe_end_cold_function): New function.
5406 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5409 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5410 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5413 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5415 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5416 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5419 * multiple_target.c: Include tree-inline.h and intl.h.
5420 (expand_target_clones): Diagnose and fail if node->definition and
5421 !tree_versionable_function_p (node->decl).
5423 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5425 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5427 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5428 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5429 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5431 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5434 * combine.c (try_combine): Use reg_or_subregno instead of handling
5435 just paradoxical SUBREGs and REGs.
5437 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5439 * config/arc/arc.c (arc_finalize_pic): Remove function.
5440 (arc_must_save_register): We use single base PIC register, remove
5441 checks to save/restore the PIC register.
5442 (arc_expand_prologue): Likewise.
5443 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5445 (arc_verify_short): Likewise.
5446 (arc_attr_type): Likewise.
5447 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5448 (walk_stores): Likewise.
5449 (arc_address_cost): Make it static.
5450 (arc_verify_short): Likewise.
5451 (branch_dest): Likewise.
5452 (arc_attr_type): Likewise.
5453 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5454 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5455 (arc_final_prescan_insn): Remove inserting the nops due to
5456 hardware hazards. It is done in reorg step.
5457 (insn_length_variant_t): Remove.
5458 (insn_length_parameters_t): Likewise.
5459 (arc_insn_length_parameters): Likewise.
5460 (arc_get_insn_variants): Likewise.
5461 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5463 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5466 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5469 PR tree-optimization/84687
5470 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5472 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5474 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5476 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5477 Rename to ppc_speculation_barrier.
5478 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5479 __builtin_ppc_speculation_barrier.
5481 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5484 * combine.c (combine_simplify_rtx): Don't try to simplify if
5485 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5488 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5490 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5491 to 32 bytes when compiling for POWER9.
5493 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5496 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5497 regparm >= 3 with no arg reg available also for calls with
5498 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5501 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5503 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5505 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5508 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5510 2018-03-05 Richard Biener <rguenther@suse.de>
5512 PR tree-optimization/84486
5513 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5514 When inserting a __builtin_assume_aligned call set the LHS
5515 SSA name alignment info accordingly.
5517 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5519 PR tree-optimization/84114
5520 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5521 Avoid reassociation of FLOAT_MODE addition.
5523 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5525 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5526 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5527 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5528 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5529 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5530 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5531 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5533 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5534 __builtin_ia32_wbinvd): New builtins.
5535 (SPECIAL_ARGS2): New.
5536 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5537 (SPECIAL_ARGS2): New.
5538 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5539 (ix86_valid_target_attribute_inner_p): Ditto.
5540 (ix86_init_mmx_sse_builtins): Add special_args2.
5541 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5542 TARGET_WBNOINVD_P): New.
5543 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5544 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5545 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5546 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5547 * config/i386/pconfigintrin.h: New file.
5548 * config/i386/wbnoinvdintrin.h: Ditto.
5549 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5550 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5552 2018-03-05 Richard Biener <rguenther@suse.de>
5554 PR tree-optimization/84670
5555 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5557 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5558 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5559 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5560 doesn't grow if all successors (recursively) were visited at least
5563 2018-03-05 Richard Biener <rguenther@suse.de>
5565 PR tree-optimization/84650
5566 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5567 if executed in the loop pipeline.
5569 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5571 * doc/configfiles.texi (Configuration Files): Move info about
5572 conditionalizing $target-protos.h to...
5573 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5574 differs from $target-protos.h.
5576 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5577 Chung-Ju Wu <jasonwucj@gmail.com>
5579 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5580 * config/nds32/nds32-multiple.md (setmemsi): Define.
5581 * config/nds32/nds32-memory-manipulation.c
5582 (nds32_gen_dup_4_byte_to_word_value): New.
5583 (emit_setmem_word_loop): New.
5584 (emit_setmem_byte_loop): New.
5585 (nds32_expand_setmem_loop): New.
5586 (nds32_expand_setmem_loop_v3m): New.
5587 (nds32_expand_setmem_unroll): New.
5588 (nds32_expand_setmem): New.
5590 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5591 Chung-Ju Wu <jasonwucj@gmail.com>
5593 * config/nds32/nds32-memory-manipulation.c
5594 (nds32_emit_load_store): New.
5595 (nds32_emit_post_inc_load_store): New.
5596 (nds32_emit_mem_move): New.
5597 (nds32_emit_mem_move_block): New.
5598 (nds32_expand_movmemsi_loop_unknown_size): New.
5599 (nds32_expand_movmemsi_loop_known_size): New.
5600 (nds32_expand_movmemsi_loop): New.
5601 (nds32_expand_movmemsi_unroll): New.
5602 (nds32_expand_movmemqi): Rename ...
5603 (nds32_expand_movmemsi): ... to this.
5604 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5605 (movmemsi): ... to this.
5606 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5607 (nds32_expand_movmemsi): ... to this.
5609 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5610 Monk Chiang <sh.chiang04@gmail.com>
5611 Chung-Ju Wu <jasonwucj@gmail.com>
5613 * config/nds32/nds32-protos.h
5614 (nds32_expand_load_multiple): New arguments.
5615 (nds32_expand_store_multiple): Ditto.
5616 (nds32_valid_multiple_load_store): Rename ...
5617 (nds32_valid_multiple_load_store_p): ... to this.
5618 * config/nds32/nds32-memory-manipulation.c
5619 (nds32_expand_load_multiple): Refine implementation.
5620 (nds32_expand_store_multiple): Ditto.
5621 * config/nds32/nds32-multiple.md
5622 (load_multiple): Update nds32_expand_load_multiple interface.
5623 (store_multiple): Update nds32_expand_store_multiple interface.
5624 * config/nds32/nds32-predicates.c
5625 (nds32_valid_multiple_load_store): Rename ...
5626 (nds32_valid_multiple_load_store_p): ... to this and refine
5628 * config/nds32/predicates.md
5629 (nds32_load_multiple_and_update_address_operation): New predicate.
5630 (nds32_store_multiple_and_update_address_operation): New predicate.
5632 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5633 Chung-Ju Wu <jasonwucj@gmail.com>
5635 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5636 (combo): New attribute.
5637 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5639 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5641 * config/nds32/nds32.opt: Change -mcmodel= default value.
5643 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5644 Monk Chiang <sh.chiang04@gmail.com>
5645 Chung-Ju Wu <jasonwucj@gmail.com>
5647 * config/nds32/constants.md (unspec_element): New enum.
5648 * config/nds32/constraints.md (Umw): New constraint.
5649 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5650 * config/nds32/nds32-intrinsic.md: Likewise.
5651 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5652 (nds32_valid_smw_lwm_base_p): New.
5653 (nds32_output_smw_single_word): New.
5654 (nds32_output_lmw_single_word): New.
5655 (nds32_expand_unaligned_load): New.
5656 (nds32_expand_unaligned_store): New.
5657 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5658 (nds32_output_smw_single_word): Declare.
5659 (nds32_output_lmw_single_word): Declare.
5660 (nds32_expand_unaligned_load): Declare.
5661 (nds32_expand_unaligned_store): Declare.
5662 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5663 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5664 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5665 NDS32_BUILTIN_UASTORE_DW.
5666 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5669 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5670 Kito Cheng <kito.cheng@gmail.com>
5671 Chung-Ju Wu <jasonwucj@gmail.com>
5673 * config/nds32/nds32-intrinsic.c
5674 (nds32_expand_builtin_null_ftype_reg): Delete.
5675 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5676 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5677 (nds32_read_argument): New.
5678 (nds32_legitimize_target): Ditto.
5679 (nds32_legitimize_argument): Ditto.
5680 (nds32_check_constant_argument): Ditto.
5681 (nds32_expand_unop_builtin): Ditto.
5682 (nds32_expand_unopimm_builtin): Ditto.
5683 (nds32_expand_binop_builtin): Ditto.
5684 (nds32_builtin_decl_impl): Ditto.
5685 (builtin_description): Ditto.
5686 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5687 (nds32_init_builtins_impl): Ditto.
5688 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5689 (nds32_builtin_decl): New.
5690 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5691 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5693 2018-03-02 Jeff Law <law@redhat.com>
5695 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5696 (redundant_insn, fill_simple_delay_slots): Likewise.
5697 (fill_slots_from_thread): Likewise.
5698 * resource.c (mark_referenced_resources): Likewise.
5699 (mark_set_resources, find_dead_or_set_registers): Likewise.
5701 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5703 * substring-locations.h (format_warning_va): Formatting fix for
5705 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5707 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5708 * substring-locations.c: Include intl.h.
5709 (format_warning_va): Turned into small wrapper around
5710 format_warning_n_va, renamed to ...
5711 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5712 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5714 (format_warning_at_substring_n): New function.
5715 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5716 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5717 format_warning_at_substring with just a shorter name instead of
5718 const function pointer.
5719 (fmtwarn_n): New function.
5720 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5721 appropriate, get rid of all the fmtstr temporaries, move conditionals
5722 with G_() wrapped string literals directly into fmtwarn arguments,
5723 cast dir.len to (int), formatting fixes.
5725 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5727 * doc/invoke.texi: Remove "Cilk Plus" references.
5729 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5730 Richard Biener <rguenther@suse.de>
5733 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5734 for error or warning attributes if CALL_FROM_THUNK_P is set.
5737 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5740 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5741 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5744 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5745 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5747 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5748 instead of -1U in last predictors element's probability member.
5750 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5753 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5754 arguments if they are comparable.
5756 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5758 PR tree-optimization/84634
5759 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5760 masks and masked_loop_p with a single loop_masks, making sure it's
5761 null for bb vectorization.
5763 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5765 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5766 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5767 loop->force_vectorize to check whether there is no alias.
5769 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5772 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5774 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5776 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5777 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5778 instead of a loop around prev_real_insn.
5779 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5783 * config/i386/i386.c (ix86_print_operand): Use conditional
5784 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5787 2018-03-02 Richard Biener <rguenther@suse.de>
5789 PR tree-optimization/84427
5790 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5791 (bitmap_set_subtract_values): Rewrite to handle multiple
5794 (prune_clobbered_mems): Likewise.
5795 (phi_translate): Take edge instead of pred/phiblock.
5796 (phi_translate_1): Likewise.
5797 (phi_translate_set): Likewise. Insert all translated
5798 exprs for a value into the set, keeping possibly multiple
5799 expressions per value.
5800 (compute_antic_aux): Adjust for phi_translate changes.
5801 When intersecting union the expressions and prune those
5802 not in the final value set, keeping possibly multiple
5803 expressions per value. Do not use value-insertion
5804 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5805 all expressions. Add verification that the value-sets
5806 only shrink during iteration.
5807 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5808 (do_pre_regular_insertion): Likewise.
5809 (do_pre_partial_partial_insertion): Likewise.
5811 2018-03-02 Richard Biener <rguenther@suse.de>
5814 * config/darwin.c (saved_debug_info_level): New static global.
5815 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5816 (darwin_asm_lto_end): Restore debug info generation settings.
5818 2018-03-01 Martin Liska <mliska@suse.cz>
5821 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5824 2018-03-01 Richard Biener <rguenther@suse.de>
5827 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5828 type annotation with LTO when debug was disabled at compile-time.
5830 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5832 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5834 (mips_final_postscan_insn): Likewise.
5836 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5838 PR rtl-optimization/84528
5839 * alias.c (init_alias_target): Add commentary.
5840 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5841 a unique base value if the frame pointer is not eliminated
5842 to the stack pointer.
5844 2018-03-01 Tom de Vries <tom@codesourcery.com>
5846 PR rtl-optimization/83327
5847 * lra-int.h (hard_regs_spilled_into): Declare.
5848 * lra.c (hard_regs_spilled_into): Define.
5849 (init_reg_info): Init hard_regs_spilled_into.
5850 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5851 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5852 (process_bb_lives): Handle hard_regs_spilled_into.
5853 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5854 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5856 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5858 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5859 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5860 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5861 * config/rs6000/aix72.h: New file.
5863 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5865 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5866 instead of warning_at with conditional singular and plural messages
5870 * stor-layout.c (update_alignment_for_field): For
5871 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5872 && !DECL_PACKED (field), do the alignment update, just use
5873 only desired_align instead of MAX (type_align, desired_align)
5875 (place_field): Don't do known_align < desired_align handling
5876 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5877 is non-NULL, instead do it after rli->prev_field handling and
5878 only if not within a bitfield word. For DECL_PACKED (field)
5879 use type_align of BITS_PER_UNIT.
5881 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5883 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5884 superfluous parentheses and trailing spaces.
5886 2018-02-28 Richard Biener <rguenther@suse.de>
5888 PR tree-optimization/84584
5889 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5890 SCoPs with fake exit edge.
5892 2018-02-28 Martin Liska <mliska@suse.cz>
5895 * timevar.c (timer::print): Fix format to properly print 100%
5898 2018-02-28 Richard Biener <rguenther@suse.de>
5901 * genmatch.c (capture_info::walk_match): Do not mark
5902 captured expressions without operands as expr_p given
5903 they act more like predicates and should be subject to
5904 "lost tail" side-effect preserving.
5906 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5908 PR rtl-optimization/81611
5909 * auto-inc-dec.c (attempt_change): Move dead note from
5910 mem_insn if it's the next use of regno
5911 (find_address): Take address use of reg holding
5912 non-incremented value. Add parm to limit search to the named
5914 (merge_in_block): Attempt to use a mem insn that is the next
5915 use of the original regno.
5917 2018-02-27 Martin Sebor <msebor@redhat.com>
5920 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5921 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5923 2018-02-27 Martin Sebor <msebor@redhat.com>
5925 PR translation/84207
5926 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5927 n argument to unsigned HOST_WIDE_INT.
5928 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5929 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5930 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5931 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5933 2018-02-27 Richard Biener <rguenther@suse.de>
5935 PR tree-optimization/84512
5936 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5937 Do not use the estimate returned from record_stmt_cost for
5938 the scalar iteration cost but sum properly using add_stmt_cost.
5940 2018-02-27 Richard Biener <rguenther@suse.de>
5942 PR tree-optimization/84466
5943 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5944 Adjust last change to less strictly validate use operands.
5946 2018-02-27 Martin Liska <mliska@suse.cz>
5948 PR gcov-profile/84548
5949 * gcov.c (process_file): Allow partial overlap and consider it
5950 also as group functions.
5951 (output_lines): Properly calculate range of lines for a group.
5953 2018-02-27 Martin Liska <mliska@suse.cz>
5955 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5956 'ggc' suffixes. Change first column width.
5957 (timer::print): Fix formatting of the column.
5959 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5961 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5962 preserve inline entry blocks for the sake of debug inline
5963 entry point markers alone.
5964 (remove_unused_locals): Suggest in comments a better place to
5965 force the preservation of inline entry blocks that are
5966 otherwise unused, but do not preserve them.
5968 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5970 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5972 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5975 * config/i386/constraints.md (Bs): Replace
5976 ix86_indirect_branch_register with
5977 TARGET_INDIRECT_BRANCH_REGISTER.
5979 * config/i386/i386.md (indirect_jump): Likewise.
5980 (tablejump): Likewise.
5981 (*sibcall_memory): Likewise.
5982 (*sibcall_value_memory): Likewise.
5983 Peepholes of indirect call and jump via memory: Likewise.
5984 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5985 (*sibcall_value_GOT_32): Likewise.
5986 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5987 (GOT_memory_operand): Likewise.
5988 (call_insn_operand): Likewise.
5989 (sibcall_insn_operand): Likewise.
5990 (GOT32_symbol_operand): Likewise.
5991 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5993 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5995 PR rtl-optimization/83496
5996 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5997 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5998 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5999 redundant insn, if any.
6000 (relax_delay_slots): Likewise.
6001 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6003 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6005 PR tree-optimization/83965
6006 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6007 that grouped statements are part of a reduction chain. Return
6008 true if the statement is not marked as a reduction itself but
6010 (vect_recog_dot_prod_pattern): Don't check whether the statement
6011 is part of a group here.
6012 (vect_recog_sad_pattern): Likewise.
6013 (vect_recog_widen_sum_pattern): Likewise.
6015 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6018 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6020 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6023 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6025 (ix86_output_indirect_function_return): New prototype.
6026 (ix86_split_simple_return_pop_internal): Likewise.
6027 * config/i386/i386.c (indirect_return_via_cx): New.
6028 (indirect_return_via_cx_bnd): Likewise.
6029 (indirect_thunk_name): Handle return va CX_REG.
6030 (output_indirect_thunk_function): Create alias for
6031 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6032 (ix86_output_indirect_jmp): Remove the bool argument.
6033 (ix86_output_indirect_function_return): New function.
6034 (ix86_split_simple_return_pop_internal): Likewise.
6035 * config/i386/i386.md (*indirect_jump): Don't pass false
6036 to ix86_output_indirect_jmp.
6037 (*tablejump_1): Likewise.
6038 (simple_return_pop_internal): Change it to define_insn_and_split.
6039 Call ix86_split_simple_return_pop_internal to split it for
6041 (simple_return_indirect_internal): Call
6042 ix86_output_indirect_function_return instead of
6043 ix86_output_indirect_jmp.
6045 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6048 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6049 memset and value initialization afterwards.
6051 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6053 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6055 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6058 * common/config/aarch64/aarch64-common.c
6059 (aarch_option_optimization_table[]): Switch
6060 off fomit-frame-pointer
6062 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6063 Chung-Ju Wu <jasonwucj@gmail.com>
6065 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6067 (store_multiple): Ditto.
6069 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6071 * config.gcc: Add --with-cpu support for nds32 target.
6072 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6073 * config/nds32/nds32.opt: Add -mcpu= option.
6075 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6077 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6078 isel=yes): Warn for these deprecated options.
6080 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6082 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6083 ISA_2_5_MASKS_EMBEDDED.
6085 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6087 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6088 p->max as pointers rather than using iterative_hash_expr.
6090 2018-02-23 Carl Love <cel@us.ibm.com>
6092 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6093 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6095 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6096 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6097 P8V_BUILTIN_VEC_VUNSIGNED2.
6099 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6102 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6103 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6105 * lra-constraints.c (curr_insn_transform): Set up
6106 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6108 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6109 * lra-eliminations.c (spill_pseudos): Ditto.
6110 (process_insn_for_elimination): Ditto.
6111 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6112 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6113 LRA_NON_CLOBBERED_ALT.
6115 2018-02-22 Martin Sebor <msebor@redhat.com>
6117 PR tree-optimization/84480
6118 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6119 to maybe_diag_stxncpy_trunc. Call it.
6120 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6121 from gimple_fold_builtin_strcpy. Print inlining stack.
6122 (handle_builtin_stxncpy): Print inlining stack.
6123 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6125 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6128 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6129 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6130 and -fcheck-pointer-bounds are used together.
6131 (indirect_thunk_prefix): New enum.
6132 (indirect_thunk_need_prefix): New function.
6133 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6134 "_nt" instead of "_bnd" for NOTRACK prefix.
6135 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6136 (output_indirect_thunk_function): Likewise.
6138 (ix86_code_end): Update output_indirect_thunk_function calls.
6139 (ix86_output_indirect_branch_via_reg): Replace
6140 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6141 (ix86_output_indirect_branch_via_push): Likewise.
6142 (ix86_output_function_return): Likewise.
6143 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6144 incompatible with -fcf-protection=branch and
6145 -fcheck-pointer-bounds.
6147 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6150 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6151 Change gcc_assert call to output_operand_lossage.
6153 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6155 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6157 2018-02-22 DJ Delorie <dj@redhat.com>
6158 Sebastian Perta <sebastian.perta@renesas.com>
6159 Oleg Endo <olegendo@gcc.gnu.org>
6161 * config/rx/rx.c (rx_rtx_costs): New function.
6162 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6164 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6166 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6168 2018-02-22 Martin Liska <mliska@suse.cz>
6171 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6172 Add "native" as a possible value.
6174 2018-02-22 Martin Liska <mliska@suse.cz>
6177 * config/i386/i386.c (ix86_option_override_internal):
6178 Add "native" as a possible value for -march and -mtune.
6180 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6183 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6184 to all type variants.
6186 PR tree-optimization/84503
6187 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6188 width as info->bitpos + info->bitsize - start.
6189 (merged_store_group::merge_overlapping): Simplify width computation.
6190 (check_no_overlap): New function.
6191 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6192 start + width and last_order of the group, fail if check_no_overlap
6194 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6195 to group if check_no_overlap fails.
6197 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6199 * config/rs6000/altivec.md: Delete contraint arguments to
6200 define_expand, define_split, and define_peephole2, and in
6201 define_insn_and_split if always unused.
6202 * config/rs6000/darwin.md: Ditto.
6203 * config/rs6000/dfp.md: Ditto.
6204 * config/rs6000/rs6000.md: Ditto.
6205 * config/rs6000/sync.md: Ditto.
6206 * config/rs6000/vector.md: Ditto.
6207 * config/rs6000/vsx.md: Ditto.
6209 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6211 * config/rs6000/altivec.md: Write output control strings as braced
6212 blocks instead of double-quoted strings.
6213 * config/rs6000/darwin.md: Ditto.
6214 * config/rs6000/rs6000.md: Ditto.
6215 * config/rs6000/vector.md: Ditto.
6216 * config/rs6000/vsx.md: Ditto.
6218 2018-02-21 Jason Merrill <jason@redhat.com>
6220 PR c++/84314 - ICE with templates and fastcall attribute.
6221 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6223 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6225 * ipa-cp.c (determine_versionability): Fix comment typos.
6227 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6230 * ipa-cp.c (determine_versionability): Do not version functions caling
6233 2018-02-21 Martin Liska <mliska@suse.cz>
6236 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6237 Add "native" as a possible value.
6238 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6239 the macro when native cpu detection is available.
6241 2018-02-21 Martin Liska <mliska@suse.cz>
6244 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6245 Add "native" as a possible value.
6246 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6247 when native cpu detection is available.
6249 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6250 Martin Sebor <msebor@redhat.com>
6252 PR tree-optimization/84478
6253 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6255 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6256 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6257 support which is conservatively correct, for 2 only stay conservative
6258 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6259 argument to the 2 argument get_range_strlen, adjust 6 arg
6260 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6262 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6263 (gimple_fold_builtin_strlen): Pass true as last argument to
6266 2018-02-20 Martin Sebor <msebor@redhat.com>
6269 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6270 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6271 (builtin_memref::builtin_memref): Factor out parts into
6272 set_base_and_offset and call it.
6274 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6277 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6278 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6279 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6280 search at the associated MODE_INT.
6282 2018-02-20 Jeff Law <law@redhat.com>
6285 PR tree-optimization/81592
6287 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6288 for range data rather than using global data.
6289 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6290 range data rather than using global data.
6291 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6292 pass it to children as needed.
6293 (struct directive::fmtresult): Similarly.
6294 (struct directive::set_width): Similarly.
6295 (struct directive::set_precision): Similarly.
6296 (format_integer, format_directive, parse_directive): Similarly.
6297 (format_none): Accept unnamed vr_values parameter.
6298 (format_percent, format_floating, format_character): Similarly.
6299 (format_string, format_plain): Similarly.
6300 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6301 the EVRP range analyzer for range data rather than using global data.
6302 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6303 gimple-ssa-evrp-analyze.h
6304 (class sprintf_dom_walker): Add after_dom_children member function.
6305 Add evrp_range_analyzer member.
6306 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6307 range analyzer as needed.
6308 (sprintf_dom_walker::after_dom_children): New member function.
6309 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6311 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6312 (evrp_range_analyzer::pop_to_marker): Likewise.
6314 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6316 PR tree-optimization/84419
6317 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6318 with the required type if its current type is compatible but
6321 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6324 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6325 after vectorization.
6327 2018-02-20 Martin Liska <mliska@suse.cz>
6330 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6331 possible values if we don't have a hint.
6333 2018-02-20 Martin Liska <mliska@suse.cz>
6337 * final.c (shorten_branches): Build align_tab array with one
6339 * opts.c (finish_options): Add alignment option limit check.
6340 (MAX_CODE_ALIGN): Likewise.
6341 (MAX_CODE_ALIGN_VALUE): Likewise.
6342 * doc/invoke.texi: Document maximum allowed option value for
6343 all -falign-* options.
6345 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6348 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6349 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6350 * var-tracking.c (emit_note_insn_var_location): Remove all references
6351 to NOTE_INSN_CALL_ARG_LOCATION.
6352 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6353 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6354 Use copy_rtx_if_shared.
6355 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6356 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6357 (dwarf2out_var_location): Remove handling of
6358 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6360 * final.c (final_scan_insn): Remove all references to
6361 NOTE_INSN_CALL_ARG_LOCATION.
6362 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6363 before dumping final insns.
6364 * except.c (emit_note_eh_region_end): Remove all references to
6365 NOTE_INSN_CALL_ARG_LOCATION.
6366 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6367 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6368 * config/arc/arc.c (hwloop_optimize): Likewise.
6369 * config/arm/arm.c (create_fix_barrier): Likewise.
6370 * config/s390/s390.c (s390_chunkify_start): Likewise.
6371 * config/sh/sh.c (find_barrier): Likewise.
6372 * config/i386/i386.c (rest_of_insert_endbranch,
6373 ix86_seh_fixup_eh_fallthru): Likewise.
6374 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6375 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6376 * config/frv/frv.c (frv_function_prologue): Likewise.
6377 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6379 (note_outside_basic_block_p): Remove all references to
6380 NOTE_INSN_CALL_ARG_LOCATION.
6381 * gengtype.c (adjust_field_rtx_def): Likewise.
6382 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6384 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6385 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6388 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6391 PR tree-optimization/84452
6392 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6393 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6396 2018-02-19 Martin Liska <mliska@suse.cz>
6399 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6401 2018-02-19 Martin Liska <mliska@suse.cz>
6402 Richard Sandiford <richard.sandiford@linaro.org>
6404 PR tree-optimization/82491
6405 * gimple-fold.c (get_base_constructor): Make earlier bail out
6408 2018-02-19 Carl Love <cel@us.ibm.com>
6410 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6411 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6413 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6414 P8V_BUILTIN_VEC_NEG.
6416 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6418 * config/rl78/rl78.md (movdf): New define expand.
6420 2018-02-19 Martin Liska <mliska@suse.cz>
6423 * doc/invoke.texi: Fix typo.
6424 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6426 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6428 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6429 handle rs6000_single_float and rs6000_double_float specially for
6432 2018-02-16 Jeff Law <law@redhat.com>
6434 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6437 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6440 * ipa-inline.c (inline_small_functions): Fix a typo.
6442 2018-02-16 Nathan Sidwell <nathan@acm.org>
6444 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6446 2018-02-16 Carl Love <cel@us.ibm.com>
6448 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6449 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6450 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6451 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6452 expansion to P8V_BUILTIN_VEC_FLOAT2.
6454 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6456 PR rtl-optimization/70023
6457 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6458 src_regno into account.
6460 2018-02-16 Carl Love <cel@us.ibm.com>
6462 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6463 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6464 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6465 * config/rs6000/rs6000.c: Remove case statements for
6466 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6467 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6468 and P9V_BUILTIN_VEC_VINSERT4B.
6469 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6470 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6471 * config/rs6000/vsx.md:
6472 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6475 2018-02-16 Carl Love <cel@us.ibm.com>
6477 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6479 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6481 * config/rs6000/rs6000-c.c: Add the definitions for
6482 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6483 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6484 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6485 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6486 definition for insert4b and define insn *insert3b_internal.
6487 * doc/extend.texi: Add documentation for vec_extract4b.
6489 2018-02-16 Nathan Sidwell <nathan@acm.org>
6491 * doc/extend.texi (Backwards Compatibility): Mention friend
6492 injection. Note for-scope is deprecated.
6493 * doc/invoke.texi (-ffriend-injection): Deprecate.
6495 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6497 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6498 that moved to I2, also allow destinations that are a paradoxical
6499 subreg (instead of a normal reg).
6501 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6504 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6507 2018-02-16 Richard Biener <rguenther@suse.de>
6509 PR tree-optimization/84037
6510 PR tree-optimization/84016
6512 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6513 Adjust vec_construct for the fact we need additional higher latency
6514 128bit inserts for AVX256 and AVX512 vector builds.
6515 (ix86_add_stmt_cost): Scale vector construction cost for
6518 2018-02-16 Richard Biener <rguenther@suse.de>
6520 PR tree-optimization/84417
6521 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6522 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6523 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6525 2018-02-16 Martin Liska <mliska@suse.cz>
6528 * internal-fn.def (ASAN_CHECK): Set proper flags.
6529 (ASAN_MARK): Likewise.
6531 2018-02-16 Julia Koval <julia.koval@intel.com>
6533 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6534 from PTA_CANNONLAKE.
