1 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5 * config/nds32/nds32.opt (minline-asm-r15): New option.
7 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
11 * config/nds32/nds32.md (abssi2): New pattern.
13 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
15 * config/i386/i386.md (rex64namesuffix): New mode attribute.
16 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
17 Merge insn pattern from sse_cvtsi2ss<round_name> and
18 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
19 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
20 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
21 using SWI48 mode iterator.
22 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
23 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
24 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
25 pattern from sse_cvttss2si<round_saeonly_name>
26 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
27 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
28 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
29 using SWI48 mode iterator.
30 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
31 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
32 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
33 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
34 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
35 using SWI48 mode iterator.
36 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
37 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
38 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
39 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
40 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
42 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
43 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
44 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
45 pattern from sse_cvttsd2si<round_saeonly_name>
46 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
48 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
50 * config/nds32/nds32-md-auxiliary.c
51 (nds32_valid_smw_lwm_base_p): Refine.
52 (nds32_output_smw_single_word): Refine.
53 (nds32_output_smw_double_word): New.
54 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
56 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
58 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
59 (nds32_output_stack_pop): Refine.
60 (nds32_expand_unaligned_load): Refine.
61 (nds32_expand_unaligned_store): Refine.
63 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
64 Chung-Ju Wu <jasonwucj@gmail.com>
66 * config/nds32/constants.md: Add TP_REGNUM constant.
67 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
68 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
70 * config/nds32/nds32-doubleword.md: Consider flag_pic.
71 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
72 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
73 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
74 and PIC code generation.
75 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
77 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
79 * config/nds32/nds32.md: Support TLS and PIC.
80 * config/nds32/nds32.c: Support TLS and PIC.
81 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
82 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
85 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
87 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
90 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
91 Chung-Ju Wu <jasonwucj@gmail.com>
93 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
94 * config/nds32/nds32-md-auxiliary.c
95 (symbolic_reference_mentioned_p): New.
96 (nds32_legitimize_ict_address): New.
97 (nds32_expand_ict_move): New.
98 (nds32_indirect_call_referenced_p): New.
99 (nds32_symbol_binds_local_p): Delete.
100 (nds32_long_call_p): Modify.
101 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
102 * config/nds32/nds32-protos.h
103 (symbolic_reference_mentioned_p): Declare.
104 (nds32_legitimize_ict_address): Declare.
105 (nds32_expand_ict_move): Declare.
106 (nds32_indirect_call_referenced_p): Declare.
107 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
108 (nds32_relax_group): Use nds32_ict_const_p as condition.
109 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
110 (nds32_asm_file_start): Output ict_model directive in asm code.
111 (nds32_legitimate_address_p): Consider indirect call.
112 (nds32_print_operand): Consider indirect call.
113 (nds32_print_operand_address): Consider indirect call.
114 (nds32_insert_attributes): Handle "indirect_call" attribute.
115 (TARGET_LEGITIMATE_ADDRESS_P): Define.
116 (TARGET_LEGITIMATE_CONSTANT_P): Define.
117 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
118 (TARGET_DELEGITIMIZE_ADDRESS): Define.
119 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
120 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
121 (TARGET_ICT_MODEL_SMALL): Define.
122 (TARGET_ICT_MODEL_LARGE): Define.
123 * config/nds32/nds32.md (movsi): Consider ict model.
124 (call, call_value): Consider ict model.
125 (sibcall, sibcall_value): Consider ict model.
126 * config/nds32/nds32.opt (mict-model): New option.
127 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
130 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
131 Monk Chiang <sh.chiang04@gmail.com>
132 Jim Wilson <jimw@sifive.com>
134 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
135 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
136 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
137 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
138 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
139 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
140 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
141 compute save_libcall_adjustment properly.
142 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
143 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
144 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
145 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
146 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
147 (ABI_SPEC): Handle mabi=ilp32e.
148 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
150 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
151 <-march>: Add rv32e as an example.
153 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
156 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
157 (intra_create_variable_infos): Handle C++ constructors.
159 2018-05-18 Martin Liska <mliska@suse.cz>
161 * passes.def: Remove a redundant pass.
163 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
166 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
168 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
170 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
172 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
174 (armv2, armv2a, armv3, armv3m): Delete architectures.
175 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
176 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
177 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
179 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
180 (*mulsidi3adddi): Likewise.
181 (mulsidi3): Likewise.
182 (*mulsidi3_nov6): Likewise.
183 (umulsidi3): Likewise.
184 (umulsidi3_nov6): Likewise.
185 (umaddsidi4): Likewise.
186 (*umulsidi3adddi): Likewise.
187 (smulsi3_highpart): Likewise.
188 (*smulsi3_highpart_nov6): Likewise.
189 (umulsi3_highpart): Likewise.
190 (*umulsi3_highpart_nov6): Likewise.
191 * config/arm/arm.h (arm_arch3m): Delete.
192 * config/arm/arm.c (arm_arch3m): Delete.
193 (arm_option_override_internal): Update armv3-related comment.
194 (arm_configure_build_target): Delete use of isa_bit_mode32.
195 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
196 (arm_rtx_costs_internal): Delete check of arm_arch3m.
197 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
200 * config/arm/arm-protos.h (arm_arch3m): Delete.
201 * config/arm/arm-tables.opt: Regenerate.
202 * config/arm/arm-tune.md: Likewise.
203 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
204 deleted architectures.
206 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
208 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
209 (armv5t, armv5te): New features.
210 (ARMv5, ARMv5e): Delete fgroups.
211 (ARMv5t, ARMv5te): Adjust for above changes.
213 (armv5, armv5e): Delete arches.
214 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
216 (*call_reg_arm): Likewise.
217 (*call_value_reg_armv5): Likewise.
218 (*call_value_reg_arm): Likewise.
219 (*call_symbol): Likewise.
220 (*call_value_symbol): Likewise.
221 (*sibcall_insn): Likewise.
222 (*sibcall_value_insn): Likewise.
224 (prefetch): Likewise.
225 (define_split and define_peephole2 dependent on arm_arch5):
227 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
229 (TARGET_ARM_QBIT): Likewise.
230 (TARGET_DSP_MULTIPLY): Likewise.
231 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
232 (arm_arch5, arm_arch5e): Delete.
233 (arm_arch5t, arm_arch5te): Declare.
234 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
235 (arm_arch5t): Declare.
236 (arm_option_reconfigure_globals): Update for the above.
237 (arm_options_perform_arch_sanity_checks): Update comment, replace
238 use of arm_arch5 with arm_arch5t.
239 (use_return_insn): Likewise.
240 (arm_emit_call_insn): Likewise.
241 (output_return_instruction): Likewise.
242 (arm_final_prescan_insn): Likewise.
243 (arm_coproc_builtin_available): Likewise.
244 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
245 arm_arch5e with arm_arch5t and arm_arch5te.
246 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
247 (arm_arch5t, arm_arch5te): Declare.
248 * config/arm/arm-tables.opt: Regenerate.
249 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
250 * config/arm/t-multilib: Likewise.
251 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
252 instead of arm_arch5.
253 (*call_reg_thumb1): Likewise.
254 (*call_value_reg_thumb1_v5): Likewise.
255 (*call_value_reg_thumb1): Likewise.
256 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
258 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
260 2018-05-18 Martin Liska <mliska@suse.cz>
262 PR gcov-profile/84846
263 * doc/gcov.texi: Document -t option of gcov tool.
265 2018-05-18 Martin Liska <mliska@suse.cz>
267 PR gcov-profile/84846
268 * gcov.c (print_usage): Add new -t option.
269 (process_args): Handle the option.
270 (generate_results): Use stdout as output when requested by
273 2018-05-18 Martin Liska <mliska@suse.cz>
275 PR gcov-profile/84846
276 * coverage.c (coverage_init): Write PWD to .gcno file.
277 * doc/gcov.texi: Document how working directory is printed.
278 * gcov-dump.c (dump_gcov_file): Print PWD.
279 * gcov.c (output_intermediate_file): Likewise.
280 (read_graph_file): Read PWD string.
281 (output_lines): Print PWD.
283 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
286 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
287 for retval and return false if all args to phi are zero.
289 2018-05-18 Richard Biener <rguenther@suse.de>
291 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
293 (evrp_dom_walker::before_dom_children): Call it.
295 2018-05-18 Richard Biener <rguenther@suse.de>
297 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
298 results when processing array refs with variable index.
300 2018-05-18 Toon Moene <toon@moene.org>
302 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
303 directly after that of -floop-interchange. Indicate that both
304 options are enabled by default when specifying -O3.
306 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
308 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
309 iterator. Delete separate integer-mode vec_set<mode> expander.
310 (aarch64_simd_vec_setv2di): Delete.
311 (vec_setv2di): Delete.
312 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
313 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
314 the "w, r" alternative.
316 2018-05-18 Martin Liska <mliska@suse.cz>
318 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
319 * tree-pass.h (make_pass_lower_switch_O0): New function.
320 * tree-switch-conversion.c (node_has_low_bound): Remove.
321 (node_has_high_bound): Likewise.
322 (node_is_bounded): Likewise.
323 (class pass_lower_switch): Make it a template type and create
325 (pass_lower_switch::execute): Add template argument.
326 (make_pass_lower_switch): New function.
327 (make_pass_lower_switch_O0): New function.
328 (do_jump_if_equal): Remove.
329 (emit_case_nodes): Simplify to just handle all 3 cases and leave
330 all the hard work to tree optimization passes.
332 2018-05-18 Martin Liska <mliska@suse.cz>
334 * dbgcnt.c (limit_low): Renamed from limit.
335 (limit_high): New variable.
336 (dbg_cnt_is_enabled): Check for upper limit.
337 (dbg_cnt): Adjust dumping.
338 (dbg_cnt_set_limit_by_index): Add new argument for high
340 (dbg_cnt_set_limit_by_name): Likewise.
341 (dbg_cnt_process_single_pair): Parse new format.
342 (dbg_cnt_process_opt): Use strtok.
343 (dbg_cnt_list_all_counters): Remove 'value' and add
345 * doc/invoke.texi: Document changes.
347 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
349 * doc/sourcebuild.texi (scalar_all_fma): Document.
350 * tree.def (FMA_EXPR): Delete.
351 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
352 * internal-fn.c (ternary_direct): New macro.
353 (expand_ternary_optab_fn): Likewise.
354 (direct_ternary_optab_supported_p): Likewise.
355 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
356 * builtins.c (fold_builtin_fma): Delete.
357 (fold_builtin_3): Don't call it.
358 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
359 * expr.c (expand_expr_real_2): Likewise.
360 * fold-const.c (operand_equal_p): Likewise.
361 (fold_ternary_loc): Likewise.
362 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
363 * gimple.c (DEFTREECODE): Likewise.
364 * gimplify.c (gimplify_expr): Likewise.
365 * optabs-tree.c (optab_for_tree_code): Likewise.
366 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
367 * tree-eh.c (operation_could_trap_p): Likewise.
368 (stmt_could_throw_1_p): Likewise.
369 * tree-inline.c (estimate_operator_cost): Likewise.
370 * tree-pretty-print.c (dump_generic_node): Likewise.
371 (op_code_prio): Likewise.
372 * tree-ssa-loop-im.c (stmt_cost): Likewise.
373 * tree-ssa-operands.c (get_expr_operands): Likewise.
374 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
375 * fold-const-call.h (fold_fma): Delete.
376 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
377 CFN_FNMA and CFN_FNMS.
379 * genmatch.c (combined_fn): New enum.
380 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
381 (commutative_op): New function.
382 (commutate): Use it. Handle more than 2 operands.
383 (dt_operand::gen_gimple_expr): Use commutative_op.
384 (parser::parse_expr): Allow :c to be used with non-binary
385 operators if the commutative operand is known.
386 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
387 CFN_FMS, CFN_FNMA and CFN_FNMS.
388 (backprop::process_assign_use): Remove FMA_EXPR handling.
389 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
390 (gen_hsa_fma): New function.
391 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
392 IFN_FNMA and IFN_FNMS.
393 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
394 * gimple-fold.h (follow_all_ssa_edges): Declare.
395 * gimple-fold.c (follow_all_ssa_edges): New function.
396 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
397 gimple_build interface and use follow_all_ssa_edges to fold the result.
398 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
399 instead of checking for optabs directly.
400 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
401 rather than FMA_EXPRs.
402 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
403 call to IFN_FMA instead of an FMA_EXPR.
405 2018-05-17 Jim Wilson <jimw@sifive.com>
407 * expr.c (do_tablejump): When converting index to Pmode, if we have a
408 sign extended promoted subreg, and the range does not have the sign bit
409 set, then do a sign extend.
411 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
412 test, check for sign extended subreg and/or constant operands, and
413 do a sign extend in that case.
415 2018-05-17 Steve Ellcey <sellcey@cavium.com>
417 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
418 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
420 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
421 Change logics_shift_reg to logics_shift_imm.
422 (thunderx2t99_fp_loadpair_basic): Delete.
423 (thunderx2t99_fp_storepair_basic): Delete.
424 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
425 (thunderx2t99_asimd_polynomial): Delete.
426 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
427 and neon_fp_mul_d_scalar_q.
428 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
429 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
430 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
431 (thunderx2t99_asimd_lut): Add missing tbl types.
432 (thunderx2t99_asimd_ext): Delete.
433 (thunderx2t99_asimd_load1_1_mult): Delete.
434 (thunderx2t99_asimd_load1_2_mult): Delete.
435 (thunderx2t99_asimd_load1_ldp): New.
436 (thunderx2t99_asimd_load1): New.
437 (thunderx2t99_asimd_load2): Add missing *load2* types.
438 (thunderx2t99_asimd_load3): New.
439 (thunderx2t99_asimd_load4): New.
440 (thunderx2t99_asimd_store1_1_mult): Delete.
441 (thunderx2t99_asimd_store1_2_mult): Delete.
442 (thunderx2t99_asimd_store2_mult): Delete.
443 (thunderx2t99_asimd_store2_onelane): Delete.
444 (thunderx2t99_asimd_store_stp): New.
445 (thunderx2t99_asimd_store1): New.
446 (thunderx2t99_asimd_store2): New.
447 (thunderx2t99_asimd_store3): New.
448 (thunderx2t99_asimd_store4): New.
450 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
452 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
453 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
455 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
456 Segher Boessenkool <segher@kernel.crashing.org>
459 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
462 2018-05-17 Richard Biener <rguenther@suse.de>
464 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
465 for pruning loop and prune defs feeding only already visited PHIs.
467 2018-05-17 Richard Biener <rguenther@suse.de>
469 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
471 2018-05-17 Bin Cheng <bin.cheng@arm.com>
472 Richard Biener <rguenther@suse.de>
474 PR tree-optimization/85793
475 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
476 for VMAT_ELEMENTWISE.
478 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
480 * internal-fn.h (lookup_internal_fn): Declare
481 * internal-fn.c (lookup_internal_fn): New function.
482 * gimple.c (gimple_build_call_from_tree): Handle calls to
484 * gimple-pretty-print.c (dump_gimple_call): Print "." before
485 internal function names.
486 * tree-pretty-print.c (dump_generic_node): Likewise.
487 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
489 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
491 * gimple-fold.h (gimple_build): Make the function forms take
492 combined_fn rather than built_in_function.
493 (gimple_simplify): Likewise.
494 * gimple-match-head.c (gimple_simplify): Likewise.
495 * gimple-fold.c (gimple_build): Likewise.
496 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
497 rather than gimple_build_call_internal.
498 (get_initial_defs_for_reduction): Likewise.
499 (vect_create_epilog_for_reduction): Likewise.
500 (vectorizable_live_operation): Likewise.
502 2018-05-17 Martin Liska <mliska@suse.cz>
504 * gimple-ssa-sprintf.c (format_directive): Do not use
505 space in between 'G_' and '('.
507 2018-05-17 Jakub Jelinek <jakub@redhat.com>
510 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
511 even if the mask is not all ones.
514 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
516 (ix86_gimple_fold_builtin): Likewise.
519 * config/i386/i386.c: Include tree-vector-builder.h.
520 (ix86_vector_shift_count): New function.
521 (ix86_fold_builtin): Fold shift builtins by scalar count.
522 (ix86_gimple_fold_builtin): Likewise.
524 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
525 _mm512_setzero): New intrinsics.
527 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
528 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
530 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
531 code generation for cases where splatting a value is not useful.
532 * simplify-rtx.c (simplify_ternary_operation): Simplify
533 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
534 mode to a vec_concat.
536 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
538 * config.gcc: Support "goldmont-plus".
539 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
541 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
542 PROCESSOR_GOLDMONT_PLUS.
543 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
544 (processor_target_table): Add "goldmont-plus".
545 (PTA_GOLDMONT_PLUS): Define.
546 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
547 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
548 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
549 (fold_builtin_cpu): Add "goldmont-plus".
550 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
551 (ix86_option_override_internal): Add "goldmont-plus".
552 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
553 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
554 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
555 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
557 2018-05-17 Richard Biener <rguenther@suse.de>
559 PR tree-optimization/85757
560 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
561 remove defs that only feed that PHI from further processing.
563 2018-05-16 Jim Wilson <jimw@sifive.com>
565 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
567 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
569 2018-05-16 Mark Wielaard <mark@klomp.org>
571 * dwarf2out.c (count_index_strings): New function.
572 (output_indirect_strings): Call count_index_strings and generate
573 header for dwarf_version >= 5.
575 2018-05-16 Mark Wielaard <mark@klomp.org>
577 * dwarf2out.c (dwarf_FORM): New function.
578 (set_indirect_string): Use dwarf_FORM.
579 (reset_indirect_string): Likewise.
580 (size_of_die): Likewise.
581 (value_format): Likewise.
582 (output_die): Likewise.
583 (add_skeleton_AT_string): Likewise.
584 (output_macinfo_op): Likewise.
585 (index_string): Likewise.
586 (output_index_string_offset): Likewise.
587 (output_index_string): Likewise.
588 (count_index_strings): Likewise.
590 2018-05-16 Carl Love <cel@us.ibm.com>
592 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
593 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
595 2018-05-16 Martin Jambor <mjambor@suse.cz>
597 * ipa-prop.c (ipa_free_all_edge_args): Remove.
598 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
600 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
602 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
603 (fnma<mode>4): Likewise.
604 (fms<mode>4): Likewise.
605 (fnms<mode>4): Likewise.
606 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
607 (aarch64_fnma<mode>4): Likewise.
608 (aarch64_fms<mode>4): Likewise.
609 (aarch64_fnms<mode>4): Likewise.
610 (aarch64_fnmadd<mode>4): Likewise.
612 2018-05-16 Jason Merrill <jason@redhat.com>
614 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
616 2018-05-16 Richard Biener <rguenther@suse.de>
618 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
619 (dump_stmt_cost): Declare.
620 (add_stmt_cost): Dump cost we add.
621 (add_stmt_costs): New function.
622 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
624 (vect_analyze_stmt): Adjust prototype.
625 (vectorizable_condition): Likewise.
626 (vectorizable_live_operation): Likewise.
627 (vectorizable_reduction): Likewise.
628 (vectorizable_induction): Likewise.
629 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
630 cost vector to pass to vectorizable_ and record afterwards.
631 (vect_model_reduction_cost): Take cost vector argument and adjust.
632 (vect_model_induction_cost): Likewise.
633 (vectorizable_reduction): Likewise.
634 (vectorizable_induction): Likewise.
635 (vectorizable_live_operation): Likewise.
636 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
637 SLP_TREE_NUMBER_OF_VEC_STMTS.
638 (vect_analyze_slp_cost_1): Remove.
639 (vect_analyze_slp_cost): Likewise.
640 (vect_slp_analyze_node_operations): Take visited args and
641 a target cost vector. Avoid processing already visited stmt sets.
642 (vect_slp_analyze_operations): Use a local cost vector to gather
643 costs and register those of non-discarded instances.
644 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
645 (vect_schedule_slp_instance): Remove copying of
646 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
648 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
649 adding cost. Record cost entry location.
650 (vect_prologue_cost_for_slp_op): Function to compute cost of
651 a constant or invariant generated for SLP vect in the prologue,
652 split out from vect_analyze_slp_cost_1.
653 (vect_model_simple_cost): Make static. Adjust for SLP costing.
654 (vect_model_promotion_demotion_cost): Likewise.
655 (vect_model_store_cost): Likewise, make static.
656 (vect_model_load_cost): Likewise.
657 (vectorizable_bswap): Add cost vector arg and adjust.
658 (vectorizable_call): Likewise.
659 (vectorizable_simd_clone_call): Likewise.
660 (vectorizable_conversion): Likewise.
661 (vectorizable_assignment): Likewise.
662 (vectorizable_shift): Likewise.
663 (vectorizable_operation): Likewise.
664 (vectorizable_store): Likewise.
665 (vectorizable_load): Likewise.
666 (vectorizable_condition): Likewise.
667 (vectorizable_comparison): Likewise.
668 (can_vectorize_live_stmts): Likewise.
669 (vect_analyze_stmt): Likewise.
670 (vect_transform_stmt): Adjust calls to vectorizable_*.
671 * tree-vectorizer.c: Include gimple-pretty-print.h.
672 (dump_stmt_cost): New function.