6536 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6539 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6540 Use ++iter rather than iter++ for std::list iterators.
6541 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6542 defer deleting them until all nodes in the forest are processed. Do
6543 free even leaf nodes. Change to_process into auto_vec.
6546 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6547 * vec.h (vec_default_construct): Use memset instead of placement new
6548 if BROKEN_VALUE_INITIALIZATION is defined.
6549 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6550 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6553 PR rtl-optimization/83723
6554 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6555 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6556 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6558 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6560 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6562 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6564 PR rtl-optimization/81443
6565 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6566 from inner REGs to paradoxical SUBREGs.
6568 2018-02-16 Richard Biener <rguenther@suse.de>
6570 PR tree-optimization/84399
6571 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6572 For operands we can analyze at their definition make sure we can
6573 analyze them at each use as well.
6575 2018-02-16 Richard Biener <rguenther@suse.de>
6577 PR tree-optimization/84190
6578 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6579 volatile accesses if the decl isn't volatile.
6581 2018-02-15 Jason Merrill <jason@redhat.com>
6583 PR c++/84314 - ICE with templates and fastcall attribute.
6584 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6585 TYPE_CANONICAL on an existing type.
6587 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6589 PR tree-optimization/84383
6590 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6591 dstoff nor call operand_equal_p if dstbase is NULL.
6593 PR tree-optimization/84334
6594 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6595 also a CONSTANT_CLASS_P, punt.
6597 2018-02-14 Jim Wilson <jimw@sifive.com>
6599 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6600 first SMALL_OPERAND check. New local min_second_step. Move assert
6601 to where locals are set. Add TARGET_RVC support.
6602 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6604 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6606 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6608 2018-02-14 Martin Sebor <msebor@redhat.com>
6610 PR tree-optimization/83698
6611 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6612 arrays constrain the offset range to their bounds.
6613 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6614 (builtin_access::overlap): Avoid setting the size of overlap if it's
6616 (maybe_diag_overlap): Also consider arrays when deciding what values
6617 of offsets to include in diagnostics.
6619 2018-02-14 Martin Sebor <msebor@redhat.com>
6622 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6623 that correspond to the kind of a declaration.
6625 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6628 * config/pa/pa.md: Load address of PIC label using the linkage table
6629 if the label is nonlocal.
6631 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6633 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6634 warning message if user requests -maltivec=be.
6635 * doc/invoke.texi: Document deprecation of -maltivec=be.
6637 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6640 * config/rs6000/rs6000-c.c: Update definitions for
6641 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6642 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6644 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6647 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6648 add _get_ssp intrinsics. Remove argument from
6649 __builtin_ia32_rdssp[d|q].
6650 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6651 * config/i386/i386-builtin.def: Remove argument from
6652 __builtin_ia32_rdssp[d|q].
6653 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6654 ix86_expand_special_args_builtin for _rdssp[d|q].
6655 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6656 Clear register before usage.
6657 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6658 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6660 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6662 PR tree-optimization/84357
6663 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6664 operand 1 of an ARRAY_REF too.
6666 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6669 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6670 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6672 (set_of_reg): New struct.
6673 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6674 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6675 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6677 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6678 Split into bitclr, bitset, bitinvert patterns if appropriate.
6679 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6680 use rx_fuse_in_memory_bitop.
6681 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6682 to named insn, correct maximum insn length.
6684 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6687 * machmode.def: Define a complex mode for PARTIAL_INT.
6688 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6690 * doc/rtl.texi: Document CSPImode.
6691 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6693 (msp430_hard_regno_nregs_with_padding): Likewise.
6695 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6698 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6700 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6702 PR rtl-optimization/84169
6703 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6704 we generated a parallel as new i3 and we split that to new i2 and i3
6705 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6706 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6707 those to i2, not i1. Partially rewrite this scan code.
6709 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6712 * stor-layout.c (place_field): For variable length fields, adjust
6713 offset_align afterwards not just based on the field's alignment,
6714 but also on the size.
6717 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6718 of exps and logs in the use_exp2 case.
6720 2018-02-13 Jeff Law <law@redhat.com>
6722 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6725 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6728 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6732 * common.opt (gas-loc-support, gas-locview-support): New.
6733 (ginline-points, ginternal-reset-location-views): New.
6734 * doc/invoke.texi: Document them. Use @itemx where intended.
6735 (gvariable-location-views): Adjust.
6736 * target.def (reset_location_view): New.
6737 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6738 (TARGET_RESET_LOCATION_VIEW): New.
6739 * doc/tm.texi: Rebuilt.
6740 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6741 (dwarf2out_default_as_locview_support): New.
6742 (output_asm_line_debug_info): Use option variables.
6743 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6744 (output_loc_list): Likewise.
6745 (add_high_low_attributes): Check option variables.
6746 Don't output entry view attribute in strict mode.
6747 (gen_inlined_subroutine_die): Check option variables.
6748 (dwarf2out_inline_entry): Likewise.
6749 (init_sections_and_labels): Likewise.
6750 (dwarf2out_early_finish): Likewise.
6751 (maybe_reset_location_view): New, from...
6752 (dwarf2out_var_location): ... here. Call it.
6753 * debug.h (dwarf2out_default_as_loc_support): Declare.
6754 (dwarf2out_default_as_locview_support): Declare.
6755 * hooks.c (hook_int_rtx_insn_0): New.
6756 * hooks.h (hook_int_rtx_insn_0): Declare.
6757 * toplev.c (process_options): Take -gas-loc-support and
6758 -gas-locview-support from dwarf2out. Enable
6759 -gvariable-location-views by default only with locview
6760 assembler support. Enable -ginternal-reset-location-views by
6761 default only if the target defines the corresponding hook.
6762 Enable -ginline-points by default if location views are
6763 enabled; force it disabled if statement frontiers are
6765 * tree-inline.c (expand_call_inline): Check option variables.
6766 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6768 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6770 PR tree-optimization/84321
6771 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6772 handling. Also check whether the anti-range contains any values
6773 that satisfy the mask; switch to a VR_RANGE if not.
6775 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6778 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6780 2018-02-13 Martin Jambor <mjambor@suse.cz>
6783 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6784 of call statements, also set location of a load to a temporary.
6786 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6788 * config/rl78/rl78.c (add_vector_labels): New function.
6789 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6790 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6791 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6792 which checks that no arguments are passed.
6793 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6794 * doc/extend.texi: Documentation for the new attribute.
6796 2018-02-13 Andreas Schwab <schwab@suse.de>
6798 * config/riscv/linux.h (CPP_SPEC): Define.
6800 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6803 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6804 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6805 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6806 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6807 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6808 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6809 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6810 temporarily for AES and PCLMUL builtins.
6812 PR tree-optimization/84339
6813 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6814 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6818 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6819 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6820 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6822 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6824 * omp-simd-clone.h: New file.
6825 * omp-simd-clone.c: Include omp-simd-clone.h.
6826 (expand_simd_clones): No longer static.
6827 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6828 cgraph.h and omp-simd-clone.h.
6829 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6830 (vect_recog_widen_shift_pattern): Formatting fix.
6831 (vect_pattern_recog_1): Don't check optab for calls.
6834 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6835 operands[2] into a REG before using gen_lowpart on it.
6837 2018-02-12 Jeff Law <law@redhat.com>
6840 * config/sh/sh.c (find_barrier): Consider a sibling call
6843 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6844 successfully back substituting a reg.
6846 2018-02-12 Richard Biener <rguenther@suse.de>
6848 PR tree-optimization/84037
6849 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6850 parameter, move visited init to caller.
6851 (vect_slp_analyze_operations): Separate cost from validity
6852 check, initialize visited once for all instances.
6853 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6855 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6857 (vect_model_promotion_demotion_cost): Likewise.
6858 (vectorizable_bswap): Guard cost modeling with !slp_node
6859 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6861 (vectorizable_call): Likewise.
6862 (vectorizable_conversion): Likewise.
6863 (vectorizable_assignment): Likewise.
6864 (vectorizable_shift): Likewise.
6865 (vectorizable_operation): Likewise.
6866 (vectorizable_store): Likewise.
6867 (vectorizable_load): Likewise.
6868 (vectorizable_condition): Likewise.
6869 (vectorizable_comparison): Likewise.
6871 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6874 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6875 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6876 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6878 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6881 * params.def (inline-min-speedup): Increase from 8 to 15.
6882 (max-inline-insns-auto): Decrease from 40 to 30.
6883 * ipa-split.c (consider_split): Add some buffer for function to
6884 be considered inlining candidate.
6885 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6888 2018-02-12 Richard Biener <rguenther@suse.de>
6890 PR tree-optimization/84037
6891 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6892 matched stmts if we cannot swap the non-matched ones.
6894 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6896 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6897 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6898 _mm_maskz_scalef_round_ss): New intrinsics.
6899 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6900 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6901 __builtin_ia32_scalefss_round): Remove.
6902 (__builtin_ia32_scalefsd_mask_round,
6903 __builtin_ia32_scalefss_mask_round): New intrinsics.
6904 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6905 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6906 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6907 "<round_constraint>")): Changed to ...
6908 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6909 "<round_scalar_constraint>")): ... this.
6910 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6911 %0, %1, %2<round_op3>}"): Changed to ...
6912 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6913 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6914 %2<round_scalar_mask_op3>}"): ... this.
6915 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6917 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6919 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6920 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6921 (_mm_maskz_sqrt_round_ss): New intrinsics.
6922 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6923 (__builtin_ia32_sqrtsd_mask_round)
6924 (__builtin_ia32_sqrtss_mask_round): New builtins.
6925 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6926 (__builtin_ia32_sqrtss_round): Remove.
6927 (__builtin_ia32_sqrtsd_mask_round)
6928 (__builtin_ia32_sqrtss_mask_round): New builtins.
6929 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6930 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6931 ((match_operand:VF_128 1 "vector_operand"
6932 "xBm,<round_constraint>")): Changed to ...
6933 ((match_operand:VF_128 1 "vector_operand"
6934 "xBm,<round_scalar_constraint>")): ... this.
6935 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6936 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6937 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6938 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6939 %<iptr>1<round_scalar_mask_op3>}): ... this.
6940 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6941 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6943 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6946 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6947 Cast vec_cmpeq result to correct type.
6948 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6949 Cast vec_cmpgt result to correct type.
6951 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6953 * final.c (final_scan_insn_1): Renamed from...
6954 (final_scan_insn): ... this. New wrapper, to recover
6955 seen from the outermost call in recursive ones.
6956 * config/sparc/sparc.c (output_return): Drop seen from call.
6957 (output_sibcall): Likewise.
6958 * config/visium/visium.c (output_branch): Likewise.
6960 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6962 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6965 2018-02-10 Alan Modra <amodra@gmail.com>
6968 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6969 Specify LR as an input.
6971 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6974 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6975 remove_member_access_dummy_vars): New functions.
6976 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6977 lower_omp_1, execute_lower_omp): Use them.
6979 PR rtl-optimization/84308
6980 * shrink-wrap.c (spread_components): Release todo vector.
6982 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6984 PR rtl-optimization/57193
6985 * ira-color.c (struct allocno_color_data): Add member
6986 conflict_allocno_hard_prefs.
6987 (update_conflict_allocno_hard_prefs): New.
6988 (bucket_allocno_compare_func): Add a preference based on
6989 conflict_allocno_hard_prefs.
6990 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6991 (color_allocnos): Remove a dead code. Initiate
6992 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6994 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6997 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6998 constraint from =wa to wa. Avoid a subreg on the output operand,
6999 instead use a pseudo and subreg it in a move.
7000 (p9_xxbrd_<mode>): Changed to ...
7001 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7002 (p9_xxbrd_v2df): New expander.
7003 (p9_xxbrw_<mode>): Changed to ...
7004 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7005 (p9_xxbrw_v4sf): New expander.
7007 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7009 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7011 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7014 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7015 multiply in 32-bit mode.
7016 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7017 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7020 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7022 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7023 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7024 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7025 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7027 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7030 * dwarf2out.c (is_trivial_indirect_ref): New function.
7031 (dwarf2out_late_global_decl): Do not generate a location
7032 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7033 and that are not defined in the current unit.
7035 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7037 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7038 instead of a libcall for UNORDERED.
7040 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7043 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7044 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7046 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7049 * config/s390/s390.c (s390_set_current_function): Invoke
7050 s390_indirect_branch_settings also if fndecl didn't change.
7052 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7054 * config/rs6000/rs6000.md (blockage): Set length to zero.
7056 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7058 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7060 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7063 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7064 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7068 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7069 PARALLEL incoming that failed vt_get_decl_and_offset check.
7072 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7073 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7075 (get_variable_section): For decls in named .bss* sections pass true as
7076 second argument to bss_initializer_p.
7078 2018-02-09 Marek Polacek <polacek@redhat.com>
7079 Jakub Jelinek <jakub@redhat.com>
7082 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7083 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7084 Sync some changes from cxx_fold_indirect_ref.
7086 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7088 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7090 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7091 (BLOCK_INLINE_ENTRY_LABEL): New.
7092 (dwarf2out_var_location): Disregard inline entry markers.
7093 (inline_entry_data): New struct.
7094 (inline_entry_data_hasher): New hashtable type.
7095 (inline_entry_data_hasher::hash): New.
7096 (inline_entry_data_hasher::equal): New.
7097 (inline_entry_data_table): New variable.
7098 (add_high_low_attributes): Add DW_AT_entry_pc and
7099 DW_AT_GNU_entry_view attributes if a pending entry is found
7100 in inline_entry_data_table. Add old entry_pc attribute only
7101 if debug nonbinding markers are disabled.
7102 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7103 markers are enabled.
7104 (block_within_block_p, dwarf2out_inline_entry): New.
7105 (dwarf2out_finish): Check that no entries remained in
7106 inline_entry_data_table.
7107 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7108 (final_scan_insn, notice_source_line): Likewise.
7109 (rest_of_clean_state): Skip inline entry markers.
7110 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7112 * gimple.c (gimple_build_debug_inline_entry): New.
7113 * gimple.h (enum gimple_debug_subcode): Add
7114 GIMPLE_DEBUG_INLINE_ENTRY.
7115 (gimple_build_debug_inline_entry): Declare.
7116 (gimple_debug_inline_entry_p): New.
7117 (gimple_debug_nonbind_marker_p): Adjust.
7118 * insn-notes.def (INLINE_ENTRY): New.
7119 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7120 inline entry marker notes.
7121 (print_insn): Likewise.
7122 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7123 (INSN_DEBUG_MARKER_KIND): Likewise.
7124 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7125 * tree-inline.c (expand_call_inline): Build and insert
7126 debug_inline_entry stmt.
7127 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7128 inline entry blocks early, if nonbind markers are enabled.
7129 (dump_scope_block): Dump fragment info.
7130 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7131 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7132 (gimple_build_debug_inline_entry): New.
7133 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7134 Enable/disable inline entry points too.
7135 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7136 (DEBUG_INSN): Describe inline entry markers.
7138 * common.opt (gvariable-location-views): New.
7139 (gvariable-location-views=incompat5): New.
7140 * config.in: Rebuilt.
7141 * configure: Rebuilt.
7142 * configure.ac: Test assembler for view support.
7143 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7144 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7145 * dwarf2out.c (var_loc_view): New typedef.
7146 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7147 (dwarf2out_locviews_in_attribute): New.
7148 (dwarf2out_locviews_in_loclist): New.
7149 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7150 (enum dw_line_info_opcode): Add LI_adv_address.
7151 (struct dw_line_info_table): Add view.
7152 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7153 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7154 (zero_view_p): New variable.
7155 (ZERO_VIEW_P): New macro.
7156 (output_asm_line_debug_info): New.
7157 (struct var_loc_node): Add view.
7158 (add_AT_view_list, AT_loc_list): New.
7159 (add_var_loc_to_decl): Add view param. Test it against last.
7160 (new_loc_list): Add view params. Record them.
7161 (AT_loc_list_ptr): Handle loc and view lists.
7162 (view_list_to_loc_list_val_node): New.
7163 (print_dw_val): Handle dw_val_class_view_list.
7164 (size_of_die): Likewise.
7165 (value_format): Likewise.
7166 (loc_list_has_views): New.
7167 (gen_llsym): Set vl_symbol too.
7168 (maybe_gen_llsym, skip_loc_list_entry): New.
7169 (dwarf2out_maybe_output_loclist_view_pair): New.
7170 (output_loc_list): Output view list or entries too.
7171 (output_view_list_offset): New.
7172 (output_die): Handle dw_val_class_view_list.
7173 (output_dwarf_version): New.
7174 (output_compilation_unit_header): Use it.
7175 (output_skeleton_debug_sections): Likewise.
7176 (output_rnglists, output_line_info): Likewise.
7177 (output_pubnames, output_aranges): Update version comments.
7178 (output_one_line_info_table): Output view numbers in asm comments.
7179 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7180 Call maybe_gen_llsym.
7181 (loc_list_from_tree_1): Adjust.
7182 (add_AT_location_description): Create view list attribute if
7183 needed, check it's absent otherwise.
7184 (convert_cfa_to_fb_loc_list): Adjust.
7185 (maybe_emit_file): Call output_asm_line_debug_info for test.
7186 (dwarf2out_var_location): Reset views as needed. Precompute
7187 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7188 attribute. Set view.
7189 (new_line_info_table): Reset next view.
7190 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7191 (dwarf2out_source_line): Likewise. Output view resets and labels to
7192 the assembler, or select appropriate line info opcodes.
7193 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7194 (optimize_string_length): Catch it. Adjust.
7195 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7196 dw_val_class_view_list, and remove it if no longer needed.
7197 (hash_loc_list): Hash view numbers.
7198 (loc_list_hasher::equal): Compare them.
7199 (optimize_location_lists): Check whether a view list symbol is
7200 needed, and whether the locview attribute is present, and
7201 whether they match. Remove the locview attribute if no longer
7203 (index_location_lists): Call skip_loc_list_entry for test.
7204 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7205 Use output_dwarf_version.
7206 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7207 (struct dw_val_node): Add val_view_list.
7208 * final.c (SEEN_NEXT_VIEW): New.
7209 (set_next_view_needed): New.
7210 (clear_next_view_needed): New.
7211 (maybe_output_next_view): New.
7212 (final_start_function): Rename to...
7213 (final_start_function_1): ... this. Take pointer to FIRST,
7214 add SEEN parameter. Emit param bindings in the initial view.
7215 (final_start_function): Reintroduce SEEN-less interface.
7216 (final): Rename to...
7217 (final_1): ... this. Take SEEN parameter. Output final pending
7218 next view at the end.
7219 (final): Reintroduce seen-less interface.
7220 (final_scan_insn): Output pending next view before switching
7221 sections or ending a block. Mark the next view as needed when
7222 outputting variable locations. Notify debug backend of section
7223 changes, and of location view changes.
7224 (rest_of_handle_final): Adjust.
7225 * toplev.c (process_options): Autodetect value for debug variable
7226 location views option. Warn on incompat5 without -gdwarf-5.
7227 * doc/invoke.texi (gvariable-location-views): New.
7228 (gvariable-location-views=incompat5): New.
7229 (gno-variable-location-views): New.
7231 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7233 PR tree-optimization/84136
7234 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7235 that the result of find_edge is non-NULL.
7237 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7240 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7241 storing integer register in SImode. Fix cost of 256 and 512
7242 byte aligned SSE register store.
7244 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7246 * config/i386/i386.c (ix86_multiplication_cost): Fix
7247 multiplication cost for TARGET_AVX512DQ.
7249 2018-02-08 Marek Polacek <polacek@redhat.com>
7251 PR tree-optimization/84238
7252 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7255 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7257 PR tree-optimization/84265
7258 * tree-vect-stmts.c (vectorizable_store): Don't treat
7259 VMAT_CONTIGUOUS accesses as grouped.
7260 (vectorizable_load): Likewise.
7262 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7264 PR tree-optimization/81635
7265 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7266 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7267 (test_round_for_mask): New functions.
7268 (wide_int_cc_tests): Call test_round_for_mask.
7269 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7270 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7271 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7272 range returned by get_range_info.
7274 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7277 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7278 * symtab.c: Include builtins.h
7279 (symtab_node::output_to_lto_symbol_table_p): Move here
7280 from lto-streamer-out.c:output_symbol_p.
7281 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7282 (output_symbol_p): Move all logic to symtab.c
7283 (produce_symtab): Update.
7285 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7287 * config/s390/s390-opts.h (enum indirect_branch): Define.
7288 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7289 (s390_indirect_branch_via_thunk)
7290 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7291 (enum s390_indirect_branch_type): Define.
7292 * config/s390/s390.c (struct s390_frame_layout, struct
7293 machine_function): Remove.
7294 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7295 (indirect_branch_table_label_no, indirect_branch_table_name):
7297 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7298 (enum s390_indirect_branch_option): Define.
7299 (s390_return_addr_from_memory): New function.
7300 (s390_handle_string_attribute): New function.
7301 (s390_attribute_table): Add new attribute handler.
7302 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7303 (s390_indirect_branch_via_thunk): New function.
7304 (s390_indirect_branch_via_inline_thunk): New function.
7305 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7306 sibling call optimization for non z10 compiles.
7307 (s390_emit_call): Force indirect branch target to be a single
7308 register. Add r1 clobber for non-z10 compiles.
7309 (s390_emit_epilogue): Emit return jump via return_use expander.
7310 (s390_reorg): Handle JUMP_INSNs as execute targets.
7311 (s390_option_override_internal): Perform validity checks for the
7312 new command line options.
7313 (s390_indirect_branch_attrvalue): New function.
7314 (s390_indirect_branch_settings): New function.
7315 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7316 (s390_output_indirect_thunk_function): New function.
7317 (s390_code_end): Implement target hook.
7318 (s390_case_values_threshold): Implement target hook.
7319 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7321 * config/s390/s390.h (struct s390_frame_layout)
7322 (struct machine_function): Move here from s390.c.
7323 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7324 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7325 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7326 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7327 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7328 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7329 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7330 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7331 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7332 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7333 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7334 (mnemonic attribute): Add values which aren't recognized
7336 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7337 pattern for branch conversion. Fix mnemonic attribute.
7338 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7339 indirect branch via thunk if requested.
7340 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7341 ("*indirect_jump"): Disable for branch conversion using out of
7343 ("indirect_jump_via_thunk<mode>_z10")
7344 ("indirect_jump_via_thunk<mode>")
7345 ("indirect_jump_via_inlinethunk<mode>_z10")
7346 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7347 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7348 ("casesi_jump_via_inlinethunk<mode>_z10")
7349 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7350 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7351 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7352 ("*indirect2_jump"): Disable for branch conversion.
7353 ("casesi_jump"): Turn into expander and expand patterns for branch
7355 ("return_use"): New expander.
7356 ("*return"): Emit return via thunk and rename it to ...
7357 ("*return<mode>"): ... this one.
7358 * config/s390/s390.opt: Add new options and and enum for the
7361 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7363 * lra-constraints.c (match_reload): Unconditionally use
7364 gen_lowpart_SUBREG, rather than selecting between that
7365 and equivalent gen_rtx_SUBREG code.
7367 2018-02-08 Richard Biener <rguenther@suse.de>
7369 PR tree-optimization/84233
7370 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7371 changed flag instead of boguously re-using phi_inserted.
7373 2018-02-08 Martin Jambor <mjambor@suse.cz>
7375 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7376 static local variables.
7378 2018-02-08 Richard Biener <rguenther@suse.de>
7380 PR tree-optimization/84278
7381 * tree-vect-stmts.c (vectorizable_store): When looking for
7382 smaller vector types to perform grouped strided loads/stores
7383 make sure the mode is supported by the target.
7384 (vectorizable_load): Likewise.
7386 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7388 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7389 Increase LDP/STP opportunities by adding adjacent callee-saves.
7391 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7393 PR rtl-optimization/84068
7394 PR rtl-optimization/83459
7395 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7397 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7399 PR tree-optimization/84224
7400 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7401 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7404 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7407 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7408 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7409 count, remove one USE.
7411 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7413 * doc/install.texi (Configuration): Document the
7414 --with-long-double-format={ibm,ieee} PowerPC configuration
7418 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7419 Convert from define_expand to be define_insn_and_split. Rework
7420 float/double/_Float128 conversions to QI/HI/SImode to work with
7421 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7422 conversions to QI/HImode types did a store and then a load to
7423 truncate the value. For conversions to VSX registers, don't split
7424 the insn, instead emit the code directly. Use the code iterator
7425 any_fix to combine signed and unsigned conversions.
7426 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7427 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7428 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7429 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7430 (fix_<mode>di2_hw): Likewise.
7431 (fixuns_<mode>di2_hw): Likewise.
7432 (fix_<mode>si2_hw): Likewise.
7433 (fixuns_<mode>si2_hw): Likewise.
7434 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7435 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7436 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7437 fix<uns>_trunc<SFDF:mode>si2_p8.
7438 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7440 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7441 (fix<uns>_<mode>_mem): Likewise.
7442 (fctiw<u>z_<mode>_mem): Likewise.
7443 (fix<uns>_<mode>_mem): Likewise.
7444 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7445 the register allocator from doing a direct move to the GPRs to do
7446 a store, and instead use the ISA 3.0 store byte/half-word from
7447 vector register instruction. For IEEE 128-bit floating point,
7448 also optimize stores of 32-bit ints.
7449 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7451 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7453 * genextract.c (push_pathstr_operand): New function to support
7455 (walk_rtx): Call push_pathstr_operand.
7456 (print_path): Support [a-zA-Z].
7458 2018-02-07 Richard Biener <rguenther@suse.de>
7460 PR tree-optimization/84037
7461 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7462 (cse_and_gimplify_to_preheader): Declare.
7463 (vect_get_place_in_interleaving_chain): Likewise.
7464 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7466 (_loop_vec_info::~_loop_vec_info): Delete it.
7467 (cse_and_gimplify_to_preheader): New function.
7468 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7469 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7470 (vectorizable_load): Likewise. For grouped stores always base
7471 the IV on the first element.
7472 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7473 condition before gimplifying.
7475 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7477 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7478 *DIV_EXPR and *MOD_EXPR.
7480 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7483 * config/i386/i386.c (ix86_option_override_internal): Mask out
7484 the CF_SET bit when checking -fcf-protection.
7486 2018-02-07 Tom de Vries <tom@codesourcery.com>
7489 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7492 2018-02-07 Richard Biener <rguenther@suse.de>
7494 PR tree-optimization/84204
7495 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7498 PR tree-optimization/84205
7499 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7500 special-case isl_ast_op_zdiv_r.
7502 PR tree-optimization/84223
7503 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7504 Only add conditions from within the region.
7505 (gather_bbs::after_dom_children): Adjust.
7507 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7510 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7511 * config/avr/avr.md: Only post-reload split REG-REG moves if
7512 either register is GENERAL_REG_P.
7514 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7516 PR tree-optimization/84235
7517 * tree-ssa-scopedtables.c
7518 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7519 if the subtraction is performed in floating point type where NaNs are
7520 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7521 build 1. Formatting fix.
7523 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7526 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7527 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7528 and skip it regardless of bb boundaries. Use CALL_P macro,
7529 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7530 unnecessarily, formatting fix.
7532 2018-02-06 Michael Collison <michael.collison@arm.com>
7534 * config/arm/thumb2.md:
7535 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7536 (*thumb_mov_notscc): Ditto.
7538 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7541 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7542 unsigned_fix, not "s".
7544 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7546 * configure.ac (gcc_fn_eh_frame_ro): New function.
7547 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7548 correct .eh_frame permissions.
7549 * configure: Regenerate.
7551 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7553 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7556 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7558 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7559 Display warning message for -mno-speculate-indirect-jumps.
7561 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7563 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7565 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7567 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7569 PR tree-optimization/84225
7570 * tree-eh.c (find_trapping_overflow): Only call
7571 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7573 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7576 * config/i386/i386.c: Reimplement the check of possible options
7577 -mibt/-mshstk conbination. Change error messages.
7578 * doc/invoke.texi: Fix a typo: remove extra '='.
7580 2018-02-06 Marek Polacek <polacek@redhat.com>
7582 PR tree-optimization/84228
7583 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7585 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7588 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7589 emitted arch directives.
7590 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7591 __ARM_FEATURE_COPROC before changing architectures.
7593 2018-02-06 Richard Biener <rguenther@suse.de>
7595 * config/i386/i386.c (print_reg): Fix typo.
7596 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7598 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7600 * configure: Regenerate.
7602 2018-02-05 Martin Sebor <msebor@redhat.com>
7604 PR tree-optimization/83369
7605 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7608 2018-02-05 Martin Liska <mliska@suse.cz>
7610 * doc/invoke.texi: Cherry-pick upstream r323995.
7612 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7614 * ira.c (ira_init_register_move_cost): Adjust comment.
7616 2018-02-05 Martin Liska <mliska@suse.cz>
7618 PR gcov-profile/84137
7619 * doc/gcov.texi: Fix typo in documentation.
7621 2018-02-05 Martin Liska <mliska@suse.cz>
7623 PR gcov-profile/83879
7624 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7625 using dlopen functionality.
7627 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7629 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7630 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7631 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7632 _mm_maskz_range_round_ss): New intrinsics.
7633 (__builtin_ia32_rangesd128_round)
7634 (__builtin_ia32_rangess128_round): Remove.
7635 (__builtin_ia32_rangesd128_mask_round,
7636 __builtin_ia32_rangess128_mask_round): New builtins.
7637 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7638 __builtin_ia32_rangess128_round): Remove.
7639 (__builtin_ia32_rangesd128_mask_round,
7640 __builtin_ia32_rangess128_mask_round): New builtins.
7641 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7642 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7643 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7644 "<round_saeonly_constraint>")): Changed to ...
7645 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7646 "<round_saeonly_scalar_constraint>")): ... this.
7647 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7648 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7649 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7650 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7651 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7653 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7655 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7657 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7658 Remove all values except native, 8540 and 8548.
7660 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7662 * config/i386/i386.c (ix86_output_function_return): Pass
7663 INVALID_REGNUM, instead of -1, as invalid register number to
7664 indirect_thunk_name and output_indirect_thunk.
7666 2018-02-02 Julia Koval <julia.koval@intel.com>
7668 * config.gcc: Add -march=icelake.
7669 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7670 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7671 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7672 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7673 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7674 (processor_target_table): Add icelake.
7675 (ix86_option_override_internal): Handle new PTAs.
7676 (get_builtin_code_for_version): Handle icelake.
7677 (M_INTEL_COREI7_ICELAKE): New.
7678 (fold_builtin_cpu): Handle icelake.
7679 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7680 * doc/invoke.texi: Add -march=icelake.
7682 2018-02-02 Julia Koval <julia.koval@intel.com>
7684 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7685 to wide_int_bitmask.
7686 * wide-int-bitmask.h: New.
7688 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7691 * config/i386/i386.md: Replace Pmode with word_mode in
7692 builtin_setjmp_setup and builtin_longjmp to support x32.
7694 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7698 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7700 (rs6000_supported_cpu_names): New static variable.
7701 (linux_cpu_translation_table): Likewise.
7702 (elf_platform) <cpu>: Define new static variable and use it.
7703 Translate kernel AT_PLATFORM name to canonical name if needed.
7704 Error if platform name is unknown.
7706 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7709 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7711 2018-02-01 Jeff Law <law@redhat.com>
7714 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7715 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7716 the scratch if RELEASE_VIA_POP is false.
7717 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7718 If we have to save a temporary register, decrement SIZE appropriately.
7719 Pass new arguments to release_scratch_register_on_entry.
7720 (ix86_adjust_stack_and_probe): Likewise.
7721 (ix86_emit_probe_stack_range): Pass new arguments to
7722 release_scratch_register_on_entry.
7724 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7726 PR rtl-optimization/84157
7727 * combine.c (change_zero_ext): Use REG_P predicate in
7728 front of HARD_REGISTER_P predicate.
7730 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7732 * config/avr/avr.c (avr_option_override): Move disabling of
7733 -fdelete-null-pointer-checks to...
7734 * common/config/avr/avr-common.c (avr_option_optimization_table):
7737 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7739 PR tree-optimization/81635
7740 * tree-data-ref.c (split_constant_offset_1): For types that
7741 wrap on overflow, try to use range info to prove that wrapping
7744 2018-02-01 Renlin Li <renlin.li@arm.com>
7747 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7749 (aarch64_register_move_cost): Likewise.
7750 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7752 (REG_CLASS_NAMES): Likewise.
7753 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7754 TAILCALL_ADDR_REGS. Remove IP registers.
7755 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7757 2018-02-01 Richard Biener <rguenther@suse.de>
7759 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7760 for specifying RPO order and allow NULL for that.
7761 * domwalk.c (dom_walker::dom_walker): Likewise.
7762 (dom_walker::walk): Handle NULL RPO order.
7763 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7765 (rewrite_update_dom_walker): Likewise.
7766 (mark_def_dom_walker): Likewise.
7768 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7770 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7771 (aarch64_maybe_expand_sve_subreg_move): Declare.
7772 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7773 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7775 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7776 that are semantically a reverse operation.
7777 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7778 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7779 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7781 (aarch64_can_change_mode_class): For big-endian, forbid changes
7782 between two SVE modes if they have different element sizes.
7784 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7786 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7787 the TImode handling for big-endian targets.
7789 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7791 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7792 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7794 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7795 Remove BSWAP handing for big-endian targets and use the form of
7796 LD1RQ appropariate for the mode.
7798 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7800 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7801 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7804 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7807 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7808 check for operands that need to go through aarch64_sve_reload_be.
7810 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7812 PR tree-optimization/81661
7813 PR tree-optimization/84117
7814 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7815 * tree-eh.c: Include gimplify.h.
7816 (find_trapping_overflow, replace_trapping_overflow,
7817 rewrite_to_non_trapping_overflow): New functions.
7818 * tree-vect-loop.c: Include tree-eh.h.
7819 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7820 * tree-data-ref.c: Include tree-eh.h.
7821 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7823 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7825 PR rtl-optimization/84123
7826 * combine.c (change_zero_ext): Check if hard register satisfies
7827 can_change_dest_mode before calling gen_lowpart_SUBREG.
7829 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7832 * ira.c (ira_init_register_move_cost): Remove assert.
7834 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7836 PR rtl-optimization/84071
7837 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7838 * doc/tm.texi: Regenerate.
7840 2018-01-31 Richard Biener <rguenther@suse.de>
7842 PR tree-optimization/84132
7843 * tree-data-ref.c (analyze_miv_subscript): Properly
7844 check whether evolution_function_is_affine_multivariate_p
7845 before calling gcd_of_steps_may_divide_p.
7847 2018-01-31 Julia Koval <julia.koval@intel.com>
7850 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7851 * config/i386/i386.md (rdpid_rex64) New.
7852 (rdpid): Make 32bit only.
7854 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7857 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7858 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7860 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7863 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7865 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7867 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7869 PR rtl-optimization/84071
7870 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7871 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7873 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7875 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7876 (arc_attribute_table): Add 'aux' attribute.
7877 (arc_in_small_data_p): Consider aux like variables.
7878 (arc_is_aux_reg_p): New function.
7879 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7880 (arc_get_aux_arg): New function.
7881 (prepare_move_operands): Handle aux-register access.
7882 (arc_handle_aux_attribute): New function.
7883 * doc/extend.texi (ARC Variable attributes): Add subsection.
7885 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7887 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7888 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7889 (arc_attribute_table): Add 'uncached' attribute.
7890 (arc_print_operand): Print '.di' flag for uncached memory
7892 (arc_in_small_data_p): Do not consider for small data the uncached
7894 (arc_is_uncached_mem_p): New function.
7895 * config/arc/predicates.md (compact_store_memory_operand): Check
7896 for uncached memory accesses.
7897 (nonvol_nonimm_operand): Likewise.
7898 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7900 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7903 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7904 falign-loops=): Add Optimization flag.
7906 2018-01-30 Jeff Law <law@redhat.com>
7909 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7910 INT_REGISTERS_SAVED. Check it prior to calling
7911 get_scratch_register_on_entry.
7912 (ix86_adjust_stack_and_probe): Similarly.
7913 (ix86_emit_probe_stack_range): Similarly.
7914 (ix86_expand_prologue): Corresponding changes.
7916 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7919 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7920 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7922 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7925 * lra-constraints.c (curr_insn_transform): Process AND in the
7928 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7930 PR rtl-optimization/83986
7931 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7932 dependence against last_pending_memory_flush in addition to
7935 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7937 PR tree-optimization/81611
7938 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7941 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7944 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7947 2018-01-30 Richard Biener <rguenther@suse.de>
7948 Jakub Jelinek <jakub@redhat.com>
7950 PR tree-optimization/84111
7951 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7952 inner loops added during recursion, as they don't have up-to-date
7955 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7958 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7959 (can_inline_edge_by_limits_p): ... here.
7960 (can_early_inline_edge_p, check_callers,
7961 update_caller_keys, update_callee_keys, recursive_inlining,
7962 add_new_edges_to_heap, speculation_useful_p,
7963 inline_small_functions,
7964 inline_small_functions, flatten_function,
7965 inline_to_all_callers_1): Update.
7967 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7969 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7972 2018-01-30 Richard Biener <rguenther@suse.de>
7974 PR tree-optimization/83008
7975 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7976 invariant and constant vector uses in stmts when they need
7979 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7982 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7983 * configure: Regenerate.
7985 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7987 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7989 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7990 Use gen_rtx_REG rather than gen_lowpart.
7992 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7994 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7995 rather than 0 when creating partial subregs.
7997 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7999 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8002 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8005 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8006 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8007 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8008 flags. This restores the settings used before the 2017-07-24.
8009 Turning off pre increment/decrement/modify allows IVOPTS to
8010 optimize DF/SF loops where the index is an int.
8012 2018-01-29 Richard Biener <rguenther@suse.de>
8013 Kelvin Nilsen <kelvin@gcc.gnu.org>
8016 * tree-vect-stmts.c (vectorizable_call): Don't call
8017 targetm.vectorize_builtin_md_vectorized_function if callee is
8020 2018-01-22 Carl Love <cel@us.ibm.com>
8022 * doc/extend.tex: Fix typo in second arg in
8023 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8025 2018-01-29 Richard Biener <rguenther@suse.de>
8027 PR tree-optimization/84086
8028 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8029 (flush_ssaname_freelist): When SSA names were released reset
8030 the SCEV hash table.
8032 2018-01-29 Richard Biener <rguenther@suse.de>
8034 PR tree-optimization/84057
8035 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8036 removed paths when removing edges.
8038 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8040 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8041 -mfunction-return=@var{choice}.
8043 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8046 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8048 (layout::print_source_line): Likewise.
8049 (test_get_line_width_without_trailing_whitespace): Add test cases.
8051 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8054 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8057 2018-01-26 Jim Wilson <jimw@sifive.com>
8059 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8061 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8064 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8066 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8067 and CMP + SUB-immediate -> SUBS.
8069 2018-01-26 Martin Sebor <msebor@redhat.com>
8071 PR tree-optimization/83896
8072 * tree-ssa-strlen.c (get_string_len): Rename...
8073 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8074 Avoid assuming length is constant.
8075 (handle_char_store): Use HOST_WIDE_INT for string length.
8077 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8080 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8081 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8083 2018-01-26 Richard Biener <rguenther@suse.de>
8085 PR rtl-optimization/84003
8086 * dse.c (record_store): Only record redundant stores when
8087 the earlier store aliases at least all accesses the later one does.
8089 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8091 PR rtl-optimization/83985
8092 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8093 REG_CFA_RESTORE insns.
8094 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8095 REG_CFA_RESTORE insns here.
8098 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8099 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8101 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8103 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8104 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8105 (arc_init): Likewise.
8106 (arc_override_options): Likewise.
8107 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8109 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8111 * config/arc/arc.h (TARGET_DBNZ): Define.
8112 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8113 properly set the tune attribute.
8114 (dbnz): Use TARGET_DBNZ guard.
8115 * config/arc/arc.opt (mtune): Add core3 option.
8117 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8119 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8120 recognize new pic like addresses.
8121 (arc_delegitimize_address): Clean up.
8123 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8125 * config/arc/arc-arches.def: Option mrf16 valid for all
8127 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8128 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8129 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8130 * config/arc/arc-tables.opt: Regenerate.
8131 * config/arc/arc.c (arc_conditional_register_usage): Handle
8132 reduced register file case.
8133 (arc_file_start): Set must have build attributes.
8134 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8136 * config/arc/arc.opt (mrf16): Add new option.
8137 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8138 * config/arc/genmultilib.awk: Handle new mrf16 option.
8139 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8140 * config/arc/t-multilib: Regenerate.
8141 * doc/invoke.texi (ARC Options): Document mrf16 option.
8143 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8145 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8146 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8147 (arc_attribute_table): Add 'secure_call' attribute.
8148 (arc_print_operand): Print secure call operand.
8149 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8151 (arc_is_secure_call_p): New function. * config/arc/arc.md
8152 (call_i): Add support for sjli instruction.
8153 (call_value_i): Likewise.
8154 * config/arc/constraints.md (Csc): New constraint.
8156 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8157 John Eric Martin <John.Martin@emmicro-us.com>
8159 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8160 * config/arc/arc.c (_arc_jli_section): New struct.
8161 (arc_jli_section): New type.
8162 (rc_jli_sections): New static variable.
8163 (arc_handle_jli_attribute): New function.
8164 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8165 (arc_file_end): New function.
8166 (TARGET_ASM_FILE_END): Define.
8167 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8168 (arc_add_jli_section): New function.
8169 (jli_call_scan): Likewise.
8170 (arc_reorg): Call jli_call_scan.
8171 (arc_output_addsi): Remove 'S' from printing asm operand.
8172 (arc_is_jli_call_p): New function.
8173 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8175 (movhi_insn): Likewise.
8176 (movsi_insn): Likewise.
8177 (movsi_set_cc_insn): Likewise.
8178 (loadqi_update): Likewise.
8179 (load_zeroextendqisi_update): Likewise.
8180 (load_signextendqisi_update): Likewise.
8181 (loadhi_update): Likewise.
8182 (load_zeroextendhisi_update): Likewise.
8183 (load_signextendhisi_update): Likewise.
8184 (loadsi_update): Likewise.
8185 (loadsf_update): Likewise.
8186 (movsicc_insn): Likewise.
8187 (bset_insn): Likewise.
8188 (bxor_insn): Likewise.
8189 (bclr_insn): Likewise.
8190 (bmsk_insn): Likewise.
8191 (bicsi3_insn): Likewise.
8192 (cmpsi_cc_c_insn): Likewise.
8193 (movsi_ne): Likewise.
8194 (movsi_cond_exec): Likewise.
8195 (clrsbsi2): Likewise.
8206 (call_i): Remove 'S' asm letter, add jli instruction.
8207 (call_value_i): Likewise.
8208 * config/arc/arc.op (mjli-always): New option.
8209 * config/arc/constraints.md (Cji): New constraint.
8210 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8212 (subsf3_fpx): Likewise.
8213 (mulsf3_fpx): Likewise.
8214 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8216 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8218 * doc/invoke.texi (ARC): Document mjli-always option.
8220 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8222 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8223 avoid addition with 0 and use incw and decw where possible.
8225 2018-01-26 Richard Biener <rguenther@suse.de>
8227 PR tree-optimization/81082
8228 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8229 association if it requires casting to unsigned.
8230 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8231 from fold_plusminus_mult_expr to catch important cases late when
8232 range info is available.
8234 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8236 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8237 * configure.ac (hidden_linkonce): New test.
8238 * configure: Regenerate.
8239 * config.in: Regenerate.
8241 2018-01-26 Julia Koval <julia.koval@intel.com>
8243 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8244 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8245 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8246 _mm_mask_bitshuffle_epi64_mask): Fix type.
8247 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8248 USI_FTYPE_V4DI_V4DI_USI): Remove.
8249 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8250 __builtin_ia32_vpshufbitqmb256_mask,
8251 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8252 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8253 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8255 2018-01-26 Alan Modra <amodra@gmail.com>
8258 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8259 UNSPEC_VBPERMQ. Sort other unspecs.
8261 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8263 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8265 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8268 * predict.c (drop_profile): Do not push/pop cfun; update also
8270 (handle_missing_profiles): Fix logic looking for zero profiles.
8272 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8275 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8276 on functions with #pragma omp declare simd or functions with simd
8278 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8279 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8280 Remove trailing \n from warning_at calls.
8282 2018-01-25 Tom de Vries <tom@codesourcery.com>
8285 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8286 for neutered workers.
8288 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8291 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8292 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8294 2018-01-24 Jeff Law <law@redhat.com>
8297 * i386.c (get_probe_interval): Move to earlier point.
8298 (ix86_compute_frame_layout): If -fstack-clash-protection and
8299 the frame is larger than the probe interval, then use pushes
8300 to save registers rather than reg->mem moves.
8301 (ix86_expand_prologue): Remove conditional for int_registers_saved
8304 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8307 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8308 min/max for never referenced object.
8310 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8313 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8315 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8316 attributes from DECL_ATTRIBUTES (decl) without affecting
8317 DECL_ATTRIBUTES (current_function_decl).
8318 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8319 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8321 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8323 PR tree-optimization/83979
8324 * fold-const.c (fold_comparison): Use constant_boolean_node
8325 instead of boolean_{true,false}_node.
8327 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8329 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8332 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8334 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8335 Simplify the clause that sets the length attribute.
8336 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8337 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8338 clause that sets the length attribute.
8339 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8341 2018-01-24 Tom de Vries <tom@codesourcery.com>
8344 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8345 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8346 Add strict parameter.
8347 (prevent_branch_around_nothing): Insert dummy insn between branch to
8348 label and label with no ptx insn inbetween.
8349 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8351 2018-01-24 Tom de Vries <tom@codesourcery.com>
8354 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8355 for neutered threads in warp.
8356 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8358 2018-01-24 Richard Biener <rguenther@suse.de>
8360 PR tree-optimization/83176
8361 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8364 2018-01-24 Richard Biener <rguenther@suse.de>
8366 PR tree-optimization/82819
8367 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8368 code generating pluses that are no-ops in the target precision.
8370 2018-01-24 Richard Biener <rguenther@suse.de>
8373 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8375 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8377 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8378 to merge probabilities.
8379 * predict.c (probably_never_executed): Also mark as cold functions
8380 with global 0 profile and guessed local profile.
8381 * profile-count.c (profile_probability::combine_with_count): New
8383 * profile-count.h (profile_probability::operator*,
8384 profile_probability::operator*=, profile_probability::operator/,
8385 profile_probability::operator/=): Reduce precision to adjusted
8386 and set value to guessed on contradictory divisions.
8387 (profile_probability::combine_with_freq): Remove.
8388 (profile_probability::combine_wiht_count): Declare.
8389 (profile_count::force_nonzero):: Set to adjusted.
8390 (profile_count::probability_in):: Set quality to adjusted.
8391 * tree-ssa-tail-merge.c (replace_block_by): Use
8394 2018-01-23 Andrew Waterman <andrew@sifive.com>
8395 Jim Wilson <jimw@sifive.com>
8397 * config/riscv/riscv.c (riscv_stack_boundary): New.
8398 (riscv_option_override): Set riscv_stack_boundary. Handle
8399 riscv_preferred_stack_boundary_arg.
8400 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8401 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8402 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8403 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8404 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8405 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8407 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8410 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8411 of struct ix86_frame.
8412 (ix86_expand_epilogue): Likewise. Add a local variable for
8413 the reg_save_offset field in struct ix86_frame.
8415 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8417 PR tree-optimization/82604
8418 * tree-loop-distribution.c (enum partition_kind): New enum item
8419 PKIND_PARTIAL_MEMSET.
8420 (partition_builtin_p): Support above new enum item.
8421 (generate_code_for_partition): Ditto.
8422 (compute_access_range): Differentiate cases that equality can be
8423 proven at all loops, the innermost loops or no loops.
8424 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8425 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8426 (finalize_partitions, distribute_loop): Don't fuse partition of
8427 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8428 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8431 2018-01-23 Martin Liska <mliska@suse.cz>
8433 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8434 order to ignore the predictor.
8435 (PRED_POLYMORPHIC_CALL): Likewise.
8436 (PRED_RECURSIVE_CALL): Likewise.
8438 2018-01-23 Martin Liska <mliska@suse.cz>
8440 * tree-profile.c (tree_profiling): Print function header to
8441 aware reader which function we are working on.
8442 * value-prof.c (gimple_find_values_to_profile): Do not print
8443 not interesting value histograms.
8445 2018-01-23 Martin Liska <mliska@suse.cz>
8447 * profile-count.h (enum profile_quality): Add
8448 profile_uninitialized as the first value. Do not number values
8449 as they are zero based.
8450 (profile_count::verify): Update sanity check.
8451 (profile_probability::verify): Likewise.
8453 2018-01-23 Nathan Sidwell <nathan@acm.org>
8455 * doc/invoke.texi (ffor-scope): Deprecate.
8457 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8459 PR tree-optimization/83510
8460 * domwalk.c (set_all_edges_as_executable): New function.
8461 (dom_walker::dom_walker): Convert bool param
8462 "skip_unreachable_blocks" to enum reachability. Move setup of
8463 edge flags to set_all_edges_as_executable and only do it when
8464 reachability is REACHABLE_BLOCKS.
8465 * domwalk.h (enum dom_walker::reachability): New enum.
8466 (dom_walker::dom_walker): Convert bool param
8467 "skip_unreachable_blocks" to enum reachability.
8468 (set_all_edges_as_executable): New decl.
8469 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8470 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8472 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8473 but converting true to REACHABLE_BLOCKS.
8474 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8476 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8477 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8478 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8480 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8481 if check_all_array_refs will be called.
8483 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8485 * tree.c (selftest::test_location_wrappers): Add more test
8488 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8490 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8491 (selftest::test_bit_in_range): Likewise.
8493 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8496 * doc/sourcebuild.texi (vect_float): Say that the selector
8497 only describes the situation when -funsafe-math-optimizations is on.
8498 (vect_float_strict): Document.
8500 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8502 PR tree-optimization/83965
8503 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8504 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8505 instead of checking only for a reduction.
8506 (vect_recog_widen_sum_pattern): Likewise.
8508 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8510 * predict.c (probably_never_executed): Only use precise profile info.
8511 (compute_function_frequency): Skip after inlining hack since we now
8512 have quality checking.
8514 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8516 * profile-count.h (profile_probability::very_unlikely,
8517 profile_probability::unlikely, profile_probability::even): Set
8518 precision to guessed.
8520 2018-01-23 Richard Biener <rguenther@suse.de>
8522 PR tree-optimization/83963
8523 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8524 Properly terminate dominator walk when crossing the exit edge not
8525 when visiting its source block.
8527 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8530 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8531 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8533 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8535 PR tree-optimization/83957
8536 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8537 semicolon after for body surrounded by braces.
8539 PR tree-optimization/83081
8540 * profile-count.h (profile_probability::split): New method.
8541 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8542 Use profile_probability::split.
8543 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8544 when splitting a single conditional jump into 2.
8546 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8548 PR tree-optimization/69452
8549 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8552 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8554 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8555 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8556 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8558 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8560 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8561 * config/rl78/rl78.md (movdi): New define_expand.
8562 * config/rl78/rl78.c (rl78_split_movdi): New function.
8564 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8567 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8569 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8570 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8571 128-bit to produce an UNSPEC move to get the double word with the
8572 signbit and then a shift directly to do signbit.
8573 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8574 implementation with a new version that just does either a direct
8575 move or a regular move. Move memory interface to separate insns.
8576 Move insns so they are next to the expander.
8577 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8578 with signbit move. Split big and little endian case.
8579 (signbit<mode>2_dm_mem_le): Likewise.
8580 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8581 (signbit<mode>2_dm2): Likewise.
8583 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8585 * config/rl78/rl78.md (anddi3): New define_expand.
8587 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8589 * config/rl78/rl78.md (umindi3): New define_expand.
8591 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8593 * config/rl78/rl78.md (smindi3): New define_expand.
8595 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8597 * config/rl78/rl78.md (smaxdi3): New define_expand.
8599 2018-01-22 Carl Love <cel@us.ibm.com>
8601 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8602 LVX_V1TI): Add macro expansion.
8603 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8604 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8605 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8606 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8607 Change check to determine if the instruction is a byte reversing
8608 entry. Fix typo in comment.
8609 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8610 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8611 Add def_builtin calls for new builtins.
8612 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8613 Add define_insn expansion.
8615 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8617 * config/rl78/rl78.md (umaxdi3): New define_expand.
8619 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8621 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8622 for non-QImode registers.
8624 2018-01-22 Richard Biener <rguenther@suse.de>
8626 PR tree-optimization/83963
8627 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8628 including the loop exit block.
8629 (scop_detection::merge_sese): Likewise.
8630 (scop_detection::add_scop): Do it here instead.
8632 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8634 * doc/sourcebuild.texi (arm_softfloat): Document.
8636 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8639 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8640 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8641 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8643 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8644 David Edelsohn <dje.gcc@gmail.com>
8647 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8648 Change "crset eq" to "crset 2".
8649 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8650 (*call_indirect_aix<mode>_nospec): Likewise.
8651 (*call_value_indirect_aix<mode>_nospec): Likewise.
8652 (*call_indirect_elfv2<mode>_nospec): Likewise.
8653 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8654 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8655 change assembly output from . to $.
8656 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8657 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8658 (*tablejump<mode>_internal1_nospec): Likewise.
8660 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8663 * config/sh/sh_optimize_sett_clrt.cc:
8664 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8666 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8668 PR tree-optimization/83940
8669 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8670 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8671 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8672 use it to pass back the definition type.
8673 (vect_check_store_rhs): Likewise rhs_dt_out.
8674 (vect_build_gather_load_calls): Add a mask_dt argument and use
8675 it instead of a call to vect_is_simple_use.
8676 (vectorizable_store): Update calls to vect_check_load_store_mask
8677 and vect_check_store_rhs. Use the dt returned by the latter instead
8678 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8679 instead of calls to vect_is_simple_use. Pass the scalar rather
8680 than the vector operand to vect_is_simple_use when handling
8681 second and subsequent copies of an rhs value.
8682 (vectorizable_load): Update calls to vect_check_load_store_mask
8683 and vect_build_gather_load_calls. Use the cached mask_dt and
8684 gs_info.offset_dt instead of calls to vect_is_simple_use.
8686 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8689 * tree-emutls.c: Include gimplify.h.
8690 (lower_emutls_2): New function.
8691 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8692 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8693 it before further processing.
8696 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8697 UINTVAL (trueop1) instead of INTVAL (op1).
8699 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8703 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8704 INCOMING_FRAME_SP_OFFSET if not defined.
8705 (scan_trace): Add ENTRY argument. If true and
8706 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8707 emit a note to adjust the CFA offset.
8708 (create_cfi_notes): Adjust scan_trace callers.
8709 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8710 INCOMING_FRAME_SP_OFFSET in the CIE.
8711 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8712 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8714 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8715 * doc/tm.texi: Regenerated.
8717 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8719 PR rtl-optimization/83147
8720 * lra-constraints.c (remove_inheritance_pseudos): Use
8721 lra_substitute_pseudo_within_insn.
8723 2018-01-19 Tom de Vries <tom@codesourcery.com>
8724 Cesar Philippidis <cesar@codesourcery.com>
8727 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8729 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8732 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8733 spaces for function labels.
8735 2018-01-19 Martin Liska <mliska@suse.cz>
8737 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8738 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8739 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8740 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8741 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8742 (PRED_CONST_RETURN): Change from 69 to 65.
8743 (PRED_NULL_RETURN): Change from 91 to 71.
8744 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8745 (PRED_LOOP_GUARD): Change from 66 to 73.
8747 2018-01-19 Martin Liska <mliska@suse.cz>
8749 * predict.c (predict_insn_def): Add new assert.
8750 (struct branch_predictor): Change type to signed integer.
8751 (test_prediction_value_range): Amend test to cover
8753 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8754 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8755 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8756 (PRED_LOOP_IV_COMPARE): Likewise.
8757 * predict.h (PROB_UNINITIALIZED): Define new constant.
8759 2018-01-19 Martin Liska <mliska@suse.cz>
8761 * predict.c (dump_prediction): Add new format for
8762 analyze_brprob.py script which is enabled with -details
8764 * profile-count.h (precise_p): New function.
8766 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8768 PR tree-optimization/83922
8769 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8770 there are no statements that need masking.
8771 (vect_active_double_reduction_p): New function.
8772 (vect_analyze_loop_operations): Use it when handling phis that
8773 are not in the loop header.
8775 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8777 PR tree-optimization/83914
8778 * tree-vect-loop.c (vectorizable_induction): Don't convert
8779 init_expr or apply the peeling adjustment for inductions
8780 that are nested within the vectorized loop.
8782 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8784 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8787 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8791 * function.h (gimplify_parameters): Add gimple_seq * argument.
8792 * function.c: Include gimple.h and options.h.
8793 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8794 for the added local temporaries if needed.
8795 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8796 if there are any parameter cleanups, wrap whole body into a
8797 try/finally with the cleanups.