674 2018-05-16 Richard Biener <rguenther@suse.de>
676 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
677 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
678 * tree-ssa-dse.c: Include tree-ssa-loop.h.
679 (check_name): New callback.
680 (dse_classify_store): Track cycles via a visited bitmap of PHI
681 defs and simplify handling of in-loop and across loop dead stores
682 and properly fail for loop-variant refs. Handle byte-tracking with
683 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
686 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
688 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
689 (vect_get_mask_type_for_stmt): Likewise.
690 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
692 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
693 to determine the statement's vector type and the vector type that
694 should be used for calculating nunits. Deal with cases in which
695 the type has to be deferred.
696 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
697 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
698 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
699 (vect_determine_vf_for_stmt): New functions, split out from...
700 (vect_determine_vectorization_factor): ...here.
701 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
702 (vect_get_mask_type_for_stmt): New functions, split out from
703 vect_determine_vectorization_factor.
705 2018-05-16 Richard Biener <rguenther@suse.de>
707 * tree-cfg.c (verify_gimple_assign_ternary): Properly
708 verify the [VEC_]COND_EXPR embedded comparison.
710 2018-05-15 Martin Sebor <msebor@redhat.com>
712 PR tree-optimization/85753
713 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
714 RECORD_TYPE in addition to ARRAY_TYPE.
716 2018-05-15 Martin Sebor <msebor@redhat.com>
719 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
721 2018-05-15 Richard Biener <rguenther@suse.de>
723 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
724 add by_clobber_p one. Change algorithm to collect all defs
725 representing uses we need to walk and try reducing them to
726 a single one before failing.
727 (dse_dom_walker::dse_optimize_stmt): Adjust.
729 2018-05-13 Mark Wielaard <mark@klomp.org>
731 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
732 (size_of_loc_descr): Likewise.
733 (output_loc_operands): Likewise.
734 (output_loc_operands_raw): Likewise.
735 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
736 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
737 (hash_loc_operands): Likewise.
738 (compare_loc_operands): Likewise.
740 2018-05-14 Mark Wielaard <mark@klomp.org>
742 * dwarf2out.c (count_index_addrs): New function.
743 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
745 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
747 PR tree-optimization/83648
748 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
749 return value as malloc candidate.
751 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
754 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
755 as true in call to suggest_attribute.
757 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
759 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
762 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
764 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
766 (aarch64_crypto_aesd_fused): Likewise.
768 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
770 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
771 (movsi_aarch64): Likewise.
772 (load_pairsi): Likewise.
773 (load_pairdi): Likewise.
774 (store_pairsi): Likewise.
775 (store_pairdi): Likewise.
776 (load_pairsf): Likewise.
777 (load_pairdf): Likewise.
778 (store_pairsf): Likewise.
779 (store_pairdf): Likewise.
780 (zero_extend): Likewise.
781 (trunc): Swap alternatives.
782 (fcvt_target): Add '?' to prefer w over r.
784 2018-05-14 Jakub Jelinek <jakub@redhat.com>
787 * config/i386/i386.md: Disallow non-commutative arithmetics in
788 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
789 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
790 in the peephole2 before it.
792 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
794 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
795 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
796 (ix86_handle_option): Handle -mcldemote.
797 * config.gcc: New header.
798 * config/i386/cldemoteintrin.h: New file.
799 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
800 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
802 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
803 OPTION_MASK_ISA_CLDEMOTE.
804 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
805 (ix86_valid_target_attribute_inner_p): Ditto.
806 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
807 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
808 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
809 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
810 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
812 * config/i386/i386.opt: Add -mcldemote.
813 * config/i386/x86intrin.h: New header.
814 * doc/invoke.texi: Add -mcldemote.
816 2018-05-14 Richard Biener <rguenther@suse.de>
818 * doc/match-and-simplify.texi: Adjust :s documentation.
820 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
822 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
823 intended memcpy size.
824 (REORDER_45): Likewise.
826 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
829 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
830 * vec.c (qsort_chk): Use gcc_qsort.
831 * Makefile.in (OBJS-libcommon): Add sort.o.
832 (build/sort.o): New target. Use it...
833 (BUILD_RTL): ... here, and...
834 (build/gencfn-macros): ... here, and...
835 (build/genmatch): ... here.
837 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
838 Chung-Ju Wu <jasonwucj@gmail.com>
840 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
841 * config/nds32/nds32-graywolf.md: New file.
842 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
843 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
845 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
846 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
847 * config/nds32/nds32.md (pipeline_model): Add graywolf.
848 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
849 * config/nds32/pipelines.md: Include n15 settings.
851 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
852 Chung-Ju Wu <jasonwucj@gmail.com>
854 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
855 * config/nds32/nds32-n13.md: New file.
856 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
857 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
859 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
860 * config/nds32/nds32.md (pipeline_model): Add n13.
861 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
862 * config/nds32/pipelines.md: Include n13 settings.
864 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
865 Chung-Ju Wu <jasonwucj@gmail.com>
867 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
868 * config/nds32/nds32-n10.md: New file.
869 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
870 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
872 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
873 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
874 * config/nds32/nds32.md (pipeline_model): Add n10.
875 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
876 * config/nds32/pipelines.md: Include n10 settings.
878 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
879 Kito Cheng <kito.cheng@gmail.com>
880 Chung-Ju Wu <jasonwucj@gmail.com>
882 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
883 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
884 Add enum values for DSP extension instructions.
885 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
887 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
888 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
890 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
891 * config/nds32/nds32-dspext.md: New file for DSP implementation.
892 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
893 * config/nds32/nds32-intrinsic.md: Likewise.
894 * config/nds32/nds32_intrinsic.h: Likewise.
895 * config/nds32/nds32-md-auxiliary.c: Likewise.
896 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
897 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
898 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
899 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
900 * config/nds32/nds32-protos.h: New declarations for DSP extension.
901 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
902 TYPE_DMAC in switch statement.
903 * config/nds32/nds32.c: New checking and implementation for DSP
904 extension instructions.
905 * config/nds32/nds32.h: Likewise.
906 * config/nds32/nds32.md: Likewise.
907 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
908 * config/nds32/predicates.md: Implement new predicates for DSP
911 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
913 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
914 Reformat alternatives and attributes so it is easier to identify
915 which constraints/attributes go with which instruction.
916 (mov<mode>_hardfloat32, FMOVE64): Likewise.
917 (mov<mode>_softfloat32, FMOVE64): Likewise.
918 (mov<mode>_hardfloat64, FMOVE64): Likewise.
919 (mov<mode>_softfloat64, FMOVE64): Likewise.
921 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
923 * doc/extend.texi (PowerPC Built-in Functions): Rename this
925 (Basic PowerPC Built-in Functions): The new name of the
926 subsection previously known as "PowerPC Built-in Functions".
927 (Basic PowerPC Built-in Functions Available on all Configurations):
929 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
930 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
931 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
932 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
934 2018-05-11 Martin Jambor <mjambor@suse.cz>
937 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
940 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
943 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
945 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
947 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
948 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
949 (ix86_handle_option): Handle -mwaitpkg.
950 * config.gcc: New header.
951 * config/i386/cpuid.h (bit_WAITPKG): New bit.
952 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
953 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
955 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
956 OPTION_MASK_ISA_WAITPKG.
957 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
958 (ix86_option_override_internal): Add PTA_WAITPKG.
959 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
960 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
962 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
963 __builtin_ia32_umwait and __builtin_ia32_tpause.
964 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
965 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
966 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
967 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
968 UNSPECV_TPAUSE): New.
969 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
970 * config/i386/i386.opt: Add -mwaitpkg.
971 * config/i386/waitpkgintrin.h: New file.
972 * config/i386/x86intrin.h: New header.
973 * doc/invoke.texi: Add -mwaitpkg.
975 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
978 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
980 (cortex-m0): Use armv6s-m isa.
981 (cortex-m0plus): Likewise.
982 (cortex-m1): Likewise.
983 (cortex-m0.small-multiply): Likewise.
984 (cortex-m0plus.small-multiply): Likewise.
985 (cortex-m1.small-multiply): Likewise.
987 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
988 Jakub Jelinek <jakub@redhat.com>
990 PR tree-optimization/85692
991 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
992 source permute as well.
994 2018-05-11 Martin Liska <mliska@suse.cz>
997 * doc/extend.texi: Document LLVM style format for no_sanitize
1000 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1002 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1003 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1004 (mode_supports_vsx_dform_quad): Likewise.
1005 (mode_supports_vmx_dform): Move these functions to be next to the
1006 other mode_supports functions.
1007 (mode_supports_dq_form): Likewise.
1008 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1009 mode_supports_dq_form.
1010 (reg_offset_addressing_ok_p): Likewise.
1011 (offsettable_ok_by_alignment): Likewise.
1012 (rs6000_legitimate_offset_address_p): Likewise.
1013 (legitimate_lo_sum_address_p): Likewise.
1014 (rs6000_legitimize_address): Likewise.
1015 (rs6000_legitimize_reload_address): Likewise.
1016 (rs6000_secondary_reload_inner): Likewise.
1017 (rs6000_preferred_reload_class): Likewise.
1018 (rs6000_output_move_128bit): Likewise.
1020 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1022 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1023 Generate SImode target register for null target.
1024 <case IX86_BUILTIN_XGETBV>: Ditto.
1025 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1026 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1028 2018-05-10 Carl Love <cel@us.ibm.com>
1030 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1031 dcbtt and dcbtstt if operands[2] is 0.
1033 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1036 * config/i386/sse.md (usadv64qi): New expander.
1038 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1040 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1041 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1042 -maltivec=be support.
1043 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1044 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1045 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1046 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1047 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1048 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1049 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1050 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1051 altivec_vsumsws): Adjust.
1052 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1053 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1054 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1056 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1057 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1058 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1059 (altivec_lve<VI_char>x): Delete expand.
1060 (*altivec_lve<VI_char>x_internal): Rename to...
1061 (altivec_lve<VI_char>x): ... this.
1062 (altivec_lvxl_<mode>): Delete expand.
1063 (*altivec_lvxl_<mode>_internal): Rename to ...
1064 (altivec_lvxl_<mode>): ... this.
1065 (altivec_stvxl_<mode>): Delete expand.
1066 (*altivec_stvxl_<mode>_internal): Rename to ...
1067 (altivec_stvxl_<mode>): ... this.
1068 (altivec_stve<VI_char>x): Delete expand.
1069 (*altivec_stve<VI_char>x_internal): Rename to ...
1070 (altivec_stve<VI_char>x): ... this.
1071 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1072 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1073 reduc_plus_scal_<mode>): Adjust.
1074 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1076 (rs6000_cpu_cpp_builtins): Adjust.