8799 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8802 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8803 Use GET_MODE_CLASS for scalar floating point.
8805 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8809 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8810 Fix call of call_cgraph_insertion_hooks.
8812 2018-01-18 Martin Sebor <msebor@redhat.com>
8814 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8816 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8819 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8822 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8825 * common.opt: (-ffile-prefix-map): New option.
8826 * opts.c (common_handle_option): Defer it.
8827 * opts-global.c (handle_common_deferred_options): Handle it.
8828 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8829 * file-prefix-map.h: New file.
8830 (remap_debug_filename, add_debug_prefix_map): ...here.
8831 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8832 * final.c (debug_prefix_map, add_debug_prefix_map
8833 remap_debug_filename): Move to...
8834 * file-prefix-map.c: New file.
8835 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8836 generalize, get rid of alloca(), use strrchr() instead of strchr().
8837 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8838 Implement in terms of add_prefix_map().
8839 (remap_macro_filename, remap_debug_filename): Implement in term of
8841 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8842 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8843 * dbxout.c: Include file-prefix-map.h.
8844 * varasm.c: Likewise.
8845 * vmsdbgout.c: Likewise.
8846 * xcoffout.c: Likewise.
8847 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8848 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8849 * doc/invoke.texi (-ffile-prefix-map): Document.
8850 (-fdebug-prefix-map): Update description.
8852 2018-01-18 Martin Liska <mliska@suse.cz>
8854 * config/i386/i386.c (indirect_thunk_name): Document that also
8856 (output_indirect_thunk): Document why both instructions
8857 (pause and lfence) are generated.
8859 2018-01-18 Richard Biener <rguenther@suse.de>
8861 PR tree-optimization/83887
8862 * graphite-scop-detection.c
8863 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8864 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8865 (scop_detection::merge_sese): Re-implement with a flood-fill
8866 algorithm that properly finds a SESE region if it exists.
8868 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8871 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8872 pointer_diff optimizations use view_convert instead of convert.
8874 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8876 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8877 Generate different code for -mno-speculate-indirect-jumps.
8878 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8879 (*call_indirect_aix<mode>): Disable for
8880 -mno-speculate-indirect-jumps.
8881 (*call_indirect_aix<mode>_nospec): New define_insn.
8882 (*call_value_indirect_aix<mode>): Disable for
8883 -mno-speculate-indirect-jumps.
8884 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8885 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8886 -mno-speculate-indirect-jumps.
8887 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8889 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8891 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8892 long double type, set the flags for noting the default long double
8893 type, even if we don't pass or return a long double type.
8895 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8898 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8901 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8903 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8904 support for merge[hl].
8905 (fold_mergehl_helper): New helper function.
8906 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8907 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8908 (altivec_vmrglw_direct): Add xxmrglw insn.
8910 2018-01-17 Andrew Waterman <andrew@sifive.com>
8912 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8913 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8915 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8918 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8919 call the lto_location_cache before reading the
8920 DECL_SOURCE_LOCATION of the types.
8922 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8923 Richard Sandiford <richard.sandiford@linaro.org>
8925 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8926 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8927 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8928 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8929 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8931 * config/aarch64/constraints.md (aarch64_movti_operand):
8933 * config/aarch64/predicates.md (Uti): Add new constraint.
8935 2018-01-17 Carl Love <cel@us.ibm.com>
8937 * config/rs6000/vsx.md (define_expand xl_len_r,
8938 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8939 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8941 (define_expand, define_insn): Move the shift left from the
8942 define_insn to the define_expand for lxvl and stxvl instructions.
8943 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8944 and XL_LEN_R definitions to PURE.
8946 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8948 * config/i386/i386.c (indirect_thunk_name): Declare regno
8949 as unsigned int. Compare regno with INVALID_REGNUM.
8950 (output_indirect_thunk): Ditto.
8951 (output_indirect_thunk_function): Ditto.
8952 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8953 in the call to output_indirect_thunk_function.
8955 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8958 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8959 rather than the size of inner_type to determine the stack slot size
8960 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8962 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8965 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8968 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8970 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8971 endian Linux systems to optionally enable multilibs for selecting
8972 the long double type if the user configured an explicit type.
8973 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8974 have no long double multilibs if not defined.
8975 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8976 warn if the user used -mabi={ieee,ibm}longdouble and we built
8977 multilibs for long double.
8978 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8979 appropriate multilib option.
8980 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8982 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8983 for building long double multilibs.
8984 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8986 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8988 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8991 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8993 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8996 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8999 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9002 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9004 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9005 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9007 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9009 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9010 different rtl trees depending on TARGET_64BIT.
9011 (rs6000_gen_lvx): Likewise.
9013 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9015 * config/visium/visium.md (nop): Tweak comment.
9016 (hazard_nop): Likewise.
9018 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9020 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9021 -mspeculate-indirect-jumps.
9022 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9023 for -mno-speculate-indirect-jumps.
9024 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9025 (*call_value_indirect_elfv2<mode>): Disable for
9026 -mno-speculate-indirect-jumps.
9027 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9028 (indirect_jump): Emit different RTL for
9029 -mno-speculate-indirect-jumps.
9030 (*indirect_jump<mode>): Disable for
9031 -mno-speculate-indirect-jumps.
9032 (*indirect_jump<mode>_nospec): New define_insn.
9033 (tablejump): Emit different RTL for
9034 -mno-speculate-indirect-jumps.
9035 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9036 (tablejumpsi_nospec): New define_expand.
9037 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9038 (tablejumpdi_nospec): New define_expand.
9039 (*tablejump<mode>_internal1): Disable for
9040 -mno-speculate-indirect-jumps.
9041 (*tablejump<mode>_internal1_nospec): New define_insn.
9042 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9045 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9047 * caller-save.c (insert_save): Drop unnecessary parameter. All
9050 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9051 Richard Biener <rguenth@suse.de>
9054 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9055 return early, inline manually is_gimple_sizepos. Make sure if we
9056 call gimplify_expr we don't end up with a gimple constant.
9057 * tree.c (variably_modified_type_p): Don't return true for
9058 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9059 * gimplify.h (is_gimple_sizepos): Remove.
9061 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9063 PR tree-optimization/83857
9064 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9065 vectorizable_live_operation for pure SLP statements.
9066 (vectorizable_live_operation): Handle PHIs.
9068 2018-01-16 Richard Biener <rguenther@suse.de>
9070 PR tree-optimization/83867
9071 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9072 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9074 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9077 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9078 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9079 If off is not INTEGER_CST, issue a may not be aligned warning
9080 rather than isn't aligned. Use isn%'t rather than isn't.
9081 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9083 <case MULT_EXPR>: Improve the case when bottom and one of the
9084 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9085 operand, in that case check if the other operand is multiple of
9086 bottom divided by the INTEGER_CST operand.
9088 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9091 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9092 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9093 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9094 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9095 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9096 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9097 (pa_function_arg_size): New function.
9099 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9101 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9102 in a separate statement.
9104 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9106 PR tree-optimization/83847
9107 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9108 group gathers and scatters.
9110 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9112 PR rtl-optimization/86620
9113 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9115 PR rtl-optimization/83213
9116 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9117 to last if both are JUMP_INSNs.
9119 PR tree-optimization/83843
9120 * gimple-ssa-store-merging.c
9121 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9122 store_immediate_info for bswap/nop orig_stores.
9124 2018-01-15 Andrew Waterman <andrew@sifive.com>
9126 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9128 <UDIV>: Increase cost if !TARGET_DIV.
9130 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9132 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9133 (define_attr "cr_logical_3op"): New.
9134 (cceq_ior_compare): Adjust.
9135 (cceq_ior_compare_complement): Adjust.
9136 (*cceq_rev_compare): Adjust.
9137 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9138 (is_cracked_insn): Adjust.
9139 (insn_must_be_first_in_group): Adjust.
9140 * config/rs6000/40x.md: Adjust.
9141 * config/rs6000/440.md: Adjust.
9142 * config/rs6000/476.md: Adjust.
9143 * config/rs6000/601.md: Adjust.
9144 * config/rs6000/603.md: Adjust.
9145 * config/rs6000/6xx.md: Adjust.
9146 * config/rs6000/7450.md: Adjust.
9147 * config/rs6000/7xx.md: Adjust.
9148 * config/rs6000/8540.md: Adjust.
9149 * config/rs6000/cell.md: Adjust.
9150 * config/rs6000/e300c2c3.md: Adjust.
9151 * config/rs6000/e500mc.md: Adjust.
9152 * config/rs6000/e500mc64.md: Adjust.
9153 * config/rs6000/e5500.md: Adjust.
9154 * config/rs6000/e6500.md: Adjust.
9155 * config/rs6000/mpc.md: Adjust.
9156 * config/rs6000/power4.md: Adjust.
9157 * config/rs6000/power5.md: Adjust.
9158 * config/rs6000/power6.md: Adjust.
9159 * config/rs6000/power7.md: Adjust.
9160 * config/rs6000/power8.md: Adjust.
9161 * config/rs6000/power9.md: Adjust.
9162 * config/rs6000/rs64.md: Adjust.
9163 * config/rs6000/titan.md: Adjust.
9165 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9167 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9168 ix86_indirect_branch_register logic.
9170 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9172 * config/i386/constraints.md (Bs): Update
9173 ix86_indirect_branch_register check. Don't check
9174 ix86_indirect_branch_register with GOT_memory_operand.
9176 * config/i386/predicates.md (GOT_memory_operand): Don't check
9177 ix86_indirect_branch_register here.
9178 (GOT32_symbol_operand): Likewise.
9180 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9182 * config/i386/predicates.md (constant_call_address_operand):
9183 Rewrite ix86_indirect_branch_register logic.
9184 (sibcall_insn_operand): Likewise.
9186 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9188 * config/i386/constraints.md (Bs): Replace
9189 ix86_indirect_branch_thunk_register with
9190 ix86_indirect_branch_register.
9192 * config/i386/i386.md (indirect_jump): Likewise.
9193 (tablejump): Likewise.
9194 (*sibcall_memory): Likewise.
9195 (*sibcall_value_memory): Likewise.
9196 Peepholes of indirect call and jump via memory: Likewise.
9197 * config/i386/i386.opt: Likewise.
9198 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9199 (GOT_memory_operand): Likewise.
9200 (call_insn_operand): Likewise.
9201 (sibcall_insn_operand): Likewise.
9202 (GOT32_symbol_operand): Likewise.
9204 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9207 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9208 type rather than type addr's type points to.
9209 (expand_omp_atomic_mutex): Likewise.
9210 (expand_omp_atomic): Likewise.
9212 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9215 * config/i386/i386.c (output_indirect_thunk_function): Use
9216 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9217 for __x86_return_thunk.
9219 2018-01-15 Richard Biener <rguenther@suse.de>
9222 * expmed.c (extract_bit_field_1): Fix typo.
9224 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9227 * config/arm/iterators.md (VF): New mode iterator.
9228 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9229 Remove integer-related logic from pattern.
9230 (neon_vabd<mode>_3): Likewise.
9232 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9235 * common.opt (fstrict-overflow): No longer an alias.
9236 (fwrapv-pointer): New option.
9237 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9238 also for pointer types based on flag_wrapv_pointer.
9239 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9240 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9241 opts->x_flag_wrapv got set.
9242 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9243 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9244 POINTER_TYPE_OVERFLOW_UNDEFINED.
9245 * match.pd: Likewise in address comparison pattern.
9246 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9248 2018-01-15 Richard Biener <rguenther@suse.de>
9251 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9252 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9253 Reset type names to their identifier if their TYPE_DECL doesn't
9254 have linkage (and thus is used for ODR and devirt).
9255 (save_debug_info_for_decl): Remove.
9256 (save_debug_info_for_type): Likewise.
9257 (add_tree_to_fld_list): Adjust.
9258 * tree-pretty-print.c (dump_generic_node): Make dumping of
9259 type names more robust.
9261 2018-01-15 Richard Biener <rguenther@suse.de>
9263 * BASE-VER: Bump to 8.0.1.
9265 2018-01-14 Martin Sebor <msebor@redhat.com>
9268 * builtins.c (check_access): Avoid warning when the no-warning bit
9271 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9273 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9274 * ira-color (allocno_hard_regs_compare): Likewise.
9276 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9279 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9280 Use .pushsection/.popsection.
9282 2018-01-14 Martin Sebor <msebor@redhat.com>
9285 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9287 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9289 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9290 entry from extra_headers.
9291 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9292 extra_headers, make the list bitwise identical to the i?86-*-* one.
9294 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9296 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9297 -mcmodel=large with -mindirect-branch=thunk,
9298 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9299 -mfunction-return=thunk-extern.
9300 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9301 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9302 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9304 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9306 * config/i386/i386.c (print_reg): Print the name of the full
9307 integer register without '%'.
9308 (ix86_print_operand): Handle 'V'.
9309 * doc/extend.texi: Document 'V' modifier.
9311 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9313 * config/i386/constraints.md (Bs): Disallow memory operand for
9314 -mindirect-branch-register.
9316 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9317 (GOT_memory_operand): Likewise.
9318 (call_insn_operand): Likewise.
9319 (sibcall_insn_operand): Likewise.
9320 (GOT32_symbol_operand): Likewise.
9321 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9322 for -mindirect-branch-register.
9323 (tablejump): Likewise.
9324 (*sibcall_memory): Likewise.
9325 (*sibcall_value_memory): Likewise.
9326 Disallow peepholes of indirect call and jump via memory for
9327 -mindirect-branch-register.
9328 (*call_pop): Replace m with Bw.
9329 (*call_value_pop): Likewise.
9330 (*sibcall_pop_memory): Replace m with Bs.
9331 * config/i386/i386.opt (mindirect-branch-register): New option.
9332 * doc/invoke.texi: Document -mindirect-branch-register option.
9334 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9336 * config/i386/i386-protos.h (ix86_output_function_return): New.
9337 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9338 set function_return_type.
9339 (indirect_thunk_name): Add ret_p to indicate thunk for function
9341 (output_indirect_thunk_function): Pass false to
9342 indirect_thunk_name.
9343 (ix86_output_indirect_branch_via_reg): Likewise.
9344 (ix86_output_indirect_branch_via_push): Likewise.
9345 (output_indirect_thunk_function): Create alias for function
9346 return thunk if regno < 0.
9347 (ix86_output_function_return): New function.
9348 (ix86_handle_fndecl_attribute): Handle function_return.
9349 (ix86_attribute_table): Add function_return.
9350 * config/i386/i386.h (machine_function): Add
9351 function_return_type.
9352 * config/i386/i386.md (simple_return_internal): Use
9353 ix86_output_function_return.
9354 (simple_return_internal_long): Likewise.
9355 * config/i386/i386.opt (mfunction-return=): New option.
9356 (indirect_branch): Mention -mfunction-return=.
9357 * doc/extend.texi: Document function_return function attribute.
9358 * doc/invoke.texi: Document -mfunction-return= option.
9360 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9362 * config/i386/i386-opts.h (indirect_branch): New.
9363 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9364 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9365 with local indirect jump when converting indirect call and jump.
9366 (ix86_set_indirect_branch_type): New.
9367 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9368 (indirectlabelno): New.
9369 (indirect_thunk_needed): Likewise.
9370 (indirect_thunk_bnd_needed): Likewise.
9371 (indirect_thunks_used): Likewise.
9372 (indirect_thunks_bnd_used): Likewise.
9373 (INDIRECT_LABEL): Likewise.
9374 (indirect_thunk_name): Likewise.
9375 (output_indirect_thunk): Likewise.
9376 (output_indirect_thunk_function): Likewise.
9377 (ix86_output_indirect_branch_via_reg): Likewise.
9378 (ix86_output_indirect_branch_via_push): Likewise.
9379 (ix86_output_indirect_branch): Likewise.
9380 (ix86_output_indirect_jmp): Likewise.
9381 (ix86_code_end): Call output_indirect_thunk_function if needed.
9382 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9384 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9385 (ix86_attribute_table): Add indirect_branch.
9386 * config/i386/i386.h (machine_function): Add indirect_branch_type
9387 and has_local_indirect_jump.
9388 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9390 (tablejump): Likewise.
9391 (*indirect_jump): Use ix86_output_indirect_jmp.
9392 (*tablejump_1): Likewise.
9393 (simple_return_indirect_internal): Likewise.
9394 * config/i386/i386.opt (mindirect-branch=): New option.
9395 (indirect_branch): New.
9398 (thunk-inline): Likewise.
9399 (thunk-extern): Likewise.
9400 * doc/extend.texi: Document indirect_branch function attribute.
9401 * doc/invoke.texi: Document -mindirect-branch= option.
9403 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9406 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9408 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9410 * ipa-inline.c (want_inline_small_function_p): Return false if
9411 inlining has already failed with CIF_FINAL_ERROR.
9412 (update_caller_keys): Call want_inline_small_function_p before
9414 (update_callee_keys): Likewise.
9416 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9418 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9420 (rs6000_quadword_masked_address_p): Likewise.
9421 (quad_aligned_load_p): Likewise.
9422 (quad_aligned_store_p): Likewise.
9423 (const_load_sequence_p): Add comment to describe the outer-most loop.
9424 (mimic_memory_attributes_and_flags): New function.
9425 (rs6000_gen_stvx): Likewise.
9426 (replace_swapped_aligned_store): Likewise.
9427 (rs6000_gen_lvx): Likewise.
9428 (replace_swapped_aligned_load): Likewise.
9429 (replace_swapped_load_constant): Capitalize argument name in
9430 comment describing this function.
9431 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9432 and stores that access quad-word aligned addresses and replace
9433 with stvx or lvx instructions when appropriate.
9434 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9435 New function prototype.
9436 (rs6000_quadword_masked_address_p): Likewise.
9437 (rs6000_gen_lvx): Likewise.
9438 (rs6000_gen_stvx): Likewise.
9439 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9440 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9441 when memory address is aligned.
9442 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9443 this split to select lvx instruction when memory address is aligned.
9444 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9445 instruction when memory address is aligned.
9446 (*vsx_le_perm_load_v16qi): Likewise.
9447 (four unnamed splitters): Modify to select the stvx instruction
9448 when memory is aligned.
9450 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9452 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9453 which appears in the queue multiple times.
9455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9456 Alan Hayward <alan.hayward@arm.com>
9457 David Sherwood <david.sherwood@arm.com>
9459 * tree-vectorizer.h (vec_lower_bound): New structure.
9460 (_loop_vec_info): Add check_nonzero and lower_bounds.
9461 (LOOP_VINFO_CHECK_NONZERO): New macro.
9462 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9463 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9464 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9465 fields. Make seg_len the distance travelled, not including the
9467 (dr_direction_indicator): Declare.
9468 (dr_zero_step_indicator): Likewise.
9469 (dr_known_forward_stride_p): Likewise.
9470 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9472 (runtime_alias_check_p): Allow runtime alias checks with
9474 (operator ==): Compare access_size and align.
9475 (prune_runtime_alias_test_list): Rework for new distinction between
9476 the access_size and seg_len.
9477 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9479 (get_segment_min_max): New function.
9480 (create_intersect_range_checks): Use it.
9481 (dr_step_indicator): New function.
9482 (dr_direction_indicator): Likewise.
9483 (dr_zero_step_indicator): Likewise.
9484 (dr_known_forward_stride_p): Likewise.
9485 * tree-loop-distribution.c (data_ref_segment_size): Return
9486 DR_STEP * (niters - 1).
9487 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9489 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9490 (vect_preserves_scalar_order_p): New function, split out from...
9491 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9492 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9493 (vect_vfa_access_size): New function.
9494 (vect_vfa_align): Likewise.
9495 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9496 (dump_lower_bound): New function.
9497 (vect_check_lower_bound): Likewise.
9498 (vect_small_gap_p): Likewise.
9499 (vectorizable_with_step_bound_p): Likewise.
9500 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9501 depencies if the vectorization factor is 1. Convert the checks
9502 for nonzero steps into checks on the bounds of DR_STEP. Try using
9503 a bunds check for variable steps if the minimum required step is
9504 relatively small. Update calls to the dr_with_seg_len
9505 constructor and to vect_compile_time_alias.
9506 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9508 (vect_loop_versioning): Call it.
9509 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9511 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9514 Alan Hayward <alan.hayward@arm.com>
9515 David Sherwood <david.sherwood@arm.com>
9517 * doc/sourcebuild.texi (vect_scatter_store): Document.
9518 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9520 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9522 * genopinit.c (main): Add supports_vec_scatter_store and
9523 supports_vec_scatter_store_cached to target_optabs.
9524 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9525 IFN_MASK_SCATTER_STORE.
9526 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9528 * internal-fn.h (internal_store_fn_p): Declare.
9529 (internal_fn_stored_value_index): Likewise.
9530 * internal-fn.c (scatter_store_direct): New macro.
9531 (expand_scatter_store_optab_fn): New function.
9532 (direct_scatter_store_optab_supported_p): New macro.
9533 (internal_store_fn_p): New function.
9534 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9535 IFN_MASK_SCATTER_STORE.
9536 (internal_fn_mask_index): Likewise.
9537 (internal_fn_stored_value_index): New function.
9538 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9540 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9541 * optabs-query.c (supports_vec_scatter_store_p): New function.
9542 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9543 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9544 true for scatter stores.
9545 (vect_gather_scatter_fn_p): Handle scatter stores too.
9546 (vect_check_gather_scatter): Consider using scatter stores if
9547 supports_vec_scatter_store_p.
9548 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9550 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9551 internal_fn_stored_value_index.
9552 (check_load_store_masking): Handle scatter stores too.
9553 (vect_get_store_rhs): Make public.
9554 (vectorizable_call): Use internal_store_fn_p.
9555 (vectorizable_store): Handle scatter store internal functions.
9556 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9557 when deciding whether the end of the group has been reached.
9558 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9559 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9560 (mask_scatter_store<mode>): New insns.
9562 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9563 Alan Hayward <alan.hayward@arm.com>
9564 David Sherwood <david.sherwood@arm.com>
9566 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9567 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9568 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9570 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9571 Use vect_truncate_gather_scatter_offset if we can't treat the
9572 operation as a normal gather load or scatter store.
9573 (get_group_load_store_type): Take the gather_scatter_info
9574 as argument. Try using a gather load or scatter store for
9575 single-element groups.
9576 (get_load_store_type): Update calls to get_group_load_store_type
9577 and vect_use_strided_gather_scatters_p.
9579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9580 Alan Hayward <alan.hayward@arm.com>
9581 David Sherwood <david.sherwood@arm.com>
9583 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9584 optional tree argument.
9585 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9587 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9588 but continue to use the current value as a fallback.
9589 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9590 to compare the updates.
9591 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9592 (get_load_store_type): Use it when handling a strided access.
9593 (vect_get_strided_load_store_ops): New function.
9594 (vect_get_data_ptr_increment): Likewise.
9595 (vectorizable_load): Handle strided gather loads. Always pass
9596 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9598 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9599 Alan Hayward <alan.hayward@arm.com>
9600 David Sherwood <david.sherwood@arm.com>
9602 * doc/md.texi (gather_load@var{m}): Document.
9603 (mask_gather_load@var{m}): Likewise.
9604 * genopinit.c (main): Add supports_vec_gather_load and
9605 supports_vec_gather_load_cached to target_optabs.
9606 * optabs-tree.c (init_tree_optimization_optabs): Use
9607 ggc_cleared_alloc to allocate target_optabs.
9608 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9609 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9611 * internal-fn.h (internal_load_fn_p): Declare.
9612 (internal_gather_scatter_fn_p): Likewise.
9613 (internal_fn_mask_index): Likewise.
9614 (internal_gather_scatter_fn_supported_p): Likewise.
9615 * internal-fn.c (gather_load_direct): New macro.
9616 (expand_gather_load_optab_fn): New function.
9617 (direct_gather_load_optab_supported_p): New macro.
9618 (direct_internal_fn_optab): New function.
9619 (internal_load_fn_p): Likewise.
9620 (internal_gather_scatter_fn_p): Likewise.
9621 (internal_fn_mask_index): Likewise.
9622 (internal_gather_scatter_fn_supported_p): Likewise.
9623 * optabs-query.c (supports_at_least_one_mode_p): New function.
9624 (supports_vec_gather_load_p): Likewise.
9625 * optabs-query.h (supports_vec_gather_load_p): Declare.
9626 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9627 and memory_type field.
9628 (NUM_PATTERNS): Bump to 15.
9629 * tree-vect-data-refs.c: Include internal-fn.h.
9630 (vect_gather_scatter_fn_p): New function.
9631 (vect_describe_gather_scatter_call): Likewise.
9632 (vect_check_gather_scatter): Try using internal functions for
9633 gather loads. Recognize existing calls to a gather load function.
9634 (vect_analyze_data_refs): Consider using gather loads if
9635 supports_vec_gather_load_p.
9636 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9637 (vect_get_gather_scatter_offset_type): Likewise.
9638 (vect_convert_mask_for_vectype): Likewise.
9639 (vect_add_conversion_to_patterm): Likewise.
9640 (vect_try_gather_scatter_pattern): Likewise.
9641 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9642 (vect_vect_recog_func_ptrs): Add it.
9643 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9644 internal_fn_mask_index and internal_gather_scatter_fn_p.
9645 (check_load_store_masking): Take the gather_scatter_info as an
9646 argument and handle gather loads.
9647 (vect_get_gather_scatter_ops): New function.
9648 (vectorizable_call): Check internal_load_fn_p.
9649 (vectorizable_load): Likewise. Handle gather load internal
9651 (vectorizable_store): Update call to check_load_store_masking.
9652 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9653 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9654 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9655 (aarch64_gather_scale_operand_d): New predicates.
9656 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9657 (mask_gather_load<mode>): New insns.
9659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9660 Alan Hayward <alan.hayward@arm.com>
9661 David Sherwood <david.sherwood@arm.com>
9663 * optabs.def (fold_left_plus_optab): New optab.
9664 * doc/md.texi (fold_left_plus_@var{m}): Document.
9665 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9666 * internal-fn.c (fold_left_direct): Define.
9667 (expand_fold_left_optab_fn): Likewise.
9668 (direct_fold_left_optab_supported_p): Likewise.
9669 * fold-const-call.c (fold_const_fold_left): New function.
9670 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9671 * tree-parloops.c (valid_reduction_p): New function.
9672 (gather_scalar_reductions): Use it.
9673 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9674 (vect_finish_replace_stmt): Declare.
9675 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9676 (needs_fold_left_reduction_p): New function, split out from...
9677 (vect_is_simple_reduction): ...here. Accept reductions that
9678 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9679 (vect_force_simple_reduction): Also store the reduction type in
9680 the assignment's STMT_VINFO_REDUC_TYPE.
9681 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9682 (merge_with_identity): New function.
9683 (vect_expand_fold_left): Likewise.
9684 (vectorize_fold_left_reduction): Likewise.
9685 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9686 scalar phi in place for it. Check for target support and reject
9687 cases that would reassociate the operation. Defer the transform
9688 phase to vectorize_fold_left_reduction.
9689 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9690 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9691 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9693 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9695 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9696 call to ifc_temp_var.
9698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9699 Alan Hayward <alan.hayward@arm.com>
9700 David Sherwood <david.sherwood@arm.com>
9702 * target.def (legitimize_address_displacement): Take the original
9703 offset as a poly_int.
9704 * targhooks.h (default_legitimize_address_displacement): Update
9706 * targhooks.c (default_legitimize_address_displacement): Likewise.
9707 * doc/tm.texi: Regenerate.
9708 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9709 as an argument, moving assert of ad->disp == ad->disp_term to...
9710 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9711 Try calling targetm.legitimize_address_displacement before expanding
9712 the address rather than afterwards, and adjust for the new interface.
9713 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9714 Match the new hook interface. Handle SVE addresses.
9715 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9720 * Makefile.in (OBJS): Add early-remat.o.
9721 * target.def (select_early_remat_modes): New hook.
9722 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9723 * doc/tm.texi: Regenerate.
9724 * targhooks.h (default_select_early_remat_modes): Declare.
9725 * targhooks.c (default_select_early_remat_modes): New function.
9726 * timevar.def (TV_EARLY_REMAT): New timevar.
9727 * passes.def (pass_early_remat): New pass.
9728 * tree-pass.h (make_pass_early_remat): Declare.
9729 * early-remat.c: New file.
9730 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9732 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9734 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9735 Alan Hayward <alan.hayward@arm.com>
9736 David Sherwood <david.sherwood@arm.com>
9738 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9739 vfm1 with a bound_epilog parameter.
9740 (vect_do_peeling): Update calls accordingly, and move the prologue
9741 call earlier in the function. Treat the base bound_epilog as 0 for
9742 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9743 this base when peeling for gaps.