1077 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1078 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1079 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1080 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1081 -maltivec=be support.
1082 (rs6000_split_vec_extract_var): Adjust.
1083 (rs6000_split_v4si_init): Adjust.
1084 (swap_selector_for_mode): Delete.
1085 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1086 altivec_expand_stvex_be): Delete.
1087 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1088 -maltivec=be support.
1089 (rs6000_gimple_fold_builtin): Ditto.
1090 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1092 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1093 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1094 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1095 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1096 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1097 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1098 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1099 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1100 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1101 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1102 anonymous split): Adjust.
1103 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1104 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1106 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1108 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1109 when --with-gxx-include-dir is also specified.
1110 * configure: Regenerate.
1112 2018-05-09 Jim Wilson <jimw@sifive.com>
1115 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1116 * config/riscv/t-withmultilib: New.
1117 * config/riscv/withmultilib.h: New.
1118 * doc/install.texi: Document RISC-V --with-multilib-list support.
1120 2018-05-09 Richard Biener <rguenther@suse.de>
1122 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1124 (vect_bb_vectorization_profitable_p): Adjust. Compute
1125 actual scalar cost using the cost vector and the add_stmt_cost
1128 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1130 PR rtl-optimization/85645
1131 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1132 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1134 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1136 PR rtl-optimization/85645
1137 * shrink-wrap.c (spread_components): Return a boolean saying if
1138 anything was changed.
1139 (try_shrink_wrapping_separate): Iterate spread_components until
1140 nothing changes anymore.
1142 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1144 PR rtl-optimization/85645
1145 * regrename.c (build_def_use): Also kill the chains that include the
1146 destination of a REG_CFA_REGISTER note.
1148 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1150 PR rtl-optimization/85645
1151 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1152 insn that has a REG_CFA_REGISTER note.
1154 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1156 * cfgexpand.c (expand_clobber): New function.
1157 (expand_gimple_stmt_1): Use it.
1158 * tree-vect-stmts.c (vect_clobber_variable): New function,
1160 (vectorizable_simd_clone_call): ...here.
1161 (vectorizable_store): Emit a clobber either side of an
1162 IFN_STORE_LANES sequence.
1163 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1165 2018-05-09 Tom de Vries <tom@codesourcery.com>
1168 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1169 (define_insn "trap_if_false"): Add exit after trap.
1171 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1173 PR rtl-optimization/85638
1174 * bb-reorder.c: Include common/common-target.h.
1175 (create_forwarder_block): New function extracted from...
1176 (fix_up_crossing_landing_pad): ...here. Rename into...
1177 (dw2_fix_up_crossing_landing_pad): ...this.
1178 (sjlj_fix_up_crossing_landing_pad): New function.
1179 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1180 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1181 from both partitions and exit the loop after one iteration.
1183 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1186 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1188 (Basic PowerPC Built-in Functions): The new name of the
1189 subsection previously known as "PowerPC Built-in Functions".
1190 (Basic PowerPC Built-in Functions Available on all Configurations):
1192 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1194 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1196 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1198 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1201 2018-05-08 Jim Wilson <jimw@sifive.com>
1203 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1204 (LD_EMUL_SUFFIX): New.
1205 (LINK_SPEC): Use it.
1207 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1209 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1211 (Basic PowerPC Built-in Functions): The new name of the
1212 subsection previously known as "PowerPC Built-in Functions".
1213 (Basic PowerPC Built-in Functions Available on all Configurations):
1215 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1217 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1219 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1221 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1224 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1227 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1228 after cmpelim optimization.
1230 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1232 * config.gcc: Support "goldmont".
1233 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1234 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1236 * config/i386/i386.c (m_GOLDMONT): Define.
1237 (processor_target_table): Add "goldmont".
1238 (PTA_GOLDMONT): Define.
1239 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1240 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1241 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1242 (fold_builtin_cpu): Add "goldmont".
1243 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1244 (ix86_option_override_internal): Add "goldmont".
1245 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1246 (processor_type): Add PROCESSOR_GOLDMONT.
1247 * config/i386/i386.md: Add CPU "glm".
1248 * config/i386/glm.md: New file.
1249 * config/i386/x86-tune.def: Add m_GOLDMONT.
1250 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1252 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1255 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1257 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1258 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1259 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1262 * config/i386/i386.c (ix86_fold_builtin): Handle
1263 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1266 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1267 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1269 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1272 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1273 (check_arch): Likewise.
1274 (check_fpu): Return the result rather than printing it.
1275 (end arch): Fix operator precedence.
1276 (end cpu): Likewise.
1277 (END): Print the result from check_fpu.
1279 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1280 Alan Hayward <alan.hayward@arm.com>
1281 David Sherwood <david.sherwood@arm.com>
1283 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1284 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1285 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1286 (*fcmuo<mode>_and): New patterns.
1288 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1290 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1291 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1292 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1293 (cmp_op, sve_imm_con): New code attributes.
1294 (SVE_COND_INT_CMP, imm_con): Delete.
1295 (cmp_op): Remove above unspecs from int attribute.
1296 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1298 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1299 comparison-specific unspecs.
1300 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1301 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1302 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1303 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1304 (*vec_fcm<cmp_op><mode>): Rename to...
1305 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1306 (*vec_fcmuo<mode>): Rename to...
1307 (*fcmuo<mode>): ...this and adjust likewise.
1308 (*pred_fcm<cmp_op><mode>): New pattern.
1309 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1310 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1312 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1314 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1315 (aarch64_emit_sve_predicated_cond): New function.
1316 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1317 (aarch64_emit_unspec_cond_or): Replace with...
1318 (aarch64_emit_sve_or_conds): ...this new function. Use
1319 aarch64_emit_sve_ptrue_op for the individual comparisons and
1320 aarch64_emit_binop to OR them together.
1321 (aarch64_emit_inverted_unspec_cond): Replace with...
1322 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1323 aarch64_emit_sve_ptrue_op for the comparison and
1324 aarch64_emit_unop to invert the result.
1325 (aarch64_expand_sve_vec_cmp_float): Update after the above
1326 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1328 2018-05-07 Nathan Sidwell <nathan@acm.org>
1330 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1331 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1332 (Backwards Compatibility): Likewise.
1334 2018-05-07 Luis Machado <luis.machado@linaro.org>
1338 2018-05-07 Luis Machado <luis.machado@linaro.org>
1340 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1341 <prefetch_dynamic_strides>: New const bool field.
1342 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1343 prefetch_dynamic_strides.
1344 (exynosm1_prefetch_tune): Likewise.
1345 (thunderxt88_prefetch_tune): Likewise.
1346 (thunderx_prefetch_tune): Likewise.
1347 (thunderx2t99_prefetch_tune): Likewise.
1348 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1349 (aarch64_override_options_internal): Update to set
1350 PARAM_PREFETCH_DYNAMIC_STRIDES.
1351 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1352 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1353 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1354 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1355 prefetch-dynamic-strides setting.
1357 2018-05-07 Luis Machado <luis.machado@linaro.org>
1359 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1360 <minimum_stride>: New const int field.
1361 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1362 minimum_stride field.
1363 (exynosm1_prefetch_tune): Likewise.
1364 (thunderxt88_prefetch_tune): Likewise.
1365 (thunderx_prefetch_tune): Likewise.
1366 (thunderx2t99_prefetch_tune): Likewise.
1367 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1368 (aarch64_override_options_internal): Update to set
1369 PARAM_PREFETCH_MINIMUM_STRIDE.
1370 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1371 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1372 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1373 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1374 stride is constant and is below the minimum stride threshold.
1376 2018-05-07 Luis Machado <luis.machado@linaro.org>
1378 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1381 2018-05-07 Luis Machado <luis.machado@linaro.org>
1383 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1384 <prefetch_dynamic_strides>: New const bool field.
1385 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1386 prefetch_dynamic_strides.
1387 (exynosm1_prefetch_tune): Likewise.
1388 (thunderxt88_prefetch_tune): Likewise.
1389 (thunderx_prefetch_tune): Likewise.
1390 (thunderx2t99_prefetch_tune): Likewise.
1391 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1392 (aarch64_override_options_internal): Update to set
1393 PARAM_PREFETCH_DYNAMIC_STRIDES.
1394 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1395 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1396 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1397 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1398 prefetch-dynamic-strides setting.
1400 2018-05-07 Luis Machado <luis.machado@linaro.org>
1402 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1403 <minimum_stride>: New const int field.
1404 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1405 minimum_stride field.
1406 (exynosm1_prefetch_tune): Likewise.
1407 (thunderxt88_prefetch_tune): Likewise.
1408 (thunderx_prefetch_tune): Likewise.
1409 (thunderx2t99_prefetch_tune): Likewise.
1410 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1411 (aarch64_override_options_internal): Update to set
1412 PARAM_PREFETCH_MINIMUM_STRIDE.
1413 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1414 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1415 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1416 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1417 stride is constant and is below the minimum stride threshold.
1419 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1422 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1423 the type is addressable. Don't force op into register if it has
1426 2018-05-05 Roland McGrath <mcgrathr@google.com>
1429 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1430 any section for which we don't know a specific type it should have,
1431 regardless of name. Previously this was done only for the exact
1432 names ".init_array", ".fini_array", and ".preinit_array".
1433 (default_elf_asm_named_section): Add comment about
1434 relationship with default_section_type_flags and SECTION_NOTYPE.
1435 (get_section): Don't consider it a type conflict if one side has
1436 SECTION_NOTYPE and the other doesn't, as long as neither has the
1437 SECTION_BSS et al used in the default_section_type_flags logic.
1439 2018-05-05 Tom de Vries <tom@codesourcery.com>
1442 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1443 (workaround_barsyncs): New function.
1444 (nvptx_reorg): Use workaround_barsyncs.
1445 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1446 (define_expand "nvptx_membar_cta"): New define_expand.
1447 (define_insn "*nvptx_membar_cta"): New insn.
1449 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1451 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1452 To improve optimization opportunities.
1453 * builtin-types.def: The new needed builtin types for the above.
1455 2018-05-04 Richard Biener <rguenther@suse.de>
1457 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1458 * gimple-ssa-store-merging.c
1459 (imm_store_chain_info::output_merged_store): Remove redundant create,
1460 release split_store vector contents on failure.
1461 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1462 scalar stmt vector on cache hit.
1464 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1466 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1468 * config.gcc (powerpc-xilinx-eabi*): Remove.
1469 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1471 (fusion_addis_mem_combo_load): Ditto.
1472 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1474 (rs6000_cpu_cpp_builtins): Ditto.
1475 * config/rs6000/rs6000-linux.c
1476 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1477 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1478 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1480 (rs6000_setup_reg_addr_masks): Ditto.