9744 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9745 with fully-masked loops.
9746 (vect_estimate_min_profitable_iters): Handle the single peeled
9747 iteration in that case.
9749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9750 Alan Hayward <alan.hayward@arm.com>
9751 David Sherwood <david.sherwood@arm.com>
9753 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9754 single-element interleaving even if the size is not a power of 2.
9755 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9756 accesses for single-element interleaving if the group size is
9759 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9760 Alan Hayward <alan.hayward@arm.com>
9761 David Sherwood <david.sherwood@arm.com>
9763 * doc/md.texi (fold_extract_last_@var{m}): Document.
9764 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9765 * optabs.def (fold_extract_last_optab): New optab.
9766 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9767 * internal-fn.c (fold_extract_direct): New macro.
9768 (expand_fold_extract_optab_fn): Likewise.
9769 (direct_fold_extract_optab_supported_p): Likewise.
9770 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9771 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9772 EXTRACT_LAST_REDUCTION.
9773 (get_initial_def_for_reduction): Do not create an initial vector
9774 for EXTRACT_LAST_REDUCTION reductions.
9775 (vectorizable_reduction): Leave the scalar phi in place for
9776 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9777 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9778 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9779 transform phase to vectorizable_condition.
9780 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9782 (vect_finish_stmt_generation): ...here.
9783 (vect_finish_replace_stmt): New function.
9784 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9785 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9787 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9790 Alan Hayward <alan.hayward@arm.com>
9791 David Sherwood <david.sherwood@arm.com>
9793 * doc/md.texi (extract_last_@var{m}): Document.
9794 * optabs.def (extract_last_optab): New optab.
9795 * internal-fn.def (EXTRACT_LAST): New internal function.
9796 * internal-fn.c (cond_unary_direct): New macro.
9797 (expand_cond_unary_optab_fn): Likewise.
9798 (direct_cond_unary_optab_supported_p): Likewise.
9799 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9800 loops using EXTRACT_LAST.
9801 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9802 (extract_last_<mode>): ...this optab.
9803 (vec_extract<mode><Vel>): Update accordingly.
9805 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9806 Alan Hayward <alan.hayward@arm.com>
9807 David Sherwood <david.sherwood@arm.com>
9809 * target.def (empty_mask_is_expensive): New hook.
9810 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9811 * doc/tm.texi: Regenerate.
9812 * targhooks.h (default_empty_mask_is_expensive): Declare.
9813 * targhooks.c (default_empty_mask_is_expensive): New function.
9814 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9815 if the target says that empty masks are expensive.
9816 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9818 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9821 Alan Hayward <alan.hayward@arm.com>
9822 David Sherwood <david.sherwood@arm.com>
9824 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9825 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9826 (vect_use_loop_mask_for_alignment_p): New function.
9827 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9828 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9829 niters_skip argument. Make sure that the first niters_skip elements
9830 of the first iteration are inactive.
9831 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9832 Update call to vect_set_loop_masks_directly.
9833 (get_misalign_in_elems): New function, split out from...
9834 (vect_gen_prolog_loop_niters): ...here.
9835 (vect_update_init_of_dr): Take a code argument that specifies whether
9836 the adjustment should be added or subtracted.
9837 (vect_update_init_of_drs): Likewise.
9838 (vect_prepare_for_masked_peels): New function.
9839 (vect_do_peeling): Skip prologue peeling if we're using a mask
9840 instead. Update call to vect_update_inits_of_drs.
9841 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9843 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9844 alignment. Do not include the number of peeled iterations in
9845 the minimum threshold in that case.
9846 (vectorizable_induction): Adjust the start value down by
9847 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9848 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9849 Take the number of skipped iterations into account when calculating
9851 * tree-vect-stmts.c (vect_gen_while_not): New function.
9853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9854 Alan Hayward <alan.hayward@arm.com>
9855 David Sherwood <david.sherwood@arm.com>
9857 * doc/sourcebuild.texi (vect_fully_masked): Document.
9858 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9860 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9862 (vect_analyze_loop_2): ...here. Don't check the vectorization
9863 factor against the number of loop iterations if the loop is
9866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9867 Alan Hayward <alan.hayward@arm.com>
9868 David Sherwood <david.sherwood@arm.com>
9870 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9871 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9872 (dump_groups): Update accordingly.
9873 (iv_use::mem_type): New member variable.
9874 (address_p): New function.
9875 (record_use): Add a mem_type argument and initialize the new
9877 (record_group_use): Add a mem_type argument. Use address_p.
9878 Remove obsolete null checks of base_object. Update call to record_use.
9879 (find_interesting_uses_op): Update call to record_group_use.
9880 (find_interesting_uses_cond): Likewise.
9881 (find_interesting_uses_address): Likewise.
9882 (get_mem_type_for_internal_fn): New function.
9883 (find_address_like_use): Likewise.
9884 (find_interesting_uses_stmt): Try find_address_like_use before
9885 calling find_interesting_uses_op.
9886 (addr_offset_valid_p): Use the iv mem_type field as the type
9887 of the addressed memory.
9888 (add_autoinc_candidates): Likewise.
9889 (get_address_cost): Likewise.
9890 (split_small_address_groups_p): Use address_p.
9891 (split_address_groups): Likewise.
9892 (add_iv_candidate_for_use): Likewise.
9893 (autoinc_possible_for_pair): Likewise.
9894 (rewrite_groups): Likewise.
9895 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9896 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9897 (get_alias_ptr_type_for_ptr_address): New function.
9898 (rewrite_use_address): Rewrite address uses in calls that were
9899 identified by find_address_like_use.
9901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9902 Alan Hayward <alan.hayward@arm.com>
9903 David Sherwood <david.sherwood@arm.com>
9905 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9907 * gimple-expr.h (is_gimple_addressable: Likewise.
9908 * gimple-expr.c (is_gimple_address): Likewise.
9909 * internal-fn.c (expand_call_mem_ref): New function.
9910 (expand_mask_load_optab_fn): Use it.
9911 (expand_mask_store_optab_fn): Likewise.
9913 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9914 Alan Hayward <alan.hayward@arm.com>
9915 David Sherwood <david.sherwood@arm.com>
9917 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9918 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9919 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9920 (cond_umax@var{mode}): Document.
9921 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9922 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9923 (cond_umin_optab, cond_umax_optab): New optabs.
9924 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9925 (COND_IOR, COND_XOR): New internal functions.
9926 * internal-fn.h (get_conditional_internal_fn): Declare.
9927 * internal-fn.c (cond_binary_direct): New macro.
9928 (expand_cond_binary_optab_fn): Likewise.
9929 (direct_cond_binary_optab_supported_p): Likewise.
9930 (get_conditional_internal_fn): New function.
9931 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9932 Cope with reduction statements that are vectorized as calls rather
9934 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9935 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9936 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9937 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9938 (UNSPEC_COND_EOR): New unspecs.
9939 (optab): Add mappings for them.
9940 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9941 (sve_int_op, sve_fp_op): New int attributes.
9943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9944 Alan Hayward <alan.hayward@arm.com>
9945 David Sherwood <david.sherwood@arm.com>
9947 * optabs.def (while_ult_optab): New optab.
9948 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9949 * internal-fn.def (WHILE_ULT): New internal function.
9950 * internal-fn.h (direct_internal_fn_supported_p): New override
9951 that takes two types as argument.
9952 * internal-fn.c (while_direct): New macro.
9953 (expand_while_optab_fn): New function.
9954 (convert_optab_supported_p): Likewise.
9955 (direct_while_optab_supported_p): New macro.
9956 * wide-int.h (wi::udiv_ceil): New function.
9957 * tree-vectorizer.h (rgroup_masks): New structure.
9958 (vec_loop_masks): New typedef.
9959 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9961 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9962 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9963 (vect_max_vf): New function.
9964 (slpeel_make_loop_iterate_ntimes): Delete.
9965 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9966 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9967 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9968 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9969 internal-fn.h, stor-layout.h and optabs-query.h.
9970 (vect_set_loop_mask): New function.
9971 (add_preheader_seq): Likewise.
9972 (add_header_seq): Likewise.
9973 (interleave_supported_p): Likewise.
9974 (vect_maybe_permute_loop_masks): Likewise.
9975 (vect_set_loop_masks_directly): Likewise.
9976 (vect_set_loop_condition_masked): Likewise.
9977 (vect_set_loop_condition_unmasked): New function, split out from
9978 slpeel_make_loop_iterate_ntimes.
9979 (slpeel_make_loop_iterate_ntimes): Rename to..
9980 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9981 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9982 (vect_do_peeling): Update call accordingly.
9983 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9985 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9986 mask_compare_type, can_fully_mask_p and fully_masked_p.
9987 (release_vec_loop_masks): New function.
9988 (_loop_vec_info): Use it to free the loop masks.
9989 (can_produce_all_loop_masks_p): New function.
9990 (vect_get_max_nscalars_per_iter): Likewise.
9991 (vect_verify_full_masking): Likewise.
9992 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9993 retries, and free the mask rgroups before retrying. Check loop-wide
9994 reasons for disallowing fully-masked loops. Make the final decision
9995 about whether use a fully-masked loop or not.
9996 (vect_estimate_min_profitable_iters): Do not assume that peeling
9997 for the number of iterations will be needed for fully-masked loops.
9998 (vectorizable_reduction): Disable fully-masked loops.
9999 (vectorizable_live_operation): Likewise.
10000 (vect_halve_mask_nunits): New function.
10001 (vect_double_mask_nunits): Likewise.
10002 (vect_record_loop_mask): Likewise.
10003 (vect_get_loop_mask): Likewise.
10004 (vect_transform_loop): Handle the case in which the final loop
10005 iteration might handle a partial vector. Call vect_set_loop_condition
10006 instead of slpeel_make_loop_iterate_ntimes.
10007 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10008 (check_load_store_masking): New function.
10009 (prepare_load_store_mask): Likewise.
10010 (vectorizable_store): Handle fully-masked loops.
10011 (vectorizable_load): Likewise.
10012 (supportable_widening_operation): Use vect_halve_mask_nunits for
10014 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10015 (vect_gen_while): New function.
10016 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10017 (aarch64_uqdec<mode>): New insn.
10019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10020 Alan Hayward <alan.hayward@arm.com>
10021 David Sherwood <david.sherwood@arm.com>
10023 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10024 (reduc_xor_scal_optab): New optabs.
10025 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10026 (reduc_xor_scal_@var{m}): Document.
10027 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10028 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10029 internal functions.
10030 * fold-const-call.c (fold_const_call): Handle them.
10031 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10032 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10033 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10034 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10035 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10036 (UNSPEC_XORV): New unspecs.
10037 (optab): Add entries for them.
10038 (BITWISEV): New int iterator.
10039 (bit_reduc_op): New int attributes.
10041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10042 Alan Hayward <alan.hayward@arm.com>
10043 David Sherwood <david.sherwood@arm.com>
10045 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10046 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10047 * optabs.def (vec_shl_insert_optab): New optab.
10048 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10049 (duplicate_and_interleave): Likewise.
10050 * tree-vect-loop.c: Include internal-fn.h.
10051 (neutral_op_for_slp_reduction): New function, split out from
10052 get_initial_defs_for_reduction.
10053 (get_initial_def_for_reduction): Handle option 2 for variable-length
10054 vectors by loading the neutral value into a vector and then shifting
10055 the initial value into element 0.
10056 (get_initial_defs_for_reduction): Replace the code argument with
10057 the neutral value calculated by neutral_op_for_slp_reduction.
10058 Use gimple_build_vector for constant-length vectors.
10059 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10060 but the first group_size elements have a neutral value.
10061 Use duplicate_and_interleave otherwise.
10062 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10063 Update call to get_initial_defs_for_reduction. Handle SLP
10064 reductions for variable-length vectors by creating one vector
10065 result for each scalar result, with the elements associated
10066 with other scalar results stubbed out with the neutral value.
10067 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10068 Require IFN_VEC_SHL_INSERT for double reductions on
10069 variable-length vectors, or SLP reductions that have
10070 a neutral value. Require can_duplicate_and_interleave_p
10071 support for variable-length unchained SLP reductions if there
10072 is no neutral value, such as for MIN/MAX reductions. Also require
10073 the number of vector elements to be a multiple of the number of
10074 SLP statements when doing variable-length unchained SLP reductions.
10075 Update call to vect_create_epilog_for_reduction.
10076 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10077 and remove initial values.
10078 (duplicate_and_interleave): Make public.
10079 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10080 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10082 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10083 Alan Hayward <alan.hayward@arm.com>
10084 David Sherwood <david.sherwood@arm.com>
10086 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10087 (can_duplicate_and_interleave_p): New function.
10088 (vect_get_and_check_slp_defs): Take the vector of statements
10089 rather than just the current one. Remove excess parentheses.
10090 Restriction rejectinon of vect_constant_def and vect_external_def
10091 for variable-length vectors to boolean types, or types for which
10092 can_duplicate_and_interleave_p is false.
10093 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10094 (duplicate_and_interleave): New function.
10095 (vect_get_constant_vectors): Use gimple_build_vector for
10096 constant-length vectors and suitable variable-length constant
10097 vectors. Use duplicate_and_interleave for other variable-length
10098 vectors. Don't defer the update when inserting new statements.
10100 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10101 Alan Hayward <alan.hayward@arm.com>
10102 David Sherwood <david.sherwood@arm.com>
10104 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10105 min_profitable_iters doesn't go negative.
10107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10108 Alan Hayward <alan.hayward@arm.com>
10109 David Sherwood <david.sherwood@arm.com>
10111 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10112 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10113 * optabs.def (vec_mask_load_lanes_optab): New optab.
10114 (vec_mask_store_lanes_optab): Likewise.
10115 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10116 (MASK_STORE_LANES): Likewise.
10117 * internal-fn.c (mask_load_lanes_direct): New macro.
10118 (mask_store_lanes_direct): Likewise.
10119 (expand_mask_load_optab_fn): Handle masked operations.
10120 (expand_mask_load_lanes_optab_fn): New macro.
10121 (expand_mask_store_optab_fn): Handle masked operations.
10122 (expand_mask_store_lanes_optab_fn): New macro.
10123 (direct_mask_load_lanes_optab_supported_p): Likewise.
10124 (direct_mask_store_lanes_optab_supported_p): Likewise.
10125 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10127 (vect_load_lanes_supported): Likewise.
10128 * tree-vect-data-refs.c (strip_conversion): New function.
10129 (can_group_stmts_p): Likewise.
10130 (vect_analyze_data_ref_accesses): Use it instead of checking
10131 for a pair of assignments.
10132 (vect_store_lanes_supported): Take a masked_p parameter.
10133 (vect_load_lanes_supported): Likewise.
10134 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10135 vect_store_lanes_supported and vect_load_lanes_supported.
10136 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10137 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10138 parameter. Don't allow gaps for masked accesses.
10139 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10140 and vect_load_lanes_supported.
10141 (get_load_store_type): Take a masked_p parameter and update
10142 call to get_group_load_store_type.
10143 (vectorizable_store): Update call to get_load_store_type.
10144 Handle IFN_MASK_STORE_LANES.
10145 (vectorizable_load): Update call to get_load_store_type.
10146 Handle IFN_MASK_LOAD_LANES.
10148 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10149 Alan Hayward <alan.hayward@arm.com>
10150 David Sherwood <david.sherwood@arm.com>
10152 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10154 * config/aarch64/aarch64-protos.h
10155 (aarch64_sve_struct_memory_operand_p): Declare.
10156 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10157 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10158 (VPRED, vpred): Handle SVE structure modes.
10159 * config/aarch64/constraints.md (Utx): New constraint.
10160 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10161 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10162 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10163 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10164 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10165 structure modes. Split into pieces after RA.
10166 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10167 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10169 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10170 SVE structure modes.
10171 (aarch64_classify_address): Likewise.
10172 (sizetochar): Move earlier in file.
10173 (aarch64_print_operand): Handle SVE register lists.
10174 (aarch64_array_mode): New function.
10175 (aarch64_sve_struct_memory_operand_p): Likewise.
10176 (TARGET_ARRAY_MODE): Redefine.
10178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10179 Alan Hayward <alan.hayward@arm.com>
10180 David Sherwood <david.sherwood@arm.com>
10182 * target.def (array_mode): New target hook.
10183 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10184 * doc/tm.texi: Regenerate.
10185 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10186 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10187 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10188 targetm.array_mode.
10189 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10192 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10193 Alan Hayward <alan.hayward@arm.com>
10194 David Sherwood <david.sherwood@arm.com>
10196 * fold-const.c (fold_binary_loc): Check the argument types
10197 rather than the result type when testing for a vector operation.
10199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10201 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10202 * doc/tm.texi: Regenerate.
10204 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10205 Alan Hayward <alan.hayward@arm.com>
10206 David Sherwood <david.sherwood@arm.com>
10208 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10209 (sve): Document new AArch64 extension.
10210 * doc/md.texi (w): Extend the description of the AArch64
10211 constraint to include SVE vectors.
10212 (Upl, Upa): Document new AArch64 predicate constraints.
10213 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10215 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10216 (msve-vector-bits=): New option.
10217 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10218 SVE when these are disabled.
10219 (sve): New extension.
10220 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10221 modes. Adjust their number of units based on aarch64_sve_vg.
10222 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10223 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10224 aarch64_addr_query_type.
10225 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10226 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10227 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10228 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10229 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10230 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10231 (aarch64_simd_imm_zero_p): Delete.
10232 (aarch64_check_zero_based_sve_index_immediate): Declare.
10233 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10234 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10235 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10236 (aarch64_sve_float_mul_immediate_p): Likewise.
10237 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10238 rather than an rtx.
10239 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10240 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10241 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10242 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10243 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10244 (aarch64_regmode_natural_size): Likewise.
10245 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10246 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10248 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10249 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10250 for VG and the SVE predicate registers.
10251 (V_ALIASES): Add a "z"-prefixed alias.
10252 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10253 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10254 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10255 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10256 (REG_CLASS_NAMES): Add entries for them.
10257 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10258 and the predicate registers.
10259 (aarch64_sve_vg): Declare.
10260 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10261 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10262 (REGMODE_NATURAL_SIZE): Define.
10263 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10265 * config/aarch64/aarch64.c: Include cfgrtl.h.
10266 (simd_immediate_info): Add a constructor for series vectors,
10267 and an associated step field.
10268 (aarch64_sve_vg): New variable.
10269 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10270 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10271 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10272 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10273 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10274 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10275 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10276 (aarch64_get_mask_mode): New functions.
10277 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10278 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10279 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10280 predicate modes and predicate registers. Explicitly restrict
10281 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10282 to store a vector mode if it is recognized by
10283 aarch64_classify_vector_mode.
10284 (aarch64_regmode_natural_size): New function.
10285 (aarch64_hard_regno_caller_save_mode): Return the original mode
10287 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10288 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10289 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10290 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10292 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10293 does not overlap dest if the function is frame-related. Handle
10295 (aarch64_split_add_offset): New function.
10296 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10297 them aarch64_add_offset.
10298 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10299 and update call to aarch64_sub_sp.
10300 (aarch64_add_cfa_expression): New function.
10301 (aarch64_expand_prologue): Pass extra temporary registers to the
10302 functions above. Handle the case in which we need to emit new
10303 DW_CFA_expressions for registers that were originally saved
10304 relative to the stack pointer, but now have to be expressed
10305 relative to the frame pointer.
10306 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10308 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10309 IP0 and IP1 values for SVE frames.
10310 (aarch64_expand_vec_series): New function.
10311 (aarch64_expand_sve_widened_duplicate): Likewise.
10312 (aarch64_expand_sve_const_vector): Likewise.
10313 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10314 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10315 into the register, rather than emitting a SET directly.
10316 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10317 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10318 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10319 (offset_9bit_signed_scaled_p): New functions.
10320 (aarch64_replicate_bitmask_imm): New function.
10321 (aarch64_bitmask_imm): Use it.
10322 (aarch64_cannot_force_const_mem): Reject expressions involving
10323 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10324 (aarch64_classify_index): Handle SVE indices, by requiring
10325 a plain register index with a scale that matches the element size.
10326 (aarch64_classify_address): Handle SVE addresses. Assert that
10327 the mode of the address is VOIDmode or an integer mode.
10328 Update call to aarch64_classify_symbol.
10329 (aarch64_classify_symbolic_expression): Update call to
10330 aarch64_classify_symbol.
10331 (aarch64_const_vec_all_in_range_p): New function.
10332 (aarch64_print_vector_float_operand): Likewise.
10333 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10334 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10335 and the FP immediates 1.0 and 0.5.
10336 (aarch64_print_address_internal): Handle SVE addresses.
10337 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10338 (aarch64_regno_regclass): Handle predicate registers.
10339 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10341 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10342 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10343 (aarch64_convert_sve_vector_bits): New function.
10344 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10345 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10346 rather than an rtx.
10347 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10348 Handle SVE vector and predicate modes. Accept VL-based constants
10349 that need only one temporary register, and VL offsets that require
10350 no temporary registers.
10351 (aarch64_conditional_register_usage): Mark the predicate registers
10352 as fixed if SVE isn't available.
10353 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10354 Return true for SVE vector and predicate modes.
10355 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10356 rather than an unsigned int. Handle SVE modes.
10357 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10359 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10361 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10362 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10363 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10364 (aarch64_sve_float_mul_immediate_p): New functions.
10365 (aarch64_sve_valid_immediate): New function.
10366 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10367 Explicitly reject structure modes. Check for INDEX constants.
10368 Handle PTRUE and PFALSE constants.
10369 (aarch64_check_zero_based_sve_index_immediate): New function.
10370 (aarch64_simd_imm_zero_p): Delete.
10371 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10372 vector modes. Accept constants in the range of CNT[BHWD].
10373 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10374 ask for an Advanced SIMD mode.
10375 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10376 (aarch64_simd_vector_alignment): Handle SVE predicates.
10377 (aarch64_vectorize_preferred_vector_alignment): New function.
10378 (aarch64_simd_vector_alignment_reachable): Use it instead of
10380 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10381 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10383 (MAX_VECT_LEN): Delete.
10384 (expand_vec_perm_d): Add a vec_flags field.
10385 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10386 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10387 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10389 (aarch64_evpc_rev): Rename to...
10390 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10391 (aarch64_evpc_rev_global): New function.
10392 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10393 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10395 (aarch64_evpc_sve_tbl): New function.
10396 (aarch64_expand_vec_perm_const_1): Update after rename of
10397 aarch64_evpc_rev. Handle SVE permutes too, trying
10398 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10399 than aarch64_evpc_tbl.
10400 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10401 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10402 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10403 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10404 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10405 (aarch64_expand_sve_vcond): New functions.
10406 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10407 of aarch64_vector_mode_p.
10408 (aarch64_dwarf_poly_indeterminate_value): New function.
10409 (aarch64_compute_pressure_classes): Likewise.
10410 (aarch64_can_change_mode_class): Likewise.
10411 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10412 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10413 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10414 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10415 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10416 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10417 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10418 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10420 (Dn, Dl, Dr): Accept const as well as const_vector.
10421 (Dz): Likewise. Compare against CONST0_RTX.
10422 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10423 of "vector" where appropriate.
10424 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10425 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10426 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10427 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10428 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10429 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10430 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10431 (v_int_equiv): Extend to SVE modes.
10432 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10434 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10435 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10436 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10437 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10438 (SVE_COND_FP_CMP): New int iterators.
10439 (perm_hilo): Handle the new unpack unspecs.
10440 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10442 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10443 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10444 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10445 (aarch64_equality_operator, aarch64_constant_vector_operand)
10446 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10447 (aarch64_sve_nonimmediate_operand): Likewise.
10448 (aarch64_sve_general_operand): Likewise.
10449 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10450 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10451 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10452 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10453 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10454 (aarch64_sve_float_arith_immediate): Likewise.
10455 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10456 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10457 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10458 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10459 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10460 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10461 (aarch64_sve_float_arith_operand): Likewise.
10462 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10463 (aarch64_sve_float_mul_operand): Likewise.
10464 (aarch64_sve_vec_perm_operand): Likewise.
10465 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10466 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10467 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10468 as well as const_vector.
10469 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10470 in file. Use CONST0_RTX and CONSTM1_RTX.
10471 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10472 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10473 Use aarch64_simd_imm_zero.
10474 * config/aarch64/aarch64-sve.md: New file.
10475 * config/aarch64/aarch64.md: Include it.
10476 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10477 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10478 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10479 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10480 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10481 (sve): New attribute.
10482 (enabled): Disable instructions with the sve attribute unless
10484 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10485 aarch64_expand_mov_immediate.
10486 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10487 CNT[BHSD] immediates.
10488 (movti): Split CONST_POLY_INT moves into two halves.
10489 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10490 Split additions that need a temporary here if the destination
10491 is the stack pointer.
10492 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10493 (*add<mode>3_poly_1): New instruction.
10494 (set_clobber_cc): New expander.
10496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10498 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10499 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10500 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10501 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10502 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10503 Change innermode from fixed_mode_size to machine_mode.
10504 (simplify_subreg): Update call accordingly. Handle a constant-sized
10505 subreg of a variable-length CONST_VECTOR.
10507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10508 Alan Hayward <alan.hayward@arm.com>
10509 David Sherwood <david.sherwood@arm.com>
10511 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10512 (add_offset_to_base): New function, split out from...
10513 (create_mem_ref): ...here. When handling a scale other than 1,
10514 check first whether the address is valid without the offset.
10515 Add it into the base if so, leaving the index and scale as-is.
10517 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10520 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10521 fold_for_warn before checking if arg2 is INTEGER_CST.
10523 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10525 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10526 (store_multiple_operation): Delete.
10527 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10528 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10529 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10530 guarded by TARGET_STRING.
10531 (rs6000_output_load_multiple): Delete.
10532 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10533 OPTION_MASK_STRING / TARGET_STRING handling.
10534 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10535 (const rs6000_opt_masks) <"string">: Change mask to 0.
10536 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10537 (MASK_STRING): Delete.
10538 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10540 (load_multiple): Delete.
10547 (store_multiple): Delete.
10554 (movmemsi_8reg): Delete.
10555 (corresponding unnamed define_insn): Delete.
10556 (movmemsi_6reg): Delete.
10557 (corresponding unnamed define_insn): Delete.
10558 (movmemsi_4reg): Delete.
10559 (corresponding unnamed define_insn): Delete.
10560 (movmemsi_2reg): Delete.
10561 (corresponding unnamed define_insn): Delete.
10562 (movmemsi_1reg): Delete.
10563 (corresponding unnamed define_insn): Delete.
10564 * config/rs6000/rs6000.opt (mno-string): New.
10565 (mstring): Replace by deprecation warning stub.
10566 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10568 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10570 * regrename.c (regrename_do_replace): If replacing the same
10571 reg multiple times, try to reuse last created gen_raw_REG.
10574 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10575 main to workaround a bug in GDB.
10577 2018-01-12 Tom de Vries <tom@codesourcery.com>
10580 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10582 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10584 PR rtl-optimization/80481
10585 * ira-color.c (get_cap_member): New function.
10586 (allocnos_conflict_by_live_ranges_p): Use it.
10587 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10588 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10590 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10593 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10594 (*saddl_se_1): Ditto.
10595 (*ssubsi_1): Ditto.
10596 (*ssubl_se_1): Ditto.
10598 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10600 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10601 rather than wi::to_widest for DR_INITs.
10602 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10603 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10604 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10606 (vect_analyze_group_access_1): Note that here.
10608 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10610 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10611 polynomial type sizes.
10613 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10615 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10616 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10617 (gimple_add_tmp_var): Likewise.
10619 2018-01-12 Martin Liska <mliska@suse.cz>
10621 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10622 (gimple_alloc_sizes): Likewise.
10623 (dump_gimple_statistics): Use PRIu64 in printf format.
10624 * gimple.h: Change uint64_t to int.
10626 2018-01-12 Martin Liska <mliska@suse.cz>
10628 * tree-core.h: Use uint64_t instead of int.
10629 * tree.c (tree_node_counts): Likewise.
10630 (tree_node_sizes): Likewise.
10631 (dump_tree_statistics): Use PRIu64 in printf format.
10633 2018-01-12 Martin Liska <mliska@suse.cz>
10635 * Makefile.in: As qsort_chk is implemented in vec.c, add
10636 vec.o to linkage of gencfn-macros.
10637 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10638 passing the info to record_node_allocation_statistics.
10639 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10641 * ggc-common.c (struct ggc_usage): Add operator== and use
10642 it in operator< and compare function.
10643 * mem-stats.h (struct mem_usage): Likewise.