1481 (rs6000_init_hard_regno_mode_ok): Ditto.
1482 (rs6000_option_override_internal): Ditto.
1483 (legitimate_lo_sum_address_p): Ditto.
1484 (rs6000_legitimize_address): Ditto.
1485 (rs6000_legitimize_reload_address): Ditto.
1486 (rs6000_legitimate_address_p): Ditto.
1487 (abi_v4_pass_in_fpr): Ditto.
1488 (setup_incoming_varargs): Ditto.
1489 (rs6000_gimplify_va_arg): Ditto.
1490 (rs6000_split_multireg_move): Ditto.
1491 (rs6000_savres_strategy): Ditto.
1492 (rs6000_emit_prologue_components): Ditto.
1493 (rs6000_emit_epilogue_components): Ditto.
1494 (rs6000_emit_prologue): Ditto.
1495 (rs6000_emit_epilogue): Ditto.
1496 (rs6000_elf_file_end): Ditto.
1497 (rs6000_function_value): Ditto.
1498 (rs6000_libcall_value): Ditto.
1499 * config/rs6000/rs6000.h: Ditto.
1500 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1501 (TARGET_MINMAX): ... this. New.
1502 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1503 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1504 (*movsi_internal1_single): Delete.
1505 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1506 mfpu=, mxilinx-fpu): Delete.
1507 * config/rs6000/singlefp.h: Delete.
1508 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1509 * config/rs6000/t-rs6000: Ditto.
1510 * config/rs6000/t-xilinx: Delete.
1511 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1512 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1513 (VStype_simple): Delete.
1514 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1515 * config/rs6000/xfpu.h: Delete.
1516 * config/rs6000/xfpu.md: Delete.
1517 * config/rs6000/xilinx.h: Delete.
1518 * config/rs6000/xilinx.opt: Delete.
1519 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1520 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1522 2018-05-04 Tom de Vries <tom@codesourcery.com>
1525 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1528 2018-05-04 Richard Biener <rguenther@suse.de>
1531 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1532 (expand_complex_div_wide): Likewise.
1533 (expand_complex_operations_1): Likewise.
1534 (expand_complex_libcall): Preserve EH info of the original stmt.
1535 (tree_lower_complex): Handle removed blocks.
1536 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1537 on complex multiplication and division libcall builtins.
1539 2018-05-04 Richard Biener <rguenther@suse.de>
1542 * fold-const.c (negate_expr_p): Restrict negation of operand
1543 zero of a division to when we know that can happen without
1545 (fold_negate_expr_1): Likewise.
1547 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1550 * real.h (real_nextafter): Declare.
1551 * real.c (real_nextafter): New function.
1552 * fold-const-call.c (fold_const_nextafter): New function.
1553 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1554 CASE_CFN_NEXTTOWARD.
1555 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1556 even when arg1_mode is different from arg0_mode.
1558 2018-05-03 Nathan Sidwell <nathan@acm.org>
1560 * doc/extend.texi (Deprecated Features): Remove
1562 (Backwards Compatibility): Likewise.
1563 * doc/invoke.texi (C++ Language Options): Likewise.
1564 (C++ Dialect Options): Likewise.
1566 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1569 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1570 _mm512_mask_mullox_epi64): New intrinsics.
1572 2018-05-03 Tom de Vries <tom@codesourcery.com>
1575 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1576 dump files): Add offload-tree.
1578 2018-05-03 Richard Biener <rguenther@suse.de>
1580 PR tree-optimization/85615
1581 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1582 to loops not nested in BBs loop father to avoid creating multi-entry
1585 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1587 PR tree-optimization/70291
1588 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1589 arguments. Change return type to tree. Emit libcall as a new
1590 statement rather than replacing existing one when inplace_p is true.
1591 (expand_complex_multiplication_components): New function.
1592 (expand_complex_multiplication): Expand floating-point complex
1593 multiplication using the above.
1594 (expand_complex_division): Rename inner_type parameter to type.
1595 Update expand_complex_libcall call-site.
1596 (expand_complex_operations_1): Update expand_complex_multiplication
1597 and expand_complex_division call-sites.
1599 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1602 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1603 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1604 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1605 the highest significant bit of the shift count mask is clear. In
1606 check whether and[sq]i3 is needed verify that all significant bits
1607 of the shift count other than the highest are set.
1609 2018-05-02 Tom de Vries <tom@codesourcery.com>
1612 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1613 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1614 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1615 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1616 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1617 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1618 __builtin_goacc_parlevel_size.
1620 2018-05-02 Richard Biener <rguenther@suse.de>
1622 PR tree-optimization/85597
1623 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1624 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1627 2018-05-02 Tom de Vries <tom@codesourcery.com>
1630 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1631 dump files): Add ltrans-tree.
1633 2018-05-02 Tom de Vries <tom@codesourcery.com>
1636 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1637 dump files): Add wpa-ipa.
1639 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1641 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1642 powerpc*-*-linux*paired* target.
1643 * config/rs6000/750cl.h: Delete.
1644 * config/rs6000/paired.h: Delete.
1645 * config/rs6000/paired.md: Delete.
1646 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1648 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1649 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1650 comment. Remove paired float support.
1651 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1652 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1654 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1655 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1657 * config/rs6000/rs6000.c: Remove paired float support.
1658 (paired_expand_vector_init, paired_expand_vector_move,
1659 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1660 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1661 paired_expand_builtin, paired_expand_predicate_builtin,
1662 paired_init_builtins): Delete.
1663 * config/rs6000/rs6000.h: Remove paired float support.
1664 * config/rs6000/rs6000.md: Remove paired float support.
1665 (move_from_CR_ov_bit): Delete.
1666 * config/rs6000/rs6000.opt (mpaired): Delete.
1667 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1668 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1670 2018-05-02 Richard Biener <rguenther@suse.de>
1673 * gimplify.c (gimplify_save_expr): When in SSA form allow
1674 SAVE_EXPRs to compute to SSA vars.
1676 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1679 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1680 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1681 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1682 clobber operands[2], instead use a new pseudo. Formatting fixes.
1684 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1686 PR tree-optimization/85586
1687 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1688 exit early for statements in the same group if the accesses are
1691 2018-05-02 Tom de Vries <tom@codesourcery.com>
1694 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1697 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1699 PR tree-optimization/85143
1700 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1702 2018-05-01 Tom de Vries <tom@codesourcery.com>
1705 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1706 not found" error message.
1708 2018-05-01 Tom de Vries <tom@codesourcery.com>
1711 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1712 * vec.c (test_ordered_remove_if): New function.
1713 (vec_c_tests): Call test_ordered_remove_if.
1714 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1715 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1716 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1717 VEC_ORDERED_REMOVE_IF.
1719 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1721 PR tree-optimization/82665
1722 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1723 pointer subtraction where arguments come from a memchr call.
1725 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1727 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1728 --push-state --as-needed and --pop-state instead of --as-needed and
1729 --no-as-needed if ld supports it.
1730 * configure: Regenerated.
1733 * doc/install.texi2html: Replace _002d with - and _002a with * in
1734 generated html files using sed.
1736 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1739 * gcc-rich-location.c (blank_line_before_p): New function.
1740 (use_new_line): New function.
1741 (gcc_rich_location::add_fixit_insert_formatted): New function.
1742 * gcc-rich-location.h
1743 (gcc_rich_location::add_fixit_insert_formatted): New function.
1745 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1747 * selftest.c (assert_streq): Rename "expected" and "actual" to
1748 "val1" and "val2". Extend NULL-handling to cover both inputs
1749 symmetrically, while still requiring both to be non-NULL for a pass.
1750 * selftest.h (assert_streq): Rename "expected" and "actual" to
1752 (ASSERT_EQ): Likewise.
1753 (ASSERT_EQ_AT): Likewise.
1754 (ASSERT_KNOWN_EQ): Likewise.
1755 (ASSERT_KNOWN_EQ_AT): Likewise.
1756 (ASSERT_NE): Likewise.
1757 (ASSERT_MAYBE_NE): Likewise.
1758 (ASSERT_MAYBE_NE_AT): Likewise.
1759 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1760 the assertion to pass.
1761 (ASSERT_STREQ_AT): Likewise.
1763 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1765 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1766 interaction with -pie.
1768 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1770 * selftest.h: Fix alphabetization of per-source-file selftest
1773 2018-04-30 Jason Merrill <jason@redhat.com>
1775 PR c++/61982 - dead stores to destroyed objects.
1776 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1779 2018-04-30 Jason Merrill <jason@redhat.com>
1781 * tree.c (build_clobber): New.
1782 * tree.h: Declare it.
1783 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1785 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1787 * diagnostic-show-locus.c (layout::layout): Update for
1788 location_get_source_line returning a char_span.
1789 (struct char_span): Move to input.h.
1790 (struct correction): Update for fields in char_span becoming
1792 (struct source_line): Update for location_get_source_line
1793 returning a char_span.
1794 (layout::print_line): Likewise.
1795 * edit-context.c (edited_file::print_content): Likewise.
1796 (edited_file::print_diff_hunk): Likewise.
1797 (edited_file::print_run_of_changed_lines): Likewise.
1798 (edited_file::get_num_lines): Likewise.
1799 (edited_line::edited_line): Likewise.
1800 * final.c (asm_show_source): Likewise.
1801 * input.c (location_get_source_line): Convert return type
1802 from const char * to char_span, losing the final "line_len"
1804 (dump_location_info): Update for the above.
1805 (get_substring_ranges_for_loc): Likewise. Use a char_span
1806 when handling the literal within the line.
1807 (test_reading_source_line): Update for location_get_source_line
1808 returning a char_span.
1809 * input.h (class char_span): Move here from
1810 diagnostic-show-locus.c, converting from a struct to a class.
1811 Make data members private.
1812 (char_span::operator bool): New.
1813 (char_span::length): New.
1814 (char_span::get_buffer): New.
1815 (char_span::operator[]): New.
1816 (char_span::subspan): Make const.
1817 (char_span::xstrdup): New.
1818 (location_get_source_line): Convert return type from const char *
1819 to char_span, losing the final "line_size" param.
1821 2018-04-30 Jan Hubicka <jh@suse.cz>
1823 * lto-wrapper.c (ltrans_priorities): New static var.
1824 (cmp_priority): New.
1825 (run_gcc): Read priorities and if doing parallel build order
1826 the Makefile by them.
1828 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1830 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1832 2018-04-30 Richard Biener <rguenther@suse.de>
1834 * tree-cfg.c (verify_address): Remove base argument, add
1835 flag whether to check TREE_ADDRESSABLE and do that.
1836 (verify_expr): Remove.
1837 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1838 (verify_gimple_assign_single): Likewise.
1839 (verify_gimple_switch): Likewise.
1840 (verify_expr_location_1): Dereference tp once. Add (disabled)
1841 piece from verify_expr.
1842 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1844 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1846 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1848 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1850 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1851 (small_data_pattern): Likewise.
1852 (arc_rewrite_small_data): Likewise.
1853 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1854 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1855 (get_symbol_alignment): New function.
1856 (legitimate_small_data_address_p): Likewise.
1857 (legitimate_scaled_address): Update, call
1858 legitimate_small_data_address_p.
1859 (output_sdata): New static variable.
1860 (arc_print_operand): Update how we handle small data operands.
1861 (arc_print_operand_address): Likewise.
1862 (arc_legitimate_address_p): Update, use
1863 legitimate_small_data_address_p.
1864 (arc_rewrite_small_data_p): Remove.
1865 (arc_rewrite_small_data_1): Likewise.
1866 (arc_rewrite_small_data): Likewise.
1867 (small_data_pattern): Likewise.
1868 (compact_sda_memory_operand): Update to use
1869 legitimate_small_data_address_p and get_symbol_alignment.
1870 (prepare_move_operands): Don't rewite sdata pattern.
1871 (prepare_extend_operands): Remove.
1872 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1874 (zero_extendqisi2): Likewise.
1875 (zero_extendhisi2): Likewise.
1876 (extendqihi2): Likewise.
1877 (extendqisi2): Likewise.
1878 (extendhisi2): Likewise.
1882 * config/arc/constraints.md (Usd): Change it to memory constraint.
1884 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1886 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1887 as source of std instructions.
1888 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1889 allow 6-bit constants as source for store instructions.
1890 (movdi_insn): Update instruction pattern to allow 6-bit constants
1891 as source for store instructions.
1893 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1895 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1897 2018-04-30 Nathan Sidwell <nathan@acm.org>
1898 Sandra Loosemore <sandra@codesourcery.com>
1900 * dumpfile.c (dump_open): Allow '-' for stdout.
1901 * doc/invoke.texi (Developer Options): Document dump filename
1902 determination early. Document stdin/stdout selection.
1904 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1906 Microblaze Target: PIC data text relative
1908 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1909 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1911 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1912 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1913 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1914 New addressing mode for data-text relative position indepenedent code.
1915 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1916 'ADDRESS_SYMBOLIC_TXT_REL'.
1917 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1918 (microblaze_legitimate_pic_operand): Exclude function calls from
1919 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1920 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1922 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1923 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1924 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1925 for 'address + offset'.
1926 (microblaze_expand_prologue): Add new function prologue call for
1928 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1929 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1930 table in case of TARGET_PIC_DATA_TEXT_REL.
1931 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1932 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1933 Add new macros 'UNSPEC_TEXT',
1934 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1935 + exclude function calls from 'UNSPEC_PLT' in case of data text
1937 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1938 new target hook for generating address diff vector tables in case of
1940 * doc/tm.texi : Regenerate.
1941 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1942 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1943 of addr diff vector generation.
1944 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1945 target hook definition.
1946 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1947 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1948 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1949 option for data text relative.
1951 2018-04-30 Richard Biener <rguenther@suse.de>
1953 * tree-chrec.h (evolution_function_is_constant_p): Remove
1955 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1957 2018-04-30 Richard Biener <rguenther@suse.de>
1960 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1962 2018-04-30 Richard Biener <rguenther@suse.de>
1964 PR tree-optimization/28364
1965 PR tree-optimization/85275
1966 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1967 copying first exit test.
1969 2018-04-28 Mark Wielaard <mark@klomp.org>
1971 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1973 (dwarf_AT): Handle DW_AT_addr_base.
1974 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1976 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1979 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1980 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1981 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1982 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1984 2018-04-28 Richard Biener <rguenther@suse.de>
1986 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1987 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1988 to reflect use. Only add interesting stmts.
1990 2018-04-27 Martin Jambor <mjambor@suse.cz>
1993 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1994 the jump function allows for passing through aggregate values.
1996 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1998 * input.h (in_system_header_at): Convert from macro to inline
2000 (from_macro_expansion_at): Likewise.
2001 (from_macro_definition_at): Likewise.
2003 2018-04-27 Jeff Law <law@redhat.com>
2005 * config.gcc: Mark tile* targets as deprecated/obsolete.
2007 2018-04-27 Richard Biener <rguenther@suse.de>
2009 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2012 2018-04-27 Richard Biener <rguenther@suse.de>
2014 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2016 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2018 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2019 with Yd constraint. Set "preferred_for_speed" attribute from
2020 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2022 (*movdi_internal): Ditto.
2023 (movti_interunit splitters): Remove
2024 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2025 (movdi_interunit splitters): Ditto.
2026 * config/i386/constraints.md (Ye): Remove.
2027 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2029 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2032 * config/aarch64/constraints.md (Usg): Limit to 31.
2035 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2037 PR tree-optimization/85529
2038 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2039 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2040 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2041 zero extension or masking of the MSB bit.
2042 (optimize_range_tests): Add FIRST_BB argument, pass it through
2043 to optimize_range_tests_var_bound.
2044 (maybe_optimize_range_tests, reassociate_bb): Adjust
2045 optimize_range_tests callers.
2047 2018-04-26 Richard Biener <rguenther@suse.de>
2048 Jakub Jelinek <jakub@redhat.com>
2050 * cgraph.h (symbol_table): Just declare debug method here.
2051 * symtab.c (symbol_table::debug): Define.
2053 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2055 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2057 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2059 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2060 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2061 (*movdi_internal): Substitute Yi and Yj constraint with x
2062 and Ym and Yn constraint with y constraint. Update "isa"
2063 attribute and set "preferred_for_speed" attribute from
2064 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2065 (*movsi_internal): Ditto.
2066 (*movdf_internal): Ditto.
2067 (*movsf_internal): Ditto.
2068 (*zero_extendsidi2): Ditto.
2069 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2070 (sse2_loadld): Ditto.
2071 (*vec_extract<ssevecmodelower>_0): Ditto.
2072 (*vec_extractv4si_0_zext_sse4): Ditto.
2073 (vec_concatv2di): Ditto.
2074 (*vec_dup<mode>): Ditto.
2075 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2076 * config/i386/constraints.md (Yi): Remove.
2081 2018-04-26 Nathan Sidwell <nathan@acm.org>
2083 * dumpfile.c (dump_open): New.
2084 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2085 (dump_finish): Detect stdio/stderr by value not name.
2087 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2089 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2091 2018-04-26 Tom de Vries <tom@codesourcery.com>
2094 * config/nvptx/nvptx.c (verify_neutering_jumps)
2095 (verify_neutering_labels): New function
2096 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2098 2018-04-26 Tom de Vries <tom@codesourcery.com>
2101 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2102 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2105 2018-04-26 Richard Biener <rguenther@suse.de>
2106 Tom de Vries <tom@codesourcery.com>
2109 * lto-streamer-out.c (output_function): Fixup loops if required to match
2110 discovery done in the reader.
2112 2018-04-26 Richard Biener <rguenther@suse.de>
2114 PR tree-optimization/85116
2115 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2116 have a loop exit from the single latch predecessor. Remove
2117 case of header with just condition.
2118 (ch_base::copy_headers): Exclude infinite loops from any
2120 (pass_ch::execute): Record exits.
2122 2018-04-26 Richard Biener <rguenther@suse.de>
2124 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2125 prologue cost vector and pass it to vect_get_load_cost.
2126 (vect_get_peeling_costs_all_drs): Likewise.
2127 (vect_peeling_hash_get_lowest_cost): Likewise.
2128 (vect_enhance_data_refs_alignment): Likewise.
2130 2018-04-26 Richard Biener <rguenther@suse.de>
2133 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2134 checking of integer<->pointer conversions.
2135 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2136 sign-/zero-extending pointer types.
2137 (expand_omp_for_static_chunk): Likewise.
2139 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2140 Jean Lee <xiaoyur347@gmail.com>
2142 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2143 (TARGET_ASAN_SHADOW_OFFSET): Define.
2144 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2145 true for -fsanitize=address.
2147 2018-04-25 Mark Wielaard <mark@klomp.org>
2149 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2152 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2154 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2155 than "alu", remove explicit "memory" and "imm_disp" attributes.
2156 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2159 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2160 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2161 gen_lowpart_no_emit.
2163 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2166 * config/i386/i386.c (ix86_expand_builtin): Change memory
2167 operand to XI, extend p0 to Pmode.
2168 * config/i386/i386.md: Change unspec volatile and operand
2169 1 mode to XI, change operand 0 mode to P.
2171 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2173 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2174 GET_MODE_MASK before any checking.
2175 (nds32_can_use_bset_p): Likewise.
2176 (nds32_can_use_btgl_p): Likewise.
2178 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2180 * config/nds32/nds32-doubleword.md: New define_split pattern for
2181 illegal register number.
2183 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2185 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2187 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2189 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2191 2018-04-25 Richard Biener <rguenther@suse.de>
2193 * lto-streamer.h (LTO_major_version): Bump to 8.
2195 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2197 * BASE-VER: Set to 9.0.0.
2199 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2201 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2202 in __abskf2 and __powikf2.
2204 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2207 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2208 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2209 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2210 Use the above on operand 2. Reindent.
2211 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2213 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2216 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2218 * config/i386/i386.opt (mcet): Removed.
2219 * doc/install.texi: Remove -mcet documentation.
2220 * doc/invoke.texi: Likewise.
2222 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2225 * doc/install.texi: Remove -mcet from bootstrap-cet.
2227 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2230 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2231 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2235 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2236 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2237 containing a CONST_VECTOR.
2239 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2241 * doc/install.texi: Update newlib dependency for nvptx.
2243 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2246 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2247 instead of INTVAL when shifting x left.
2249 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2251 PR tree-optimization/85478
2252 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2253 vect_grouped_store_supported for single element vectors.
2255 2018-04-24 Richard Biener <rguenther@suse.de>
2258 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2259 load cost increase to the case of non-constant step.
2261 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2264 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2265 destination if any_malformed_asm.
2267 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2270 * expr.c (store_field): In the bitfield case, if the value comes from
2271 a function call and is returned in registers by means of a PARALLEL,
2272 do not change the mode of the temporary unless BLKmode and VOIDmode.
2274 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2276 PR rtl-optimization/85423
2277 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2278 dependencies to debug insns when the previous insn is non-debug.
2280 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2282 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2283 enums into a single definition.
2284 (fls): Fix predicates and printing.
2287 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2289 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2290 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2291 and short u6 immediate.
2292 (check_if_valid_sleep_operand): Remove.
2293 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2295 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2297 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2298 flag_always_save_lp condition.
2299 * config/nds32/nds32.opt (malways-save-lp): New option.