10644 * vec.c (struct vec_usage): Remove operator< and compare
10645 function. Can be simply inherited.
10647 2018-01-12 Martin Jambor <mjambor@suse.cz>
10650 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10651 * tree-ssa-math-opts.c: Include domwalk.h.
10652 (convert_mult_to_fma_1): New function.
10653 (fma_transformation_info): New type.
10654 (fma_deferring_state): Likewise.
10655 (cancel_fma_deferring): New function.
10656 (result_of_phi): Likewise.
10657 (last_fma_candidate_feeds_initial_phi): Likewise.
10658 (convert_mult_to_fma): Added deferring logic, split actual
10659 transformation to convert_mult_to_fma_1.
10660 (math_opts_dom_walker): New type.
10661 (math_opts_dom_walker::after_dom_children): New method, body moved
10662 here from pass_optimize_widening_mul::execute, added deferring logic
10664 (pass_optimize_widening_mul::execute): Moved most of code to
10665 math_opts_dom_walker::after_dom_children.
10666 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10667 * config/i386/i386.c (ix86_option_override_internal): Added
10668 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10670 2018-01-12 Richard Biener <rguenther@suse.de>
10673 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10674 inline instance vars.
10676 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10679 * config/rx/rx.c (rx_is_restricted_memory_address):
10680 Handle SUBREG case.
10682 2018-01-12 Richard Biener <rguenther@suse.de>
10684 PR tree-optimization/80846
10685 * target.def (split_reduction): New target hook.
10686 * targhooks.c (default_split_reduction): New function.
10687 * targhooks.h (default_split_reduction): Declare.
10688 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10689 target requests first reduce vectors by combining low and high
10691 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10692 (get_vectype_for_scalar_type_and_size): Export.
10693 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10694 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10695 * doc/tm.texi: Regenerate.
10696 * config/i386/i386.c (ix86_split_reduction): Implement
10697 TARGET_VECTORIZE_SPLIT_REDUCTION.
10699 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10702 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10703 in PIC mode except for TARGET_VXWORKS_RTP.
10704 * config/sparc/sparc.c: Include cfgrtl.h.
10705 (TARGET_INIT_PIC_REG): Define.
10706 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10707 (sparc_pic_register_p): New predicate.
10708 (sparc_legitimate_address_p): Use it.
10709 (sparc_legitimize_pic_address): Likewise.
10710 (sparc_delegitimize_address): Likewise.
10711 (sparc_mode_dependent_address_p): Likewise.
10712 (gen_load_pcrel_sym): Remove 4th parameter.
10713 (load_got_register): Adjust call to above. Remove obsolete stuff.
10714 (sparc_expand_prologue): Do not call load_got_register here.
10715 (sparc_flat_expand_prologue): Likewise.
10716 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10717 (sparc_use_pseudo_pic_reg): New function.
10718 (sparc_init_pic_reg): Likewise.
10719 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10720 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10722 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10724 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10725 Add item for branch_cost.
10727 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10729 PR rtl-optimization/83565
10730 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10731 not extend the result to a larger mode for rotate operations.
10732 (num_sign_bit_copies1): Likewise.
10734 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10737 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10739 Use values-Xc.o for -pedantic.
10740 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10742 2018-01-12 Martin Liska <mliska@suse.cz>
10745 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10747 (possible_polymorphic_call_targets): Use it.
10748 (ipa_devirt): Likewise.
10750 2018-01-12 Martin Liska <mliska@suse.cz>
10752 * profile-count.h (enum profile_quality): Use 0 as invalid
10753 enum value of profile_quality.
10755 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10757 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10758 -mext-string options.
10760 2018-01-12 Richard Biener <rguenther@suse.de>
10762 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10763 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10764 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10766 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10768 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10770 * configure.ac (--with-long-double-format): Add support for the
10771 configuration option to change the default long double format on
10773 * config.gcc (powerpc*-linux*-*): Likewise.
10774 * configure: Regenerate.
10775 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10776 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10777 used without modification.
10779 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10781 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10782 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10783 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10784 MISC_BUILTIN_SPEC_BARRIER.
10785 (rs6000_init_builtins): Likewise.
10786 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10788 (speculation_barrier): New define_insn.
10789 * doc/extend.texi: Document __builtin_speculation_barrier.
10791 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10794 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10795 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10796 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10798 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10799 integral modes instead of "ss" and "sd".
10800 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10801 vectors with 32-bit and 64-bit elements.
10802 (vecdupssescalarmodesuffix): New mode attribute.
10803 (vec_dup<mode>): Use it.
10805 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10808 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10809 frame if argument is passed on stack.
10811 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10814 * ree.c (combine_reaching_defs): Optimize also
10815 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10816 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10818 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10820 PR middle-end/83189
10821 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10823 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10825 PR middle-end/83718
10826 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10827 after they are computed.
10829 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10831 PR tree-optimization/83695
10832 * gimple-loop-linterchange.cc
10833 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10834 reset cached scev information after interchange.
10835 (pass_linterchange::execute): Remove call to scev_reset_htab.
10837 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10839 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10840 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10841 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10842 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10843 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10844 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10845 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10846 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10847 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10848 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10849 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10850 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10851 (V_lane_reg): Likewise.
10852 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10854 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10855 (vfmal_lane_low<mode>_intrinsic,
10856 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10857 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10858 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10859 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10860 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10861 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10863 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10865 * config/arm/arm-cpus.in (fp16fml): New feature.
10866 (ALL_SIMD): Add fp16fml.
10867 (armv8.2-a): Add fp16fml as an option.
10868 (armv8.3-a): Likewise.
10869 (armv8.4-a): Add fp16fml as part of fp16.
10870 * config/arm/arm.h (TARGET_FP16FML): Define.
10871 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10873 * config/arm/arm-modes.def (V2HF): Define.
10874 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10875 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10876 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10877 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10878 vfmsl_low, vfmsl_high): New set of builtins.
10879 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10880 (vfml_op): New code attribute.
10881 (VFMLHALVES): New int iterator.
10882 (VFML, VFMLSEL): New mode attributes.
10883 (V_reg): Define mapping for V2HF.
10884 (V_hi, V_lo): New mode attributes.
10885 (VF_constraint): Likewise.
10886 (vfml_half, vfml_half_selector): New int attributes.
10887 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10889 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10890 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10892 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10893 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10894 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10895 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10897 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10898 Document new effective target and option set.
10900 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10902 * config/arm/arm-cpus.in (armv8_4): New feature.
10903 (ARMv8_4a): New fgroup.
10904 (armv8.4-a): New arch.
10905 * config/arm/arm-tables.opt: Regenerate.
10906 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10907 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10908 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10909 Add matching rules for -march=armv8.4-a and extensions.
10910 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10912 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10915 * config/rx/rx.md (BW): New mode attribute.
10916 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10918 2018-01-11 Richard Biener <rguenther@suse.de>
10920 PR tree-optimization/83435
10921 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10922 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10923 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10925 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10926 Alan Hayward <alan.hayward@arm.com>
10927 David Sherwood <david.sherwood@arm.com>
10929 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10931 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10932 (aarch64_print_address_internal): Use it to check for a zero offset.
10934 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10935 Alan Hayward <alan.hayward@arm.com>
10936 David Sherwood <david.sherwood@arm.com>
10938 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10939 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10940 Return a poly_int64 rather than a HOST_WIDE_INT.
10941 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10942 rather than a HOST_WIDE_INT.
10943 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10944 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10945 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10946 final_offset from HOST_WIDE_INT to poly_int64.
10947 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10948 to_constant when getting the number of units in an Advanced SIMD
10950 (aarch64_builtin_vectorized_function): Check for a constant number
10952 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10954 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10955 attribute instead of GET_MODE_NUNITS.
10956 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10957 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10958 GET_MODE_SIZE for fixed-size registers.
10959 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10960 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10961 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10962 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10963 (aarch64_print_operand, aarch64_print_address_internal)
10964 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10965 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10966 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10967 Handle polynomial GET_MODE_SIZE.
10968 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10969 wider than SImode without modification.
10970 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10971 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10972 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10973 passing and returning SVE modes.
10974 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10975 rather than GEN_INT.
10976 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10977 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10978 (aarch64_allocate_and_probe_stack_space): Likewise.
10979 (aarch64_layout_frame): Cope with polynomial offsets.
10980 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10981 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10982 polynomial offsets.
10983 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10984 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10985 poly_int64 rather than a HOST_WIDE_INT.
10986 (aarch64_get_separate_components, aarch64_process_components)
10987 (aarch64_expand_prologue, aarch64_expand_epilogue)
10988 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10989 (aarch64_anchor_offset): New function, split out from...
10990 (aarch64_legitimize_address): ...here.
10991 (aarch64_builtin_vectorization_cost): Handle polynomial
10992 TYPE_VECTOR_SUBPARTS.
10993 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10995 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10996 number of elements from the PARALLEL rather than the mode.
10997 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10998 rather than GET_MODE_BITSIZE.
10999 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11000 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11001 (aarch64_expand_vec_perm_const_1): Handle polynomial
11002 d->perm.length () and d->perm elements.
11003 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11004 Apply to_constant to d->perm elements.
11005 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11006 polynomial CONST_VECTOR_NUNITS.
11007 (aarch64_move_pointer): Take amount as a poly_int64 rather
11009 (aarch64_progress_pointer): Avoid temporary variable.
11010 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11011 the mode attribute instead of GET_MODE.
11013 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11014 Alan Hayward <alan.hayward@arm.com>
11015 David Sherwood <david.sherwood@arm.com>
11017 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11018 x exists before using it.
11019 (aarch64_add_constant_internal): Rename to...
11020 (aarch64_add_offset_1): ...this. Replace regnum with separate
11021 src and dest rtxes. Handle the case in which they're different,
11022 including when the offset is zero. Replace scratchreg with an rtx.
11023 Use 2 additions if there is no spare register into which we can
11024 move a 16-bit constant.
11025 (aarch64_add_constant): Delete.
11026 (aarch64_add_offset): Replace reg with separate src and dest
11027 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11028 Use aarch64_add_offset_1.
11029 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11030 an rtx rather than an int. Take the delta as a poly_int64
11031 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11032 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11033 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11034 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11035 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11036 and aarch64_add_sp.
11037 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11038 aarch64_add_constant.
11040 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11042 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11043 Use scalar_float_mode.
11045 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11047 * config/aarch64/aarch64-simd.md
11048 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11049 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11050 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11051 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11052 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11053 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11054 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11055 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11056 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11057 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11059 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11062 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11063 targ_options->x_arm_arch_string is non NULL.
11065 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11067 * config/aarch64/aarch64.h
11068 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11070 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11073 * expmed.c (emit_store_flag_force): Swap if const op0
11074 and change VOIDmode to mode of op0.
11076 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11078 PR rtl-optimization/83761
11079 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11080 than bytes to mode_for_size.
11082 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11084 PR middle-end/83189
11085 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11086 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11089 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11091 PR middle-end/83575
11092 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11093 when in layout mode.
11094 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11095 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11098 2018-01-10 Michael Collison <michael.collison@arm.com>
11100 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11101 * config/aarch64/aarch64-option-extension.def: Add
11102 AARCH64_OPT_EXTENSION of 'fp16fml'.
11103 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11104 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11105 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11106 * config/aarch64/constraints.md (Ui7): New constraint.
11107 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11108 (VFMLA_SEL_W): Ditto.
11111 (VFMLA16_LOW): New int iterator.
11112 (VFMLA16_HIGH): Ditto.
11113 (UNSPEC_FMLAL): New unspec.
11114 (UNSPEC_FMLSL): Ditto.
11115 (UNSPEC_FMLAL2): Ditto.
11116 (UNSPEC_FMLSL2): Ditto.
11117 (f16mac): New code attribute.
11118 * config/aarch64/aarch64-simd-builtins.def
11119 (aarch64_fmlal_lowv2sf): Ditto.
11120 (aarch64_fmlsl_lowv2sf): Ditto.
11121 (aarch64_fmlalq_lowv4sf): Ditto.
11122 (aarch64_fmlslq_lowv4sf): Ditto.
11123 (aarch64_fmlal_highv2sf): Ditto.
11124 (aarch64_fmlsl_highv2sf): Ditto.
11125 (aarch64_fmlalq_highv4sf): Ditto.
11126 (aarch64_fmlslq_highv4sf): Ditto.
11127 (aarch64_fmlal_lane_lowv2sf): Ditto.
11128 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11129 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11130 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11131 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11132 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11133 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11134 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11135 (aarch64_fmlal_lane_highv2sf): Ditto.
11136 (aarch64_fmlsl_lane_highv2sf): Ditto.
11137 (aarch64_fmlal_laneq_highv2sf): Ditto.
11138 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11139 (aarch64_fmlalq_lane_highv4sf): Ditto.
11140 (aarch64_fmlsl_lane_highv4sf): Ditto.
11141 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11142 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11143 * config/aarch64/aarch64-simd.md:
11144 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11145 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11146 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11147 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11148 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11149 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11150 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11151 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11152 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11153 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11154 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11155 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11156 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11157 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11158 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11159 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11160 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11161 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11162 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11163 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11164 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11165 (vfmlsl_low_u32): Ditto.
11166 (vfmlalq_low_u32): Ditto.
11167 (vfmlslq_low_u32): Ditto.
11168 (vfmlal_high_u32): Ditto.
11169 (vfmlsl_high_u32): Ditto.
11170 (vfmlalq_high_u32): Ditto.
11171 (vfmlslq_high_u32): Ditto.
11172 (vfmlal_lane_low_u32): Ditto.
11173 (vfmlsl_lane_low_u32): Ditto.
11174 (vfmlal_laneq_low_u32): Ditto.
11175 (vfmlsl_laneq_low_u32): Ditto.
11176 (vfmlalq_lane_low_u32): Ditto.
11177 (vfmlslq_lane_low_u32): Ditto.
11178 (vfmlalq_laneq_low_u32): Ditto.
11179 (vfmlslq_laneq_low_u32): Ditto.
11180 (vfmlal_lane_high_u32): Ditto.
11181 (vfmlsl_lane_high_u32): Ditto.
11182 (vfmlal_laneq_high_u32): Ditto.
11183 (vfmlsl_laneq_high_u32): Ditto.
11184 (vfmlalq_lane_high_u32): Ditto.
11185 (vfmlslq_lane_high_u32): Ditto.
11186 (vfmlalq_laneq_high_u32): Ditto.
11187 (vfmlslq_laneq_high_u32): Ditto.
11188 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11189 (AARCH64_FL_FOR_ARCH8_4): New.
11190 (AARCH64_ISA_F16FML): New ISA flag.
11191 (TARGET_F16FML): New feature flag for fp16fml.
11192 (doc/invoke.texi): Document new fp16fml option.
11194 2018-01-10 Michael Collison <michael.collison@arm.com>
11196 * config/aarch64/aarch64-builtins.c:
11197 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11198 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11199 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11200 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11201 (AARCH64_ISA_SHA3): New ISA flag.
11202 (TARGET_SHA3): New feature flag for sha3.
11203 * config/aarch64/iterators.md (sha512_op): New int attribute.
11204 (CRYPTO_SHA512): New int iterator.
11205 (UNSPEC_SHA512H): New unspec.
11206 (UNSPEC_SHA512H2): Ditto.
11207 (UNSPEC_SHA512SU0): Ditto.
11208 (UNSPEC_SHA512SU1): Ditto.
11209 * config/aarch64/aarch64-simd-builtins.def
11210 (aarch64_crypto_sha512hqv2di): New builtin.
11211 (aarch64_crypto_sha512h2qv2di): Ditto.
11212 (aarch64_crypto_sha512su0qv2di): Ditto.
11213 (aarch64_crypto_sha512su1qv2di): Ditto.
11214 (aarch64_eor3qv8hi): Ditto.
11215 (aarch64_rax1qv2di): Ditto.
11216 (aarch64_xarqv2di): Ditto.
11217 (aarch64_bcaxqv8hi): Ditto.
11218 * config/aarch64/aarch64-simd.md:
11219 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11220 (aarch64_crypto_sha512su0qv2di): Ditto.
11221 (aarch64_crypto_sha512su1qv2di): Ditto.
11222 (aarch64_eor3qv8hi): Ditto.
11223 (aarch64_rax1qv2di): Ditto.
11224 (aarch64_xarqv2di): Ditto.
11225 (aarch64_bcaxqv8hi): Ditto.
11226 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11227 (vsha512h2q_u64): Ditto.
11228 (vsha512su0q_u64): Ditto.
11229 (vsha512su1q_u64): Ditto.
11230 (veor3q_u16): Ditto.
11231 (vrax1q_u64): Ditto.
11232 (vxarq_u64): Ditto.
11233 (vbcaxq_u16): Ditto.
11234 * config/arm/types.md (crypto_sha512): New type attribute.
11235 (crypto_sha3): Ditto.
11236 (doc/invoke.texi): Document new sha3 option.
11238 2018-01-10 Michael Collison <michael.collison@arm.com>
11240 * config/aarch64/aarch64-builtins.c:
11241 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11242 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11243 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11244 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11245 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11246 (AARCH64_ISA_SM4): New ISA flag.
11247 (TARGET_SM4): New feature flag for sm4.
11248 * config/aarch64/aarch64-simd-builtins.def
11249 (aarch64_sm3ss1qv4si): Ditto.
11250 (aarch64_sm3tt1aq4si): Ditto.
11251 (aarch64_sm3tt1bq4si): Ditto.
11252 (aarch64_sm3tt2aq4si): Ditto.
11253 (aarch64_sm3tt2bq4si): Ditto.
11254 (aarch64_sm3partw1qv4si): Ditto.
11255 (aarch64_sm3partw2qv4si): Ditto.
11256 (aarch64_sm4eqv4si): Ditto.
11257 (aarch64_sm4ekeyqv4si): Ditto.
11258 * config/aarch64/aarch64-simd.md:
11259 (aarch64_sm3ss1qv4si): Ditto.
11260 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11261 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11262 (aarch64_sm4eqv4si): Ditto.
11263 (aarch64_sm4ekeyqv4si): Ditto.
11264 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11265 (sm3part_op): Ditto.
11266 (CRYPTO_SM3TT): Ditto.
11267 (CRYPTO_SM3PART): Ditto.
11268 (UNSPEC_SM3SS1): New unspec.
11269 (UNSPEC_SM3TT1A): Ditto.
11270 (UNSPEC_SM3TT1B): Ditto.
11271 (UNSPEC_SM3TT2A): Ditto.
11272 (UNSPEC_SM3TT2B): Ditto.
11273 (UNSPEC_SM3PARTW1): Ditto.
11274 (UNSPEC_SM3PARTW2): Ditto.
11275 (UNSPEC_SM4E): Ditto.
11276 (UNSPEC_SM4EKEY): Ditto.
11277 * config/aarch64/constraints.md (Ui2): New constraint.
11278 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11279 * config/arm/types.md (crypto_sm3): New type attribute.
11280 (crypto_sm4): Ditto.
11281 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11282 (vsm3tt1aq_u32): Ditto.
11283 (vsm3tt1bq_u32): Ditto.
11284 (vsm3tt2aq_u32): Ditto.
11285 (vsm3tt2bq_u32): Ditto.
11286 (vsm3partw1q_u32): Ditto.
11287 (vsm3partw2q_u32): Ditto.
11288 (vsm4eq_u32): Ditto.
11289 (vsm4ekeyq_u32): Ditto.
11290 (doc/invoke.texi): Document new sm4 option.
11292 2018-01-10 Michael Collison <michael.collison@arm.com>
11294 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11295 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11296 (AARCH64_FL_FOR_ARCH8_4): New.
11297 (AARCH64_FL_V8_4): New flag.
11298 (doc/invoke.texi): Document new armv8.4-a option.
11300 2018-01-10 Michael Collison <michael.collison@arm.com>
11302 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11303 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11304 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11305 * config/aarch64/aarch64-option-extension.def: Add
11306 AARCH64_OPT_EXTENSION of 'sha2'.
11307 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11308 (crypto): Disable sha2 and aes if crypto disabled.
11309 (crypto): Enable aes and sha2 if enabled.
11310 (simd): Disable sha2 and aes if simd disabled.
11311 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11313 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11314 (TARGET_SHA2): New feature flag for sha2.
11315 (TARGET_AES): New feature flag for aes.
11316 * config/aarch64/aarch64-simd.md:
11317 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11318 conditional on TARGET_AES.
11319 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11320 (aarch64_crypto_sha1hsi): Make pattern conditional
11322 (aarch64_crypto_sha1hv4si): Ditto.
11323 (aarch64_be_crypto_sha1hv4si): Ditto.
11324 (aarch64_crypto_sha1su1v4si): Ditto.
11325 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11326 (aarch64_crypto_sha1su0v4si): Ditto.
11327 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11328 (aarch64_crypto_sha256su0v4si): Ditto.
11329 (aarch64_crypto_sha256su1v4si): Ditto.
11330 (doc/invoke.texi): Document new aes and sha2 options.
11332 2018-01-10 Martin Sebor <msebor@redhat.com>
11334 PR tree-optimization/83781
11335 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11338 2018-01-11 Martin Sebor <msebor@gmail.com>
11339 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11341 PR tree-optimization/83501
11342 PR tree-optimization/81703
11344 * tree-ssa-strlen.c (get_string_cst): Rename...
11345 (get_string_len): ...to this. Handle global constants.
11346 (handle_char_store): Adjust.
11348 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11349 Jim Wilson <jimw@sifive.com>
11351 * config/riscv/riscv-protos.h (riscv_output_return): New.
11352 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11353 (riscv_attribute_table, riscv_output_return),
11354 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11355 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11356 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11358 (riscv_expand_prologue): Add early return for naked function.
11359 (riscv_expand_epilogue): Likewise.
11360 (riscv_function_ok_for_sibcall): Return false for naked function.
11361 (riscv_set_current_function): New.
11362 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11363 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11364 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11365 * doc/extend.texi (RISC-V Function Attributes): New.
11367 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11369 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11370 check for 128-bit long double before checking TCmode.
11371 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11372 128-bit long doubles before checking TFmode or TCmode.
11373 (FLOAT128_IBM_P): Likewise.
11375 2018-01-10 Martin Sebor <msebor@redhat.com>
11377 PR tree-optimization/83671
11378 * builtins.c (c_strlen): Unconditionally return zero for the empty
11380 Use -Warray-bounds for warnings.
11381 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11382 for non-constant array indices with COMPONENT_REF, arrays of
11383 arrays, and pointers to arrays.
11384 (gimple_fold_builtin_strlen): Determine and set length range for
11385 non-constant character arrays.
11387 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11389 PR middle-end/81897
11390 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11393 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11395 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11397 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11400 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11401 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11402 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11403 indexed_or_indirect_operand predicate.
11404 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11405 (*vsx_le_perm_load_v8hi): Likewise.
11406 (*vsx_le_perm_load_v16qi): Likewise.
11407 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11408 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11409 (*vsx_le_perm_store_v8hi): Likewise.
11410 (*vsx_le_perm_store_v16qi): Likewise.
11411 (eight unnamed splitters): Likewise.
11413 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11415 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11416 * config/rs6000/emmintrin.h: Likewise.
11417 * config/rs6000/mmintrin.h: Likewise.
11418 * config/rs6000/xmmintrin.h: Likewise.
11420 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11423 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11425 * tree.c (tree_nop_conversion): Return true for location wrapper
11427 (maybe_wrap_with_location): New function.
11428 (selftest::check_strip_nops): New function.
11429 (selftest::test_location_wrappers): New function.
11430 (selftest::tree_c_tests): Call it.
11431 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11432 (maybe_wrap_with_location): New decl.
11433 (EXPR_LOCATION_WRAPPER_P): New macro.
11434 (location_wrapper_p): New inline function.
11435 (tree_strip_any_location_wrapper): New inline function.
11437 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11440 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11441 stack_realign_offset for the largest alignment of stack slot
11443 (ix86_find_max_used_stack_alignment): New function.
11444 (ix86_finalize_stack_frame_flags): Use it. Set
11445 max_used_stack_alignment if we don't realign stack.
11446 * config/i386/i386.h (machine_function): Add
11447 max_used_stack_alignment.
11449 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11451 * config/arm/arm.opt (-mbranch-cost): New option.
11452 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11455 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11458 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11459 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11461 2018-01-10 Richard Biener <rguenther@suse.de>
11464 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11465 early out so it also covers the case where we have a non-NULL
11468 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11470 PR tree-optimization/83753
11471 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11472 for non-strided grouped accesses if the number of elements is 1.
11474 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11477 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11478 * i386.h (TARGET_USE_GATHER): Define.
11479 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11481 2018-01-10 Martin Liska <mliska@suse.cz>
11484 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11485 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11487 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11488 CLEANUP_NO_PARTITIONING is not set.
11490 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11492 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11493 for vectors, as a partial revert of r254296.
11494 * rtl.h (const_vec_p): Delete.
11495 (const_vec_duplicate_p): Don't test for vector CONSTs.
11496 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11497 * expmed.c (make_tree): Likewise.
11500 * common.md (E, F): Use CONSTANT_P instead of checking for
11502 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11503 checking for CONST_VECTOR.
11505 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11507 PR middle-end/83575
11508 * predict.c (force_edge_cold): Handle in more sane way edges
11509 with no prediction.
11511 2018-01-09 Carl Love <cel@us.ibm.com>
11513 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11515 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11516 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11517 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11518 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11519 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11520 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11521 * config/rs6000/rs6000-protos.h: Add extern defition for
11522 rs6000_generate_float2_double_code.
11523 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11525 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11526 (float2_v2df): Add define_expand.
11528 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11531 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11532 op_mode in the force_to_mode call.
11534 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11536 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11537 instead of checking each element individually.
11538 (aarch64_evpc_uzp): Likewise.
11539 (aarch64_evpc_zip): Likewise.
11540 (aarch64_evpc_ext): Likewise.
11541 (aarch64_evpc_rev): Likewise.
11542 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11543 instead of checking each element individually. Return true without
11545 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11546 whether all selected elements come from the same input, instead of
11547 checking each element individually. Remove calls to gen_rtx_REG,
11548 start_sequence and end_sequence and instead assert that no rtl is
11551 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11553 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11554 order of HIGH and CONST checks.
11556 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11558 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11559 if the destination isn't an SSA_NAME.
11561 2018-01-09 Richard Biener <rguenther@suse.de>
11563 PR tree-optimization/83668
11564 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11566 (canonicalize_loop_form): ... here, renamed from ...
11567 (canonicalize_loop_closed_ssa_form): ... this and amended to
11568 swap successor edges for loop exit blocks to make us use
11569 the RPO order we need for initial schedule generation.
11571 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11573 PR tree-optimization/64811
11574 * match.pd: When optimizing comparisons with Inf, avoid
11575 introducing or losing exceptions from comparisons with NaN.
11577 2018-01-09 Martin Liska <mliska@suse.cz>
11580 * asan.c (shadow_mem_size): Add gcc_assert.
11582 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11584 Don't save registers in main().
11587 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11588 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11589 * config/avr/avr.c (avr_set_current_function): Don't error if
11590 naked, OS_task or OS_main are specified at the same time.
11591 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11593 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11595 * common/config/avr/avr-common.c (avr_option_optimization_table):
11596 Switch on -mmain-is-OS_task for optimizing compilations.
11598 2018-01-09 Richard Biener <rguenther@suse.de>
11600 PR tree-optimization/83572
11601 * graphite.c: Include cfganal.h.
11602 (graphite_transform_loops): Connect infinite loops to exit
11603 and remove fake edges at the end.
11605 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11607 * ipa-inline.c (edge_badness): Revert accidental checkin.
11609 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11612 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11613 symbols; not inline clones.
11615 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11618 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11619 hard registers. Formatting fixes.
11621 PR preprocessor/83722
11622 * gcc.c (try_generate_repro): Pass
11623 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11624 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11627 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11628 Kito Cheng <kito.cheng@gmail.com>
11630 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11631 (riscv_leaf_function_p): Delete.
11632 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11634 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11636 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11638 (do_ifelse): New function.
11639 (do_isel): New function.
11640 (do_sub3): New function.
11641 (do_add3): New function.
11642 (do_load_mask_compare): New function.
11643 (do_overlap_load_compare): New function.
11644 (expand_compare_loop): New function.
11645 (expand_block_compare): Call expand_compare_loop() when appropriate.
11646 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11647 option description.
11648 (-mblock-compare-inline-loop-limit): New option.
11650 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11653 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11654 Reverse order of second and third operands in first alternative.
11655 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11656 of first and second elements in UNSPEC_VPERMR vector.
11657 (altivec_expand_vec_perm_le): Likewise.