2301 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2303 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2304 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2305 * config/nds32/nds32.h
2306 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2307 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2309 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2311 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2312 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2314 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2315 Chung-Ju Wu <jasonwucj@gmail.com>
2317 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2319 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2320 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2322 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2324 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2326 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2328 * config/nds32/nds32-protos.h (nds32_data_alignment,
2329 nds32_local_alignment): Declare.
2330 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2331 nds32_local_alignment): New functions.
2332 (TARGET_CONSTANT_ALIGNMENT): Define.
2333 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2335 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2337 * config/nds32/nds32.c
2338 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2339 (TARGET_MODES_TIEABLE_P): Likewise.
2341 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2343 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2346 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2347 Chung-Ju Wu <jasonwucj@gmail.com>
2349 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2350 for unaligned access.
2351 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2353 * config/nds32/nds32-intrinsic.md: Likewise.
2354 * config/nds32/nds32_intrinsic.h: Likewise.
2355 * config/nds32/nds32.h (nds32_builtins): Likewise.
2356 * config/nds32/nds32.opt (munaligned-access): New option.
2357 * config/nds32/nds32.c (nds32_asm_file_start): Display
2358 flag_unaligned_access status.
2360 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2362 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2363 -mno-relax is present.
2364 * config/riscv/linux.h (LINK_SPEC): Ditto.
2366 2018-04-20 Martin Sebor <msebor@redhat.com>
2369 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2371 (gimple_fold_builtin_stxcpy_chk): Same.
2372 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2374 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2377 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2378 __powikf2 when long double is IEEE 128-bit.
2380 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2382 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2383 step to make sure stack always aligned.
2385 2018-04-20 Carl Love <cel@us.ibm.com>
2388 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2389 size check for arg0.
2391 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2392 Tom de Vries <tom@codesourcery.com>
2395 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2396 Emit insns for calls too.
2397 (nvptx_find_par): Always look for worker-level predecessor insn.
2398 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2400 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2401 (nvptx_process_pars): Propagate frames for calls.
2403 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2406 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2408 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2409 (ix86_handle_option): Don't handle OPT_mibt.
2410 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2412 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2414 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2415 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2416 (ix86_target_macros): Define __CET__ with flag_cf_protection
2417 for -fcf-protection.
2418 * config/i386/i386.c (isa2_opts): Remove -mibt.
2419 * config/i386/i386.h (TARGET_IBT): Removed.
2420 (TARGET_IBT_P): Likewise.
2421 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2422 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2423 * config/i386/i386.opt (mcet): Update help message.
2426 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2427 -mcet as an alias for -mshstk.
2429 2018-04-20 Richard Biener <rguenther@suse.de>
2432 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2433 complexity by forcing a single use of the multiply operand.
2435 2018-04-20 Martin Jambor <mjambor@suse.cz>
2438 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2439 recursion dependency to only apply to non-clones.
2441 2018-04-20 Martin Jambor <mjambor@suse.cz>
2444 * ipa-cp.c (create_specialized_node): Check that clones of
2445 self-recursive edges exist during IPA-CP.
2447 2018-04-19 Toon Moene <toon@moene.org>
2449 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2452 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2454 PR tree-optimization/85467
2455 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2456 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2457 VECTOR_CST element to type.
2459 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2462 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2463 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2464 (builtin_longjmp): Likewise.
2465 (save_stack_nonlocal): New pattern.
2466 (restore_stack_nonlocal): Likewise.
2468 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2471 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2472 Replace ASM_OUTPUT_LABEL with fprintf.
2474 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2477 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2478 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2479 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2480 define __IBT__ and __SHSTK__ for -fcf-protection.
2481 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2483 (ix86_trampoline_init): Likewise.
2484 (x86_output_mi_thunk): Likewise.
2485 (ix86_notrack_prefixed_insn_p): Likewise.
2486 (ix86_option_override_internal): Don't disallow -fcf-protection.
2487 * config/i386/i386.md (rdssp<mode>): Also enable for
2489 (incssp<mode>): Likewise.
2490 (nop_endbr): Likewise.
2491 * config/i386/i386.opt (mcet): Change help message to built-in
2495 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2496 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2497 enable CET built-in functions.
2499 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2501 * common/config/i386/i386-common.c
2502 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2503 OPTION_MASK_ISA_MOVDIRI_UNSET,
2504 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2505 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2506 * config.gcc (movdirintrin.h): New header.
2507 * config/i386/cpuid.h (bit_MOVDIRI,
2508 bit_MOVDIR64B): New bits.
2509 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2511 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2512 (VOID, PVOID, PCVOID)): New function types.
2513 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2514 __builtin_ia32_directstoreu_u64,
2515 __builtin_ia32_movdir64b): New builtins.
2516 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2517 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2519 (ix86_valid_target_attribute_inner_p): Ditto.
2520 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2521 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2522 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2523 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2524 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2525 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2526 (movdiri<mode>, movdir64b_<mode>): New.
2527 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2528 * config/i386/immintrin.h: Include movdirintrin.h.
2529 * config/i386/movdirintrin.h: New file.
2530 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2532 2018-04-19 Richard Biener <rguenther@suse.de>
2535 * cfg.c (clear_bb_flags): When loop state says we have
2536 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2538 2018-04-19 Richard Biener <rguenther@suse.de>
2540 PR tree-optimization/84737
2541 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2542 copying restrict info.
2543 (vect_setup_realignment): Use it.
2544 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2545 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2546 the first DR to all generated stores.
2547 (vectorizable_load): Likewise for loads.
2549 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2551 PR tree-optimization/85446
2552 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2553 the integral and pointer types to have the same precision.
2555 * doc/install.texi: Document --disable-cet being the default and
2558 2018-04-18 Martin Liska <mliska@suse.cz>
2560 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2563 2018-04-18 Martin Liska <mliska@suse.cz>
2566 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2569 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2570 arguments if they are comparable.
2572 2018-04-18 Martin Liska <mliska@suse.cz>
2575 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2578 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2581 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2584 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2585 ENDBR after calling __morestack.
2587 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2590 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2591 by using gcc_base_ver to generate a gcc_driver_version, and use
2592 it when generating GCC_DRIVER_NAME.
2593 * configure: Regenerate.
2595 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2598 * config.gcc: Obsolete powerpc*-*-*spe*.
2600 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2603 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2604 (stabstr_D): Change type of unum from unsigned int to
2605 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2608 2018-04-17 Jim Wilson <jimw@sifive.com>
2611 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2612 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2613 Set arg_pointer_offset after using pretend_args_size.
2615 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2617 PR rtl-optimization/85431
2618 * dse.c (record_store): Ignore zero width stores.
2621 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2622 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2623 __builtin_stack_restore rather than after it.
2624 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2625 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2626 argument instead of virtual_dynamic_stack_rtx.
2628 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2630 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2632 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2633 Add note to error message to explain internal mapping of overloaded
2634 built-in function name to non-overloaded built-in function name.
2635 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2638 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2641 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2642 where the inputs overlap with the output.
2644 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2646 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2647 (=v, v) alternative and explicit "memory" attribute.
2648 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2649 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2651 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2652 "sselog1" type instead of "sselog".
2653 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2654 "sselog". Remove explicit "memory" attribute.
2655 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2656 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2658 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2659 "sselog1" type instead of "sselog". Remove explicit "memory"
2661 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2662 use "sselog1" type instead of "sselog". Remove explicit "memory"
2664 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2665 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2667 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2668 "sselog1" type instead of "sselog". Remove explicit "memory"
2670 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2671 use "sselog1" type instead of "sselog". Remove explicit "memory"
2675 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2678 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2681 2018-04-17 Martin Jambor <mjambor@suse.cz>
2684 * ipa-cp.c (create_specialized_node): Call
2685 expand_all_artificial_thunks if necessary.
2687 2018-04-17 Martin Liska <mliska@suse.cz>
2690 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2691 in message, remote space in between '_G' and '('.
2693 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2696 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2697 avx512f_vmcmp<mode>3<round_saeonly_name>,
2698 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2699 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2700 avx512f_rndscale<mode><round_saeonly_name>,
2701 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2702 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2703 Use %<iptr>2 instead of %2 for -masm=intel.
2704 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2705 avx512f_vcvttss2usi<round_saeonly_name>,
2706 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2708 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2709 avx512f_vcvttsd2usi<round_saeonly_name>,
2710 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2711 Use %q1 instead of %1 for -masm=intel.
2712 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2713 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2714 of %3 for -masm=intel.
2715 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2717 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2719 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2721 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2722 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2724 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2726 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2727 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2728 %g1 and one with %0 and %1.
2729 (avx512er_vmrcp28<mode><round_saeonly_name>,
2730 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2732 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2733 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2734 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2735 of %0 and %{%4%} for -masm=intel.
2736 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2737 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2738 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2739 order of %0 and %{%5%}%{z%} for -masm=intel.
2741 2018-04-17 Jan Hubicka <jh@suse.cz>
2744 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2746 2018-04-17 Martin Liska <mliska@suse.cz>
2749 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2750 for target_clone error message. Make default implementation
2751 clone to be a local declaration.
2752 (separate_attrs): Add new argument and check for an empty
2754 (expand_target_clones): Handle it.
2755 (ipa_target_clone): Make redirection just for target_clones
2758 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2759 Tom de Vries <tom@codesourcery.com>
2762 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2763 tiled basic blocks without omp continue statements.
2765 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2768 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2769 vec_extract expression as having side effects to make sure it gets
2772 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2775 * config/i386/i386.c (get_builtin_code_for_version): Check
2778 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2781 * gcc/config.gcc: Support "skylake".
2782 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2784 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2785 (processor_target_table): Add "skylake".
2786 (ix86_option_override_internal): Add "skylake".
2787 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2788 PROCESSOR_CANNONLAKE.
2789 (get_builtin_code_for_version): Fix priority for
2790 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2791 PROCESSOR_SKYLAKE-AVX512.
2792 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2793 (processor_type): Add PROCESSOR_SKYLAKE.
2795 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2796 Jason Merrill <jason@redhat.com>
2799 * convert.c (convert_to_integer_1): Use direct recursion for
2800 enumeral types and types with a precision less than the number
2801 of bits in their mode.
2803 2018-04-16 Julia Koval <julia.koval@intel.com>
2806 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2807 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2809 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2812 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2813 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2814 and -mno-direct-move.
2816 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2819 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2820 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2821 Ensure negative shifts result in {0}.
2823 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2825 PR rtl-optimization/79916
2826 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2827 regs (if any) to define how to gnerate SD moves when LRA is in
2830 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2832 PR rtl-optimization/85393
2833 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2834 * except.c (expand_dw2_landing_pad_for_region): Make static.