11659 2018-01-08 Jeff Law <law@redhat.com>
11661 PR rtl-optimizatin/81308
11662 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11663 (process_switch): If group_case_labels makes a change, then set
11665 (pass_convert_switch::execute): If a switch is converted, then
11666 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11668 PR rtl-optimization/81308
11669 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11672 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11674 PR target/83663 - Revert r255946
11675 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11676 generation for cases where splatting a value is not useful.
11677 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11678 across a vec_duplicate and a paradoxical subreg forming a vector
11679 mode to a vec_concat.
11681 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11683 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11684 -march=armv8.3-a variants.
11685 * config/arm/t-multilib: Likewise.
11686 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11688 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11690 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11692 (cceq_ior_compare_complement): Give it a name so I can use it, and
11693 change boolean_or_operator predicate to boolean_operator so it can
11694 be used to generate a crand.
11695 (eqne): New code iterator.
11696 (bd/bd_neg): New code_attrs.
11697 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11698 a single define_insn.
11699 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11700 decrement (bdnzt/bdnzf/bdzt/bdzf).
11701 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11702 with the new names of the branch decrement patterns, and added the
11703 names of the branch decrement conditional patterns.
11705 2018-01-08 Richard Biener <rguenther@suse.de>
11707 PR tree-optimization/83563
11708 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11711 2018-01-08 Richard Biener <rguenther@suse.de>
11713 PR middle-end/83713
11714 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11716 2018-01-08 Richard Biener <rguenther@suse.de>
11718 PR tree-optimization/83685
11719 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11720 references to abnormals.
11722 2018-01-08 Richard Biener <rguenther@suse.de>
11725 * dwarf2out.c (output_indirect_strings): Handle empty
11726 skeleton_debug_str_hash.
11727 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11729 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11731 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11732 (emit_store_direct): Likewise.
11733 (arc_trampoline_adjust_address): Likewise.
11734 (arc_asm_trampoline_template): New function.
11735 (arc_initialize_trampoline): Use asm_trampoline_template.
11736 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11737 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11738 * config/arc/arc.md (flush_icache): Delete pattern.
11740 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11742 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11743 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11746 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11749 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11750 by not USED_FOR_TARGET.
11751 (make_pass_resolve_sw_modes): Likewise.
11753 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11755 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11758 2018-01-08 Richard Biener <rguenther@suse.de>
11760 PR middle-end/83580
11761 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11763 2018-01-08 Richard Biener <rguenther@suse.de>
11765 PR middle-end/83517
11766 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11768 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11770 PR middle-end/81897
11771 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11772 basic blocks with a small number of successors.
11773 (convert_control_dep_chain_into_preds): Improve handling of
11775 (dump_predicates): Split apart into...
11776 (dump_pred_chain): ...here...
11777 (dump_pred_info): ...and here.
11778 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11779 (can_chain_union_be_invalidated_p): Improve check for invalidation
11781 (uninit_uses_cannot_happen): Avoid unnecessary if
11782 convert_control_dep_chain_into_preds yielded nothing.
11784 2018-01-06 Martin Sebor <msebor@redhat.com>
11786 PR tree-optimization/83640
11787 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11788 subtracting negative offset from size.
11789 (builtin_access::overlap): Adjust offset bounds of the access to fall
11790 within the size of the object if possible.
11792 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11794 PR rtl-optimization/83699
11795 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11796 extract_bit_field_as_subreg to cases in which the extracted
11797 value is also a vector.
11799 * lra-constraints.c (process_alt_operands): Test for the equivalence
11800 substitutions when detecting a possible reload cycle.
11802 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11805 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11806 by default if flag_selective_schedling{,2}. Formatting fixes.
11808 PR rtl-optimization/83682
11809 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11810 if it has non-VECTOR_MODE element mode.
11811 (vec_duplicate_p): Likewise.
11813 PR middle-end/83694
11814 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11815 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11817 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11820 * config/i386/i386-builtin.def
11821 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11822 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11823 Require also OPTION_MASK_ISA_AVX512F in addition to
11824 OPTION_MASK_ISA_GFNI.
11825 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11826 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11827 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11828 to OPTION_MASK_ISA_GFNI.
11829 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11830 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11831 OPTION_MASK_ISA_AVX512BW.
11832 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11833 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11834 addition to OPTION_MASK_ISA_GFNI.
11835 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11836 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11837 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11838 to OPTION_MASK_ISA_GFNI.
11839 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11840 a requirement for all ISAs rather than any of them with a few
11842 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11844 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11845 bitmasks to be enabled with 3 exceptions, instead of requiring any
11846 enabled ISA with lots of exceptions.
11847 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11848 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11849 Change avx512bw in isa attribute to avx512f.
11850 * config/i386/sgxintrin.h: Add license boilerplate.
11851 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11852 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11853 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11854 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11856 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11857 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11858 temporarily sse2 rather than sse if not enabled already.
11861 * config/i386/sse.md (VI248_VLBW): Rename to ...
11862 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11863 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11864 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11865 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11866 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11867 mode iterator instead of VI248_VLBW.
11869 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11871 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11872 (record_modified): Skip clobbers; add debug output.
11873 (param_change_prob): Use sreal frequencies.
11875 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11877 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11878 punt for user-aligned variables.
11880 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11882 * tree-chrec.c (chrec_contains_symbols): Return true for
11885 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11888 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11889 of (x|y) == x for BICS pattern.
11891 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11893 PR tree-optimization/83605
11894 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11895 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11898 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11900 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11901 * config/epiphany/rtems.h: New file.
11903 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11904 Uros Bizjak <ubizjak@gmail.com>
11907 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11908 QIreg_operand instead of register_operand predicate.
11909 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11910 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11911 comments instead of -fmitigate[-_]rop.
11913 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11916 * cgraphunit.c (symbol_table::compile): Switch to text_section
11917 before calling assembly_start debug hook.
11918 * run-rtl-passes.c (run_rtl_passes): Likewise.
11921 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11923 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11924 range_int_cst_p rather than !symbolic_range_p before calling
11925 extract_range_from_multiplicative_op_1.
11927 2018-01-04 Jeff Law <law@redhat.com>
11929 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11930 redundant test in assertion.
11932 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11934 * doc/rtl.texi: Document machine_mode wrapper classes.
11936 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11938 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11939 using tree_to_uhwi.
11941 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11943 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11944 the VEC_PERM_EXPR fold to fail.
11946 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11949 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11950 to switched_sections.
11952 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11955 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11956 test for d.testing.
11958 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11961 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11962 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11964 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11967 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11968 is BLKmode and bitpos not zero or mode change is needed.
11970 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11973 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11976 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11979 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11980 instead of MULT rtx. Update all corresponding splitters.
11981 (*saddl_se): Ditto.
11982 (*ssub<modesuffix>): Ditto.
11983 (*ssubl_se): Ditto.
11984 (*cmp_sadd_di): Update split patterns.
11985 (*cmp_sadd_si): Ditto.
11986 (*cmp_sadd_sidi): Ditto.
11987 (*cmp_ssub_di): Ditto.
11988 (*cmp_ssub_si): Ditto.
11989 (*cmp_ssub_sidi): Ditto.
11990 * config/alpha/predicates.md (const23_operand): New predicate.
11991 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11992 Look for ASHIFT, not MULT inner operand.
11993 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11995 2018-01-04 Martin Liska <mliska@suse.cz>
11997 PR gcov-profile/83669
11998 * gcov.c (output_intermediate_file): Add version to intermediate
12000 * doc/gcov.texi: Document new field 'version' in intermediate
12001 file format. Fix location of '-k' option of gcov command.
12003 2018-01-04 Martin Liska <mliska@suse.cz>
12006 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12008 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12010 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12012 2018-01-03 Martin Sebor <msebor@redhat.com>
12014 PR tree-optimization/83655
12015 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12016 checking calls with invalid arguments.
12018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12020 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12021 (vectorizable_mask_load_store): Delete.
12022 (vectorizable_call): Return false for masked loads and stores.
12023 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12024 instead of gimple_assign_rhs1.
12025 (vectorizable_load): Handle IFN_MASK_LOAD.
12026 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12030 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12032 (vectorizable_mask_load_store): ...here.
12033 (vectorizable_load): ...and here.
12035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12037 * tree-vect-stmts.c (vect_build_all_ones_mask)
12038 (vect_build_zero_merge_argument): New functions, split out from...
12039 (vectorizable_load): ...here.
12041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12043 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12045 (vectorizable_mask_load_store): ...here.
12046 (vectorizable_store): ...and here.
12048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12050 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12052 (vectorizable_mask_load_store): ...here.
12054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12056 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12057 (vect_model_store_cost): Take a vec_load_store_type instead of a
12059 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12060 (vect_model_store_cost): Take a vec_load_store_type instead of a
12062 (vectorizable_mask_load_store): Update accordingly.
12063 (vectorizable_store): Likewise.
12064 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12068 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12069 IFN_MASK_LOAD calls here rather than...
12070 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12073 Alan Hayward <alan.hayward@arm.com>
12074 David Sherwood <david.sherwood@arm.com>
12076 * expmed.c (extract_bit_field_1): For vector extracts,
12077 fall back to extract_bit_field_as_subreg if vec_extract
12080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12081 Alan Hayward <alan.hayward@arm.com>
12082 David Sherwood <david.sherwood@arm.com>
12084 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12085 they are variable or constant sized.
12086 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12087 slots for constant-sized data.
12089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12090 Alan Hayward <alan.hayward@arm.com>
12091 David Sherwood <david.sherwood@arm.com>
12093 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12094 handling COND_EXPRs with boolean comparisons, try to find a better
12095 basis for the mask type than the boolean itself.
12097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12099 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12100 is calculated and how it can be overridden.
12101 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12102 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12104 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12108 Alan Hayward <alan.hayward@arm.com>
12109 David Sherwood <david.sherwood@arm.com>
12111 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12112 Remove the mode argument.
12113 (aarch64_simd_valid_immediate): Remove the mode and inverse
12115 * config/aarch64/iterators.md (bitsize): New iterator.
12116 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12117 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12118 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12119 aarch64_simd_valid_immediate.
12120 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12121 (aarch64_reg_or_bic_imm): Likewise.
12122 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12123 with an insn_type enum and msl with a modifier_type enum.
12124 Replace element_width with a scalar_mode. Change the shift
12125 to unsigned int. Add constructors for scalar_float_mode and
12126 scalar_int_mode elements.
12127 (aarch64_vect_float_const_representable_p): Delete.
12128 (aarch64_can_const_movi_rtx_p)
12129 (aarch64_simd_scalar_immediate_valid_for_move)
12130 (aarch64_simd_make_constant): Update call to
12131 aarch64_simd_valid_immediate.
12132 (aarch64_advsimd_valid_immediate_hs): New function.
12133 (aarch64_advsimd_valid_immediate): Likewise.
12134 (aarch64_simd_valid_immediate): Remove mode and inverse
12135 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12136 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12137 and aarch64_float_const_representable_p on the result.
12138 (aarch64_output_simd_mov_immediate): Remove mode argument.
12139 Update call to aarch64_simd_valid_immediate and use of
12140 simd_immediate_info.
12141 (aarch64_output_scalar_simd_mov_immediate): Update call
12144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12145 Alan Hayward <alan.hayward@arm.com>
12146 David Sherwood <david.sherwood@arm.com>
12148 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12149 (mode_nunits): Likewise CONST_MODE_NUNITS.
12150 * machmode.def (ADJUST_NUNITS): Document.
12151 * genmodes.c (mode_data::need_nunits_adj): New field.
12152 (blank_mode): Update accordingly.
12153 (adj_nunits): New variable.
12154 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12156 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12157 listed in adj_nunits.
12158 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12159 listed in adj_nunits. Don't emit case statements for such modes.
12160 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12161 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12162 nothing if adj_nunits is nonnull.
12163 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12164 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12165 (emit_mode_fbit): Update use of print_maybe_const_decl.
12166 (emit_move_size): Likewise. Treat the array as non-const
12168 (emit_mode_adjustments): Handle adj_nunits.
12170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12172 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12173 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12174 (VECTOR_MODES): Use it.
12175 (make_vector_modes): Take the prefix as an argument.
12177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12178 Alan Hayward <alan.hayward@arm.com>
12179 David Sherwood <david.sherwood@arm.com>
12181 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12182 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12183 for MODE_VECTOR_BOOL.
12184 * machmode.def (VECTOR_BOOL_MODE): Document.
12185 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12186 (make_vector_bool_mode): New function.
12187 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12189 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12190 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12192 * stor-layout.c (int_mode_for_mode): Likewise.
12193 * tree.c (build_vector_type_for_mode): Likewise.
12194 * varasm.c (output_constant_pool_2): Likewise.
12195 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12196 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12197 for MODE_VECTOR_BOOL.
12198 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12199 of mode class checks.
12200 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12201 instead of a list of mode class checks.
12202 (expand_vector_scalar_condition): Likewise.
12203 (type_for_widest_vector_mode): Handle BImode as an inner mode.
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 * machmode.h (mode_size): Change from unsigned short to
12211 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12212 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12213 or if measurement_type is not polynomial.
12214 (fixed_size_mode::includes_p): Check for constant-sized modes.
12215 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12216 return a poly_uint16 rather than an unsigned short.
12217 (emit_mode_size): Change the type of mode_size from unsigned short
12218 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12219 (emit_mode_adjustments): Cope with polynomial vector sizes.
12220 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12222 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12224 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12225 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12226 * caller-save.c (setup_save_areas): Likewise.
12227 (replace_reg_with_saved_mem): Likewise.
12228 * calls.c (emit_library_call_value_1): Likewise.
12229 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12230 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12231 (gen_lowpart_for_combine): Likewise.
12232 * convert.c (convert_to_integer_1): Likewise.
12233 * cse.c (equiv_constant, cse_insn): Likewise.
12234 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12235 (cselib_subst_to_values): Likewise.
12236 * dce.c (word_dce_process_block): Likewise.
12237 * df-problems.c (df_word_lr_mark_ref): Likewise.
12238 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12239 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12240 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12241 (rtl_for_decl_location): Likewise.
12242 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12243 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12244 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12245 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12246 (expand_expr_real_1): Likewise.
12247 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12248 (pad_below): Likewise.
12249 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12250 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12251 * ira.c (get_subreg_tracking_sizes): Likewise.
12252 * ira-build.c (ira_create_allocno_objects): Likewise.
12253 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12254 (ira_sort_regnos_for_alter_reg): Likewise.
12255 * ira-costs.c (record_operand_costs): Likewise.
12256 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12257 (resolve_simple_move): Likewise.
12258 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12259 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12260 (lra_constraints): Likewise.
12261 (CONST_POOL_OK_P): Reject variable-sized modes.
12262 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12263 (add_pseudo_to_slot, lra_spill): Likewise.
12264 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12265 * optabs-query.c (get_best_extraction_insn): Likewise.
12266 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12267 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12268 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12269 * recog.c (offsettable_address_addr_space_p): Likewise.
12270 * regcprop.c (maybe_mode_change): Likewise.
12271 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12272 * regrename.c (build_def_use): Likewise.
12273 * regstat.c (dump_reg_info): Likewise.
12274 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12275 (find_reloads, find_reloads_subreg_address): Likewise.
12276 * reload1.c (eliminate_regs_1): Likewise.
12277 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12278 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12279 (simplify_binary_operation_1, simplify_subreg): Likewise.
12280 * targhooks.c (default_function_arg_padding): Likewise.
12281 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12282 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12283 (verify_gimple_assign_ternary): Likewise.
12284 * tree-inline.c (estimate_move_cost): Likewise.
12285 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12286 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12287 (get_address_cost_ainc): Likewise.
12288 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12289 (vect_supportable_dr_alignment): Likewise.
12290 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12291 (vectorizable_reduction): Likewise.
12292 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12293 (vectorizable_operation, vectorizable_load): Likewise.
12294 * tree.c (build_same_sized_truth_vector_type): Likewise.
12295 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12296 * var-tracking.c (emit_note_insn_var_location): Likewise.
12297 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12298 (ADDR_VEC_ALIGN): Likewise.
12300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12301 Alan Hayward <alan.hayward@arm.com>
12302 David Sherwood <david.sherwood@arm.com>
12304 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12306 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12307 or if measurement_type is polynomial.
12308 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12309 * combine.c (make_extraction): Likewise.
12310 * dse.c (find_shift_sequence): Likewise.
12311 * dwarf2out.c (mem_loc_descriptor): Likewise.
12312 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12313 (extract_bit_field, extract_low_bits): Likewise.
12314 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12315 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12316 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12317 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12318 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12319 * reload.c (find_reloads): Likewise.
12320 * reload1.c (alter_reg): Likewise.
12321 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12322 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12323 * tree-if-conv.c (predicate_mem_writes): Likewise.
12324 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12325 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12326 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12327 * valtrack.c (dead_debug_insert_temp): Likewise.
12328 * varasm.c (mergeable_constant_section): Likewise.
12329 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12332 Alan Hayward <alan.hayward@arm.com>
12333 David Sherwood <david.sherwood@arm.com>
12335 * expr.c (expand_assignment): Cope with polynomial mode sizes
12336 when assigning to a CONCAT.
12338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12339 Alan Hayward <alan.hayward@arm.com>
12340 David Sherwood <david.sherwood@arm.com>
12342 * machmode.h (mode_precision): Change from unsigned short to
12344 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12346 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12347 or if measurement_type is not polynomial.
12348 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12349 in which the mode is already known to be a scalar_int_mode.
12350 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12351 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12353 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12354 for GET_MODE_PRECISION.
12355 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12356 for GET_MODE_PRECISION.
12357 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12359 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12360 (expand_field_assignment, make_extraction): Likewise.
12361 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12362 (get_last_value): Likewise.
12363 * convert.c (convert_to_integer_1): Likewise.
12364 * cse.c (cse_insn): Likewise.
12365 * expr.c (expand_expr_real_1): Likewise.
12366 * lra-constraints.c (simplify_operand_subreg): Likewise.
12367 * optabs-query.c (can_atomic_load_p): Likewise.
12368 * optabs.c (expand_atomic_load): Likewise.
12369 (expand_atomic_store): Likewise.
12370 * ree.c (combine_reaching_defs): Likewise.
12371 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12372 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12373 * tree.h (type_has_mode_precision_p): Likewise.
12374 * ubsan.c (instrument_si_overflow): Likewise.
12376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12377 Alan Hayward <alan.hayward@arm.com>
12378 David Sherwood <david.sherwood@arm.com>
12380 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12381 polynomial numbers of units.
12382 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12383 (valid_vector_subparts_p): New function.
12384 (build_vector_type): Remove temporary shim and take the number
12385 of units as a poly_uint64 rather than an int.
12386 (build_opaque_vector_type): Take the number of units as a
12387 poly_uint64 rather than an int.
12388 * tree.c (build_vector_from_ctor): Handle polynomial
12389 TYPE_VECTOR_SUBPARTS.
12390 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12391 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12392 (build_vector_from_val): If the number of units is variable,
12393 use build_vec_duplicate_cst for constant operands and
12394 VEC_DUPLICATE_EXPR otherwise.
12395 (make_vector_type): Remove temporary is_constant ().
12396 (build_vector_type, build_opaque_vector_type): Take the number of
12397 units as a poly_uint64 rather than an int.
12398 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12400 * cfgexpand.c (expand_debug_expr): Likewise.
12401 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12402 (store_constructor, expand_expr_real_1): Likewise.
12403 (const_scalar_mask_from_tree): Likewise.
12404 * fold-const-call.c (fold_const_reduction): Likewise.
12405 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12406 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12407 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12408 (fold_relational_const): Likewise.
12409 (native_interpret_vector): Likewise. Change the size from an
12410 int to an unsigned int.
12411 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12412 TYPE_VECTOR_SUBPARTS.
12413 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12414 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12415 duplicating a non-constant operand into a variable-length vector.
12416 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12417 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12418 * ipa-icf.c (sem_variable::equals): Likewise.
12419 * match.pd: Likewise.
12420 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12421 * print-tree.c (print_node): Likewise.
12422 * stor-layout.c (layout_type): Likewise.
12423 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12424 * tree-cfg.c (verify_gimple_comparison): Likewise.
12425 (verify_gimple_assign_binary): Likewise.
12426 (verify_gimple_assign_ternary): Likewise.
12427 (verify_gimple_assign_single): Likewise.
12428 * tree-pretty-print.c (dump_generic_node): Likewise.
12429 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12430 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12431 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12432 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12433 (vect_shift_permute_load_chain): Likewise.
12434 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12435 (expand_vector_condition, optimize_vector_constructor): Likewise.
12436 (lower_vec_perm, get_compute_type): Likewise.
12437 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12438 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12439 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12440 (vect_recog_mask_conversion_pattern): Likewise.
12441 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12442 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12443 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12444 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12445 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12446 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12447 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12448 (supportable_widening_operation): Likewise.
12449 (supportable_narrowing_operation): Likewise.
12450 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12452 * varasm.c (output_constant): Likewise.
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 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12459 so that both the length == 3 and length != 3 cases set up their
12460 own permute vectors. Add comments explaining why we know the
12461 number of elements is constant.
12462 (vect_permute_load_chain): Likewise.
12464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12465 Alan Hayward <alan.hayward@arm.com>
12466 David Sherwood <david.sherwood@arm.com>
12468 * machmode.h (mode_nunits): Change from unsigned char to
12470 (ONLY_FIXED_SIZE_MODES): New macro.
12471 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12472 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12473 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12475 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12476 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12477 or if measurement_type is not polynomial.
12478 * genmodes.c (ZERO_COEFFS): New macro.
12479 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12481 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12482 Use ZERO_COEFFS when emitting initializers.
12483 * data-streamer.h (bp_pack_poly_value): New function.
12484 (bp_unpack_poly_value): Likewise.
12485 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12486 for GET_MODE_NUNITS.
12487 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12488 for GET_MODE_NUNITS.
12489 * tree.c (make_vector_type): Remove temporary shim and make
12490 the real function take the number of units as a poly_uint64
12491 rather than an int.
12492 (build_vector_type_for_mode): Handle polynomial nunits.
12493 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12494 * emit-rtl.c (const_vec_series_p_1): Likewise.
12495 (gen_rtx_CONST_VECTOR): Likewise.
12496 * fold-const.c (test_vec_duplicate_folding): Likewise.
12497 * genrecog.c (validate_pattern): Likewise.
12498 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12499 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12500 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12501 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12502 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12503 * rtlanal.c (subreg_get_info): Likewise.
12504 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12505 (vect_grouped_load_supported): Likewise.
12506 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12507 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12508 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12509 (simplify_const_unary_operation, simplify_binary_operation_1)
12510 (simplify_const_binary_operation, simplify_ternary_operation)
12511 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12512 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12513 instead of CONST_VECTOR_NUNITS.
12514 * varasm.c (output_constant_pool_2): Likewise.
12515 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12516 explicit-encoded elements in the XVEC for variable-length vectors.
12518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12520 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12523 Alan Hayward <alan.hayward@arm.com>
12524 David Sherwood <david.sherwood@arm.com>
12526 * coretypes.h (fixed_size_mode): Declare.
12527 (fixed_size_mode_pod): New typedef.
12528 * builtins.h (target_builtins::x_apply_args_mode)
12529 (target_builtins::x_apply_result_mode): Change type to
12530 fixed_size_mode_pod.
12531 * builtins.c (apply_args_size, apply_result_size, result_vector)
12532 (expand_builtin_apply_args_1, expand_builtin_apply)
12533 (expand_builtin_return): Update accordingly.
12535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12537 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12538 * cselib.c (cselib_hash_rtx): Likewise.
12539 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12540 CONST_VECTOR encoding.
12542 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12543 Jeff Law <law@redhat.com>
12546 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12547 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12548 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12549 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12552 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12553 explicitly probe *sp in a noreturn function if there were any callee
12554 register saves or frame pointer is needed.
12556 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12559 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12560 BLKmode for ternary, binary or unary expressions.
12563 * var-tracking.c (delete_vta_debug_insn): New inline function.
12564 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12565 insns from get_insns () to NULL instead of each bb separately.
12566 Use delete_vta_debug_insn. No longer static.
12567 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12568 delete_vta_debug_insns callers.
12569 * rtl.h (delete_vta_debug_insns): Declare.
12570 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12571 instead of variable_tracking_main.
12573 2018-01-03 Martin Sebor <msebor@redhat.com>
12575 PR tree-optimization/83603
12576 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12577 arguments past the endof the argument list in functions declared
12578 without a prototype.
12579 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12580 Avoid checking when arguments are null.
12582 2018-01-03 Martin Sebor <msebor@redhat.com>
12585 * doc/extend.texi (attribute const): Fix a typo.
12586 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12587 issuing -Wsuggest-attribute for void functions.
12589 2018-01-03 Martin Sebor <msebor@redhat.com>
12591 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12592 offset_int::from instead of wide_int::to_shwi.
12593 (maybe_diag_overlap): Remove assertion.
12594 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12595 * gimple-ssa-sprintf.c (format_directive): Same.
12596 (parse_directive): Same.
12597 (sprintf_dom_walker::compute_format_length): Same.
12598 (try_substitute_return_value): Same.
12600 2018-01-03 Jeff Law <law@redhat.com>
12602 PR middle-end/83654
12603 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12604 non-constant residual for zero at runtime and avoid probing in
12605 that case. Reorganize code for trailing problem to mirror handling
12608 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12610 PR tree-optimization/83501
12611 * tree-ssa-strlen.c (get_string_cst): New.
12612 (handle_char_store): Call get_string_cst.
12614 2018-01-03 Martin Liska <mliska@suse.cz>
12616 PR tree-optimization/83593
12617 * tree-ssa-strlen.c: Include tree-cfg.h.
12618 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12619 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12620 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12622 (strlen_dom_walker::before_dom_children): Call
12623 gimple_purge_dead_eh_edges. Dump tranformation with details
12625 (strlen_dom_walker::before_dom_children): Update call by adding
12626 new argument cleanup_eh.
12627 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12629 2018-01-03 Martin Liska <mliska@suse.cz>
12632 * cif-code.def (VARIADIC_THUNK): New enum value.
12633 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12636 2018-01-03 Jan Beulich <jbeulich@suse.com>
12638 * sse.md (mov<mode>_internal): Tighten condition for when to use
12639 vmovdqu<ssescalarsize> for TI and OI modes.
12641 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12643 Update copyright years.
12645 2018-01-03 Martin Liska <mliska@suse.cz>
12648 * ipa-visibility.c (function_and_variable_visibility): Skip
12649 functions with noipa attribure.
12651 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12653 * gcc.c (process_command): Update copyright notice dates.
12654 * gcov-dump.c (print_version): Ditto.
12655 * gcov.c (print_version): Ditto.
12656 * gcov-tool.c (print_version): Ditto.
12657 * gengtype.c (create_file): Ditto.
12658 * doc/cpp.texi: Bump @copying's copyright year.
12659 * doc/cppinternals.texi: Ditto.
12660 * doc/gcc.texi: Ditto.
12661 * doc/gccint.texi: Ditto.
12662 * doc/gcov.texi: Ditto.
12663 * doc/install.texi: Ditto.
12664 * doc/invoke.texi: Ditto.
12666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12668 * vector-builder.h (vector_builder::m_full_nelts): Change from
12669 unsigned int to poly_uint64.
12670 (vector_builder::full_nelts): Update prototype accordingly.
12671 (vector_builder::new_vector): Likewise.
12672 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12673 (vector_builder::operator ==): Likewise.
12674 (vector_builder::finalize): Likewise.
12675 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12676 Take the number of elements as a poly_uint64 rather than an
12678 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12679 from unsigned int to poly_uint64.
12680 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12681 (vec_perm_indices::new_vector): Likewise.
12682 (vec_perm_indices::length): Likewise.
12683 (vec_perm_indices::nelts_per_input): Likewise.
12684 (vec_perm_indices::input_nelts): Likewise.
12685 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12686 number of elements per input as a poly_uint64 rather than an
12687 unsigned int. Use the original encoding for variable-length
12688 vectors, rather than clamping each individual element.
12689 For the second and subsequent elements in each pattern,
12690 clamp the step and base before clamping their sum.
12691 (vec_perm_indices::series_p): Handle polynomial element counts.
12692 (vec_perm_indices::all_in_range_p): Likewise.
12693 (vec_perm_indices_to_tree): Likewise.
12694 (vec_perm_indices_to_rtx): Likewise.
12695 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12696 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12697 (tree_vector_builder::new_binary_operation): Handle polynomial
12698 element counts. Return false if we need to know the number
12699 of elements at compile time.
12700 * fold-const.c (fold_vec_perm): Punt if the number of elements
12701 isn't known at compile time.
12703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12705 * vec-perm-indices.h (vec_perm_builder): Change element type
12706 from HOST_WIDE_INT to poly_int64.
12707 (vec_perm_indices::element_type): Update accordingly.
12708 (vec_perm_indices::clamp): Handle polynomial element_types.
12709 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12710 (vec_perm_indices::all_in_range_p): Likewise.
12711 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12713 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12714 polynomial vec_perm_indices element types.
12715 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12716 * fold-const.c (fold_vec_perm): Likewise.
12717 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12718 * tree-vect-generic.c (lower_vec_perm): Likewise.
12719 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12720 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12721 element type to HOST_WIDE_INT.
12723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12724 Alan Hayward <alan.hayward@arm.com>
12725 David Sherwood <david.sherwood@arm.com>
12727 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12728 rather than an int. Use plus_constant.
12729 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12730 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12733 Alan Hayward <alan.hayward@arm.com>
12734 David Sherwood <david.sherwood@arm.com>
12736 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12737 a HOST_WIDE_INT to a poly_int64.
12739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12740 Alan Hayward <alan.hayward@arm.com>
12741 David Sherwood <david.sherwood@arm.com>
12743 * calls.c (load_register_parameters): Cope with polynomial
12744 mode sizes. Require a constant size for BLKmode parameters
12745 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12746 forces a parameter to be padded at the lsb end in order to
12747 fill a complete number of words, require the parameter size
12748 to be ordered wrt UNITS_PER_WORD.
12750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12751 Alan Hayward <alan.hayward@arm.com>
12752 David Sherwood <david.sherwood@arm.com>
12754 * reload1.c (spill_stack_slot_width): Change element type
12755 from unsigned int to poly_uint64_pod.
12756 (alter_reg): Treat mode sizes as polynomial.
12758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12759 Alan Hayward <alan.hayward@arm.com>
12760 David Sherwood <david.sherwood@arm.com>
12762 * reload.c (complex_word_subreg_p): New function.
12763 (reload_inner_reg_of_subreg, push_reload): Use it.
12765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12766 Alan Hayward <alan.hayward@arm.com>
12767 David Sherwood <david.sherwood@arm.com>
12769 * lra-constraints.c (process_alt_operands): Reject matched
12770 operands whose sizes aren't ordered.
12771 (match_reload): Refer to this check here.
12773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12774 Alan Hayward <alan.hayward@arm.com>
12775 David Sherwood <david.sherwood@arm.com>
12777 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12778 that the mode size is in the set {1, 2, 4, 8, 16}.
12780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12781 Alan Hayward <alan.hayward@arm.com>
12782 David Sherwood <david.sherwood@arm.com>
12784 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12785 Use plus_constant instead of gen_rtx_PLUS.
12787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12788 Alan Hayward <alan.hayward@arm.com>
12789 David Sherwood <david.sherwood@arm.com>
12791 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12792 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12793 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12794 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12795 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12796 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12797 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12798 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12799 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12800 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12802 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12803 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12804 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12805 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12806 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12807 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12808 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12809 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12810 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12811 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12813 * expr.c (emit_move_resolve_push): Treat the input and result
12814 of PUSH_ROUNDING as a poly_int64.
12815 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12816 (emit_push_insn): Likewise.
12817 * lra-eliminations.c (mark_not_eliminable): Likewise.
12818 * recog.c (push_operand): Likewise.
12819 * reload1.c (elimination_effects): Likewise.
12820 * rtlanal.c (nonzero_bits1): Likewise.
12821 * calls.c (store_one_arg): Likewise. Require the padding to be
12822 known at compile time.
12824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12825 Alan Hayward <alan.hayward@arm.com>
12826 David Sherwood <david.sherwood@arm.com>
12828 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12829 Use plus_constant instead of gen_rtx_PLUS.
12831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12832 Alan Hayward <alan.hayward@arm.com>
12833 David Sherwood <david.sherwood@arm.com>
12835 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12836 rather than an int.
12838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12839 Alan Hayward <alan.hayward@arm.com>
12840 David Sherwood <david.sherwood@arm.com>
12842 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12843 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12844 via stack temporaries. Treat the mode size as polynomial too.
12846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12847 Alan Hayward <alan.hayward@arm.com>
12848 David Sherwood <david.sherwood@arm.com>
12850 * expr.c (expand_expr_real_2): When handling conversions involving
12851 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12852 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12853 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12862 Alan Hayward <alan.hayward@arm.com>
12863 David Sherwood <david.sherwood@arm.com>
12865 * combine.c (can_change_dest_mode): Handle polynomial
12866 REGMODE_NATURAL_SIZE.
12867 * expmed.c (store_bit_field_1): Likewise.
12868 * expr.c (store_constructor): Likewise.
12869 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12870 and polynomial REGMODE_NATURAL_SIZE.
12871 (gen_lowpart_common): Likewise.
12872 * reginfo.c (record_subregs_of_mode): Likewise.
12873 * rtlanal.c (read_modify_subreg_p): Likewise.
12875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12876 Alan Hayward <alan.hayward@arm.com>
12877 David Sherwood <david.sherwood@arm.com>
12879 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12880 numbers of elements.
12882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12883 Alan Hayward <alan.hayward@arm.com>
12884 David Sherwood <david.sherwood@arm.com>
12886 * match.pd: Cope with polynomial numbers of vector elements.
12888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12889 Alan Hayward <alan.hayward@arm.com>
12890 David Sherwood <david.sherwood@arm.com>
12892 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12893 in a POINTER_PLUS_EXPR.
12895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12896 Alan Hayward <alan.hayward@arm.com>
12897 David Sherwood <david.sherwood@arm.com>
12899 * omp-simd-clone.c (simd_clone_subparts): New function.
12900 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12901 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12908 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12909 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12910 (expand_vector_condition, vector_element): Likewise.
12911 (subparts_gt): New function.
12912 (get_compute_type): Use subparts_gt.
12913 (count_type_subparts): Delete.
12914 (expand_vector_operations_1): Use subparts_gt instead of
12915 count_type_subparts.
12917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12918 Alan Hayward <alan.hayward@arm.com>
12919 David Sherwood <david.sherwood@arm.com>
12921 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12922 (vect_compile_time_alias): ...this new function. Do the calculation
12923 on poly_ints rather than trees.
12924 (vect_prune_runtime_alias_test_list): Update call accordingly.
12926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12927 Alan Hayward <alan.hayward@arm.com>
12928 David Sherwood <david.sherwood@arm.com>
12930 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12932 (vect_schedule_slp_instance): Likewise.
12934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12935 Alan Hayward <alan.hayward@arm.com>
12936 David Sherwood <david.sherwood@arm.com>
12938 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12939 constant and extern definitions for variable-length vectors.
12940 (vect_get_constant_vectors): Note that the number of units
12941 is known to be constant.
12943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12944 Alan Hayward <alan.hayward@arm.com>
12945 David Sherwood <david.sherwood@arm.com>
12947 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12948 of units as polynomial. Choose between WIDE and NARROW based
12951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12952 Alan Hayward <alan.hayward@arm.com>
12953 David Sherwood <david.sherwood@arm.com>
12955 * tree-vect-stmts.c (simd_clone_subparts): New function.
12956 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12959 Alan Hayward <alan.hayward@arm.com>
12960 David Sherwood <david.sherwood@arm.com>
12962 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12963 vectors as polynomial. Use build_index_vector for
12964 IFN_GOMP_SIMD_LANE.
12966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12967 Alan Hayward <alan.hayward@arm.com>
12968 David Sherwood <david.sherwood@arm.com>
12970 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12971 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12972 for variable-length vectors.
12973 (vectorizable_mask_load_store): Treat the number of units as
12974 polynomial, asserting that it is constant if the condition has
12975 already been enforced.
12976 (vectorizable_store, vectorizable_load): Likewise.
12978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12979 Alan Hayward <alan.hayward@arm.com>
12980 David Sherwood <david.sherwood@arm.com>
12982 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12983 of units as polynomial. Punt if we can't tell at compile time
12984 which vector contains the final result.
12986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12987 Alan Hayward <alan.hayward@arm.com>
12988 David Sherwood <david.sherwood@arm.com>
12990 * tree-vect-loop.c (vectorizable_induction): Treat the number
12991 of units as polynomial. Punt on SLP inductions. Use an integer
12992 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12993 cast of such a series for variable-length floating-point
12996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12997 Alan Hayward <alan.hayward@arm.com>
12998 David Sherwood <david.sherwood@arm.com>
13000 * tree.h (build_index_vector): Declare.
13001 * tree.c (build_index_vector): New function.
13002 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13003 of units as polynomial, forcibly converting it to a constant if
13004 vectorizable_reduction has already enforced the condition.
13005 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13006 to create a {1,2,3,...} vector.
13007 (vectorizable_reduction): Treat the number of units as polynomial.
13008 Choose vectype_in based on the largest scalar element size rather
13009 than the smallest number of units. Enforce the restrictions
13012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13013 Alan Hayward <alan.hayward@arm.com>
13014 David Sherwood <david.sherwood@arm.com>
13016 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13017 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13024 * target.def (autovectorize_vector_sizes): Return the vector sizes
13025 by pointer, using vector_sizes rather than a bitmask.
13026 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13027 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13028 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13030 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13031 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13032 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13033 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13034 * omp-general.c (omp_max_vf): Likewise.
13035 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13036 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13037 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13038 * tree-vect-slp.c (vect_slp_bb): Likewise.
13039 * doc/tm.texi: Regenerate.
13040 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13042 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13043 the vector size as a poly_uint64 rather than an unsigned int.
13044 (current_vector_size): Change from an unsigned int to a poly_uint64.
13045 (get_vectype_for_scalar_type): Update accordingly.
13046 * tree.h (build_truth_vector_type): Take the size and number of
13047 units as a poly_uint64 rather than an unsigned int.
13048 (build_vector_type): Add a temporary overload that takes
13049 the number of units as a poly_uint64 rather than an unsigned int.
13050 * tree.c (make_vector_type): Likewise.
13051 (build_truth_vector_type): Take the number of units as a poly_uint64
13052 rather than an unsigned int.
13054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13055 Alan Hayward <alan.hayward@arm.com>
13056 David Sherwood <david.sherwood@arm.com>
13058 * target.def (get_mask_mode): Take the number of units and length
13059 as poly_uint64s rather than unsigned ints.
13060 * targhooks.h (default_get_mask_mode): Update accordingly.
13061 * targhooks.c (default_get_mask_mode): Likewise.
13062 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13063 * doc/tm.texi: Regenerate.
13065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13066 Alan Hayward <alan.hayward@arm.com>
13067 David Sherwood <david.sherwood@arm.com>
13069 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13070 * omp-general.c (omp_max_vf): Likewise.
13071 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13072 (expand_omp_simd): Handle polynomial safelen.
13073 * omp-low.c (omplow_simd_context): Add a default constructor.
13074 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13075 (lower_rec_simd_input_clauses): Update accordingly.
13076 (lower_rec_input_clauses): Likewise.
13078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13079 Alan Hayward <alan.hayward@arm.com>
13080 David Sherwood <david.sherwood@arm.com>
13082 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13083 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13084 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13085 (vect_analyze_slp_cost): Likewise.
13086 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13087 (vect_model_load_cost): Likewise.
13089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13090 Alan Hayward <alan.hayward@arm.com>
13091 David Sherwood <david.sherwood@arm.com>
13093 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13094 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13095 from an unsigned int * to a poly_uint64_pod *.
13096 (calculate_unrolling_factor): New function.
13097 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13100 Alan Hayward <alan.hayward@arm.com>
13101 David Sherwood <david.sherwood@arm.com>
13103 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13104 from an unsigned int to a poly_uint64.
13105 (_loop_vec_info::slp_unrolling_factor): Likewise.
13106 (_loop_vec_info::vectorization_factor): Change from an int
13108 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13109 (vect_get_num_vectors): New function.
13110 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13111 (vect_get_num_copies): Use vect_get_num_vectors.
13112 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13113 to an unsigned int *.
13114 (vect_analyze_data_refs): Change min_vf from an int * to a
13116 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13117 than an unsigned HOST_WIDE_INT.
13118 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13119 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13120 to an unsigned int *.
13121 (vect_analyze_data_ref_dependences): Likewise.
13122 (vect_compute_data_ref_alignment): Handle polynomial vf.
13123 (vect_enhance_data_refs_alignment): Likewise.
13124 (vect_prune_runtime_alias_test_list): Likewise.
13125 (vect_shift_permute_load_chain): Likewise.
13126 (vect_supportable_dr_alignment): Likewise.
13127 (dependence_distance_ge_vf): Take the vectorization factor as a
13128 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13129 (vect_analyze_data_refs): Change min_vf from an int * to a
13131 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13132 vfm1 as a poly_uint64 rather than an int. Make the same change
13133 for the returned bound_scalar.
13134 (vect_gen_vector_loop_niters): Handle polynomial vf.
13135 (vect_do_peeling): Likewise. Update call to
13136 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13137 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13139 * tree-vect-loop.c (vect_determine_vectorization_factor)
13140 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13141 (vect_get_known_peeling_cost): Likewise.
13142 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13143 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13144 (vect_transform_loop): Likewise. Use the lowest possible VF when
13145 updating the upper bounds of the loop.
13146 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13147 rather than an int.
13148 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13149 polynomial unroll factors.
13150 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13151 (vect_make_slp_decision): Likewise.
13152 (vect_supported_load_permutation_p): Likewise, and polynomial
13154 (vect_analyze_slp_cost): Handle polynomial vf.
13155 (vect_slp_analyze_node_operations): Likewise.
13156 (vect_slp_analyze_bb_1): Likewise.
13157 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13158 than an unsigned HOST_WIDE_INT.
13159 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13160 (vectorizable_load): Handle polynomial vf.
13161 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13163 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13166 Alan Hayward <alan.hayward@arm.com>
13167 David Sherwood <david.sherwood@arm.com>
13169 * match.pd: Handle bit operations involving three constants
13170 and try to fold one pair.
13172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13174 * tree-vect-loop-manip.c: Include gimple-fold.h.
13175 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13176 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13177 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13178 Add a path that uses a step of VF instead of 1, but disable it
13180 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13181 and niters_no_overflow parameters. Update calls to
13182 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13183 Create a new SSA name if the latter choses to use a ste other
13184 than zero, and return it via niters_vector_mult_vf_var.
13185 * tree-vect-loop.c (vect_transform_loop): Update calls to
13186 vect_do_peeling, vect_gen_vector_loop_niters and
13187 slpeel_make_loop_iterate_ntimes.
13188 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13189 (vect_gen_vector_loop_niters): Update declarations after above changes.
13191 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13193 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13194 128-bit round to integer instructions.
13195 (ceil<mode>2): Likewise.
13196 (btrunc<mode>2): Likewise.
13197 (round<mode>2): Likewise.
13199 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13201 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13202 unaligned VSX load/store on P8/P9.
13203 (expand_block_clear): Allow the use of unaligned VSX
13204 load/store on P8/P9.
13206 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13208 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13210 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13211 swap associated with both a load and a store.
13213 2018-01-02 Andrew Waterman <andrew@sifive.com>
13215 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13216 * config/riscv/riscv.md (clear_cache): Use it.
13218 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13220 * web.c: Remove out-of-date comment.
13222 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13224 * expr.c (fixup_args_size_notes): Check that any existing
13225 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13226 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13227 (emit_single_push_insn): ...here.
13229 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13231 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13232 (const_vector_encoded_nelts): New function.
13233 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13234 (const_vector_int_elt, const_vector_elt): Declare.
13235 * emit-rtl.c (const_vector_int_elt_1): New function.
13236 (const_vector_elt): Likewise.
13237 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13238 of CONST_VECTOR_ELT.
13240 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13242 * expr.c: Include rtx-vector-builder.h.
13243 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13244 directly on the tree encoding.
13245 (const_vector_from_tree): Likewise.
13246 * optabs.c: Include rtx-vector-builder.h.
13247 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13248 sequence of "u" values.
13249 * vec-perm-indices.c: Include rtx-vector-builder.h.
13250 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13251 directly on the vec_perm_indices encoding.
13253 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13255 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13256 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13257 * rtx-vector-builder.h: New file.
13258 * rtx-vector-builder.c: Likewise.
13259 * rtl.h (rtx_def::u2): Add a const_vector field.
13260 (CONST_VECTOR_NPATTERNS): New macro.
13261 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13262 (CONST_VECTOR_DUPLICATE_P): Likewise.
13263 (CONST_VECTOR_STEPPED_P): Likewise.
13264 (CONST_VECTOR_ENCODED_ELT): Likewise.
13265 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13266 (unwrap_const_vec_duplicate): Likewise.
13267 (const_vec_series_p): Check for a non-duplicated vector encoding.
13268 Say that the function only returns true for integer vectors.
13269 * emit-rtl.c: Include rtx-vector-builder.h.
13270 (gen_const_vec_duplicate_1): Delete.
13271 (gen_const_vector): Call gen_const_vec_duplicate instead of
13272 gen_const_vec_duplicate_1.
13273 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13274 (gen_const_vec_duplicate): Use rtx_vector_builder.
13275 (gen_const_vec_series): Likewise.
13276 (gen_rtx_CONST_VECTOR): Likewise.
13277 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13278 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13279 Build a new vector rather than modifying a CONST_VECTOR in-place.
13280 (handle_special_swappables): Update call accordingly.
13281 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13282 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13283 Build a new vector rather than modifying a CONST_VECTOR in-place.
13284 (handle_special_swappables): Update call accordingly.
13286 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13288 * simplify-rtx.c (simplify_const_binary_operation): Use
13289 CONST_VECTOR_ELT instead of XVECEXP.
13291 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13293 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13294 the selector elements to be different from the data elements
13295 if the selector is a VECTOR_CST.
13296 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13297 ssizetype for the selector.
13299 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13301 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13302 before testing each element individually.
13303 * tree-vect-generic.c (lower_vec_perm): Likewise.
13305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13307 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13308 * selftest-run-tests.c (selftest::run_tests): Call it.
13309 * vector-builder.h (vector_builder::operator ==): New function.
13310 (vector_builder::operator !=): Likewise.
13311 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13312 (vec_perm_indices::all_from_input_p): New function.
13313 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13314 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13315 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13316 instead of reading the VECTOR_CST directly. Detect whether both
13317 vector inputs are the same before constructing the vec_perm_indices,
13318 and update the number of inputs argument accordingly. Use the
13319 utility functions added above. Only construct sel2 if we need to.
13321 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13323 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13324 the broadcast of the low byte.
13325 (expand_mult_highpart): Use an explicit encoding for the permutes.
13326 * optabs-query.c (can_mult_highpart_p): Likewise.
13327 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13328 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13329 (vectorizable_bswap): Likewise.
13330 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13331 explicit encoding for the power-of-2 permutes.
13332 (vect_permute_store_chain): Likewise.
13333 (vect_grouped_load_supported): Likewise.
13334 (vect_permute_load_chain): Likewise.
13336 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13338 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13339 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13340 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13341 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13342 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13343 (vect_gen_perm_mask_any): Likewise.
13345 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13347 * int-vector-builder.h: New file.
13348 * vec-perm-indices.h: Include int-vector-builder.h.
13349 (vec_perm_indices): Redefine as an int_vector_builder.
13350 (auto_vec_perm_indices): Delete.
13351 (vec_perm_builder): Redefine as a stand-alone class.
13352 (vec_perm_indices::vec_perm_indices): New function.
13353 (vec_perm_indices::clamp): Likewise.
13354 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13355 (vec_perm_indices::new_vector): New function.
13356 (vec_perm_indices::new_expanded_vector): Update for new
13357 vec_perm_indices class.
13358 (vec_perm_indices::rotate_inputs): New function.
13359 (vec_perm_indices::all_in_range_p): Operate directly on the
13360 encoded form, without computing elided elements.
13361 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13362 encoding. Update for new vec_perm_indices class.
13363 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13364 the given vec_perm_builder.
13365 (expand_vec_perm_var): Update vec_perm_builder constructor.
13366 (expand_mult_highpart): Use vec_perm_builder instead of
13367 auto_vec_perm_indices.
13368 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13369 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13370 or double series encoding as appropriate.
13371 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13372 vec_perm_indices instead of auto_vec_perm_indices.
13373 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13374 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13375 (vect_permute_store_chain): Likewise.
13376 (vect_grouped_load_supported): Likewise.
13377 (vect_permute_load_chain): Likewise.
13378 (vect_shift_permute_load_chain): Likewise.
13379 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13380 (vect_transform_slp_perm_load): Likewise.
13381 (vect_schedule_slp_instance): Likewise.
13382 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13383 (vectorizable_mask_load_store): Likewise.
13384 (vectorizable_bswap): Likewise.
13385 (vectorizable_store): Likewise.
13386 (vectorizable_load): Likewise.
13387 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13388 vec_perm_indices instead of auto_vec_perm_indices. Use
13389 tree_to_vec_perm_builder to read the vector from a tree.
13390 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13391 vec_perm_builder instead of a vec_perm_indices.
13392 (have_whole_vector_shift): Use vec_perm_builder and
13393 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13394 truncation to calc_vec_perm_mask_for_shift.
13395 (vect_create_epilog_for_reduction): Likewise.
13396 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13397 from auto_vec_perm_indices to vec_perm_indices.
13398 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13399 instead of changing individual elements.
13400 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13401 the vector in d.perm.
13402 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13403 from auto_vec_perm_indices to vec_perm_indices.
13404 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13405 instead of changing individual elements.
13406 (arm_vectorize_vec_perm_const): Use new_vector to install
13407 the vector in d.perm.
13408 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13409 Update vec_perm_builder constructor.
13410 (rs6000_expand_interleave): Likewise.
13411 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13412 (rs6000_expand_interleave): Likewise.
13414 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13416 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13417 to qimode could truncate the indices.
13418 * optabs.c (expand_vec_perm_var): Likewise.
13420 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13422 * Makefile.in (OBJS): Add vec-perm-indices.o.
13423 * vec-perm-indices.h: New file.
13424 * vec-perm-indices.c: Likewise.
13425 * target.h (vec_perm_indices): Replace with a forward class
13427 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13428 * optabs.h: Include vec-perm-indices.h.
13429 (expand_vec_perm): Delete.
13430 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13431 (expand_vec_perm_const): Declare.
13432 * target.def (vec_perm_const_ok): Replace with...
13433 (vec_perm_const): ...this new hook.
13434 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13435 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13436 * doc/tm.texi: Regenerate.
13437 * optabs.def (vec_perm_const): Delete.
13438 * doc/md.texi (vec_perm_const): Likewise.
13439 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13440 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13441 expand_vec_perm for constant permutation vectors. Assert that
13442 the mode of variable permutation vectors is the integer equivalent
13443 of the mode that is being permuted.
13444 * optabs-query.h (selector_fits_mode_p): Declare.
13445 * optabs-query.c: Include vec-perm-indices.h.
13446 (selector_fits_mode_p): New function.
13447 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13448 is defined, instead of checking whether the vec_perm_const_optab
13449 exists. Use targetm.vectorize.vec_perm_const instead of
13450 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13451 fit in the vector mode before using a variable permute.
13452 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13453 vec_perm_indices instead of an rtx.
13454 (expand_vec_perm): Replace with...
13455 (expand_vec_perm_const): ...this new function. Take the selector
13456 as a vec_perm_indices rather than an rtx. Also take the mode of
13457 the selector. Update call to shift_amt_for_vec_perm_mask.
13458 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13459 Use vec_perm_indices::new_expanded_vector to expand the original
13460 selector into bytes. Check whether the indices fit in the vector
13461 mode before using a variable permute.
13462 (expand_vec_perm_var): Make global.
13463 (expand_mult_highpart): Use expand_vec_perm_const.
13464 * fold-const.c: Includes vec-perm-indices.h.
13465 * tree-ssa-forwprop.c: Likewise.
13466 * tree-vect-data-refs.c: Likewise.
13467 * tree-vect-generic.c: Likewise.
13468 * tree-vect-loop.c: Likewise.
13469 * tree-vect-slp.c: Likewise.
13470 * tree-vect-stmts.c: Likewise.
13471 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13473 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13474 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13475 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13476 (aarch64_vectorize_vec_perm_const): ...this new function.
13477 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13478 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13479 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13480 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13481 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13482 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13483 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13485 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13486 check for NEON modes.
13487 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13488 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13489 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13490 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13492 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13493 the old VEC_PERM_CONST conditions.
13494 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13495 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13496 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13497 (ia64_vectorize_vec_perm_const_ok): Merge into...
13498 (ia64_vectorize_vec_perm_const): ...this new function.
13499 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13500 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13501 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13502 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13503 * config/mips/mips.c (mips_expand_vec_perm_const)
13504 (mips_vectorize_vec_perm_const_ok): Merge into...
13505 (mips_vectorize_vec_perm_const): ...this new function.
13506 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13507 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13508 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13509 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13510 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13511 (rs6000_expand_vec_perm_const): Delete.
13512 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13514 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13515 (altivec_expand_vec_perm_const_le): Take each operand individually.
13516 Operate on constant selectors rather than rtxes.
13517 (altivec_expand_vec_perm_const): Likewise. Update call to
13518 altivec_expand_vec_perm_const_le.
13519 (rs6000_expand_vec_perm_const): Delete.
13520 (rs6000_vectorize_vec_perm_const_ok): Delete.
13521 (rs6000_vectorize_vec_perm_const): New function.
13522 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13523 an element count and rtx array.
13524 (rs6000_expand_extract_even): Update call accordingly.
13525 (rs6000_expand_interleave): Likewise.
13526 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13527 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13528 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13529 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13530 (rs6000_expand_vec_perm_const): Delete.
13531 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13532 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13533 (altivec_expand_vec_perm_const_le): Take each operand individually.
13534 Operate on constant selectors rather than rtxes.
13535 (altivec_expand_vec_perm_const): Likewise. Update call to
13536 altivec_expand_vec_perm_const_le.
13537 (rs6000_expand_vec_perm_const): Delete.
13538 (rs6000_vectorize_vec_perm_const_ok): Delete.
13539 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13540 reference to the SPE evmerge intructions.
13541 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13542 an element count and rtx array.
13543 (rs6000_expand_extract_even): Update call accordingly.
13544 (rs6000_expand_interleave): Likewise.
13545 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13546 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13548 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13550 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13552 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13553 vector mode and that that mode matches the mode of the data
13555 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13556 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13557 directly using expand_vec_perm_1 when forcing selectors into
13559 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13561 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13563 * optabs-query.h (can_vec_perm_p): Delete.
13564 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13565 * optabs-query.c (can_vec_perm_p): Split into...
13566 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13567 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13568 particular selector is valid.
13569 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13570 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13571 (vect_grouped_load_supported): Likewise.
13572 (vect_shift_permute_load_chain): Likewise.
13573 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13574 (vect_transform_slp_perm_load): Likewise.
13575 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13576 (vectorizable_bswap): Likewise.
13577 (vect_gen_perm_mask_checked): Likewise.
13578 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13579 implementations of variable permutation vectors into account
13580 when deciding which selector to use.
13581 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13582 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13583 with a false third argument.
13584 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13585 to test whether the constant selector is valid and can_vec_perm_var_p
13586 to test whether a variable selector is valid.
13588 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13590 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13591 * optabs-query.c (can_vec_perm_p): Likewise.
13592 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13593 instead of vec_perm_indices.
13594 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13595 (vect_gen_perm_mask_checked): Likewise,
13596 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13597 (vect_gen_perm_mask_checked): Likewise,
13599 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13601 * optabs-query.h (qimode_for_vec_perm): Declare.
13602 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13603 (qimode_for_vec_perm): ...this new function.
13604 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13606 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13608 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13609 does not have a conditional at the top.
13611 2018-01-02 Richard Biener <rguenther@suse.de>
13613 * ipa-inline.c (big_speedup_p): Fix expression.
13615 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13618 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13621 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13625 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13626 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13627 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13628 cond_taken_branch_cost 3->4.
13630 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13632 PR tree-optimization/83581
13633 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13634 TODO_cleanup_cfg if any changes have been made.
13636 PR middle-end/83608
13637 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13638 convert_modes if target mode has the right side, but different mode
13641 PR middle-end/83609
13642 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13643 last argument when extracting from CONCAT. If either from_real or
13644 from_imag is NULL, use expansion through memory. If result is not
13645 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13646 the parts directly to inner mode, if even that fails, use expansion
13649 PR middle-end/83623
13650 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13651 check for bswap in mode rather than HImode and use that in expand_unop
13654 Copyright (C) 2018 Free Software Foundation, Inc.
13656 Copying and distribution of this file, with or without modification,
13657 are permitted in any medium without royalty provided the copyright
13658 notice and this notice are preserved.