2835 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2836 a label and unconditional jump to old_bb, rather than
2837 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2840 PR rtl-optimization/85376
2841 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2842 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2843 instead of a specific value.
2845 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2846 Bin Cheng <bin.cheng@arm.com>
2848 PR tree-optimization/82965
2849 PR tree-optimization/83991
2850 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2851 by_profile_only parameter.
2852 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2853 information if the loop was predicted to iterate too many times.
2854 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2856 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2859 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2862 2018-04-13 Martin Liska <mliska@suse.cz>
2863 Jakub Jelinek <jakub@redhat.com>
2866 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2867 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2868 * builtins.c (expand_builtin_memory_copy_args): Use
2869 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2870 handle dest_addr == pc_rtx.
2872 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2875 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2876 asked to not generate direct moves.
2877 (fix_trunc<mode>si2_stfiwx): Similar.
2878 (fix_trunc<mode>si2_internal): Similar.
2880 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2883 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2884 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2885 lookup if dest in some wider mode is known to be const0_rtx and
2886 if so, record permanent equivalence for it to be ZERO_EXTEND of
2887 the narrower mode destination.
2889 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2891 * lto-streamer-out.c (output_function): Revert 259346.
2892 * omp-expand.c (expand_oacc_for): Likewise.
2894 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2896 PR rtl-optimization/85354
2897 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2898 * sel-sched.c (sel_global_init): ... here.
2900 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2903 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2904 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2905 mode for PE-COFF targets.
2906 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2907 (i386_pe_asm_lto_end): Likewise.
2908 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2909 (TARGET_ASM_LTO_END): Likewise.
2910 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2911 (i386_pe_asm_lto_start): New function.
2912 (i386_pe_asm_lto_end): Likewise.
2914 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2915 Richard Biener <rguenther@suse.de>
2918 * lto-streamer-out.c (output_function): Fix CFG loop state before
2920 * omp-expand.c (expand_oacc_for): Handle calls to internal
2921 functions like regular functions.
2923 2018-04-12 Richard Biener <rguenther@suse.de>
2926 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2927 for the early LTO debug to properly generate references to it
2928 during DIE emission. Do not re-use that for the skeleton for
2930 (dwarf2out_early_finish): Likewise.
2932 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2935 * config/i386/sse.md
2936 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2937 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2938 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2939 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2940 and output is a reg, avoid creating invalid lowpart subreg, but
2941 instead split into a 512-bit move. Don't split if not AVX512VL,
2942 input is xmm16+ reg and output is a mem.
2943 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2944 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2945 xmm16+ reg and output is a mem.
2947 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2949 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2950 also for flag_dwarf2_cfi_asm.
2952 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2954 PR rtl-optimization/85342
2955 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2956 a bool scalar var inside of the loop instead. Don't try to update
2957 recog_data.operand after failed apply_change_group.
2959 2018-04-12 Tom de Vries <tom@codesourcery.com>
2962 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2963 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2964 array with flexible array member as array without given dimension.
2965 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2966 argument for undefined param to true.
2968 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2971 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2972 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2973 from PowerPC section.
2974 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2975 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2976 help text that is too long.
2977 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2978 help text that is too long.
2979 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2980 help text that is too long.
2982 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2984 * config/alpha/alpha.md (stack_probe_internal): Rename
2985 from "probe_stack". Update all callers.
2987 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2989 PR rtl-optimization/84566
2990 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2991 sched_macro_fuse_insns.
2993 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2996 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2997 (compute_block_dependences): ... from here.
2999 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3001 PR tree-optimization/85331
3002 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3003 from int to HOST_WIDE_INT.
3005 2018-04-11 Martin Jambor <mjambor@suse.cz>
3008 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3009 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3010 not the same as the source val.
3011 (cgraph_edge_brings_value_p): New parameter.
3012 (gather_edges_for_value): Pass destination value to
3013 cgraph_edge_brings_value_p.
3014 (perhaps_add_new_callers): Likewise.
3015 (get_info_about_necessary_edges): Likewise and exclude values brought
3016 only by self-recursive edges.
3017 (create_specialized_node): Redirect only clones of self-calling edges.
3018 (+self_recursive_pass_through_p): New function.
3019 (find_more_scalar_values_for_callers_subset): Use it.
3020 (find_aggregate_values_for_callers_subset): Likewise.
3021 (known_aggs_to_agg_replacement_list): Removed.
3022 (decide_whether_version_node): Re-calculate known constants for all
3023 remaining context clones.
3025 2018-04-11 Richard Biener <rguenther@suse.de>
3028 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3029 from early DWARF output.
3030 (dwarf2out_early_finish): Output line info unconditionally into
3031 early DWARF and add reference to it.
3033 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3036 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3037 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3038 other than V2DFmode using iptr mode attribute.
3039 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3041 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3043 PR rtl-optimization/84659
3044 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3046 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3049 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3051 (output_loc_list): Pass address of a dummy size variable even in the
3052 locview handling loop.
3053 (index_location_lists): Add comment on why skip_loc_list_entry can't
3056 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3059 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3062 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3065 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3066 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3067 and -mstring-compare-inline-limit.
3069 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3072 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3073 for stack clash protection in a register whenever we need it to be in
3076 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3078 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3079 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3081 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3084 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3086 (mlong-double-): Ditto.
3087 * config/rs6000/sysv4.opt (msdata=): Ditto.
3088 (mtls-size=): Ditto.
3090 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3092 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3093 erroneous entries for
3094 "vector int vec_ldl (int, long int *)", and
3095 "vector unsigned int vec_ldl (int, unsigned long int *)".
3096 Add comments and entries for
3097 "vector bool char vec_ldl (int, bool char *)",
3098 "vector bool short vec_ldl (int, bool short *)",
3099 "vector bool int vec_ldl (int, bool int *)",
3100 "vector bool long long vec_ldl (int, bool long long *)",
3101 "vector pixel vec_ldl (int, pixel *)",
3102 "vector long long vec_ldl (int, long long *)",
3103 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3104 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3105 type tree bool_long_long_type_node and correct definition of
3106 bool_V2DI_type_node to make reference to this new type tree.
3107 (rs6000_mangle_type): Replace erroneous reference to
3108 bool_long_type_node with bool_long_long_type_node.
3109 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3110 comments to emphasize sign distinctions for char and int types and
3111 replace RS6000_BTI_bool_long constant with
3112 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3113 use of RS6000_BTI_pixel.
3114 (bool_long_type_node): Remove this macro definition.
3115 (bool_long_long_type_node): New macro definition
3117 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3119 PR rtl-optimization/85300
3120 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3121 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3122 simplify_unary_operation fails.
3124 2018-04-10 Martin Liska <mliska@suse.cz>
3126 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3127 cgraph_edge and ipa_ref.
3129 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3133 * config/i386/sse.md
3134 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3135 computation of the VEC_MERGE selector from mask.
3136 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3137 Fix decoding of the VEC_MERGE selector into mask.
3139 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3141 PR tree-optimization/85286
3142 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3144 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3146 * final.c (final_1): Set insn_last_address as well as
3147 insn_current_address.
3149 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3152 * explow.c (emit_stack_probe): Call validize_mem on memory location
3153 before passing it to gen_probe_stack. Create address operand and
3154 legitimize it for the probe_stack_address case.
3156 2018-04-09 Jan Hubicka <jh@suse.cz>
3159 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3160 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3161 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3163 (free_lang_data): Rebuild type inheritance graph.
3165 2018-04-09 Martin Sebor <msebor@redhat.com>
3167 * invoke.texi (-finline-small-functions): Mention other optimization
3169 (-findirect-inlining, -fpartial-inlining): Same.
3170 (-finline-functions-called-once): Same.
3171 (-freorder-blocks-and-partition): Same.
3173 2018-04-09 Jan Hubicka <jh@suse.cz>
3176 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3177 jumps; choose last target that matches the criteria (i.e.
3178 no partition changes for non-crossing jumps).
3179 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3180 support for redirecting crossing jumps to non-crossing.
3182 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3184 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3185 also for naked functions.
3187 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3189 * config/arc/arc.md (add_shift): New pattern.
3190 (add_shift2): Likewise.
3191 (sub_shift): Likewise.
3192 (sub_shift_cmp0_noout): Likewise.
3193 (compare_si_ashiftsi): Likewise.
3194 (xbfu_cmp0_noout): New combine pattern.
3195 (xbfu_cmp0"): Likewise.
3196 (movsi_set_cc_insn): Place the predicable variant first.
3197 (commutative_binary_cmp0_noout): Remove clobber.
3198 (commutative_binary_cmp0): New pattern.
3199 (noncommutative_binary_cmp0): Likewise.
3200 (noncommutative_binary_cmp0_noout): Likewise.
3201 (noncommutative_binary_comparison_result_used): Removed.
3202 (rsub_cmp0): New pattern.
3203 (rsub_cmp0_noout): Likewise.
3204 (extzvsi): Changed, keep only meaningful variants.
3205 (SQH, SEZ): New iterators.
3206 (SQH_postfix): New mode attribute.
3207 (SEZ_prefix): New code attribute.
3208 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3209 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3210 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3212 (noncommutative_operator): Check the availability of barrel
3215 2018-04-09 Richard Biener <rguenther@suse.de>
3217 PR tree-optimization/85284
3218 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3219 Only use the niter constraining form of simple_iv when the exit
3222 2018-04-09 Tom de Vries <tom@codesourcery.com>
3225 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3226 (define_expand "*memory_barrier"): New define_expand.
3227 (define_insn "memory_barrier"): New insn.
3229 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3231 PR rtl-optimization/80463
3232 PR rtl-optimization/83972
3233 PR rtl-optimization/83480
3235 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3236 correct producer for the insn.
3237 (tidy_control_flow): Fixup seqnos in case of debug insns.
3239 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3241 PR rtl-optimization/83913
3243 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3244 different sched-times when merging exprs.
3246 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3248 PR rtl-optimization/83962
3250 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3251 tidy_fallthru_edge and tidy_control_flow.
3253 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3255 PR rtl-optimization/83530
3257 * sel-sched.c (force_next_insn): New global variable.
3258 (remove_insn_for_debug): When force_next_insn is true, also leave only
3259 next insn in the ready list.
3260 (sel_sched_region): When the region wasn't scheduled, make another pass
3261 over it with force_next_insn set to 1.
3263 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3265 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3267 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3268 for interrupt control.
3269 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3270 functions for interrupt control.
3271 * config/nds32/nds32-intrinsic.md: Likewise.
3272 * config/nds32/nds32_intrinsic.h: Likewise.
3273 * config/nds32/nds32.h (nds32_builtins): Likewise.
3275 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3277 * config/nds32/nds32.c (nds32_init_machine_status,
3278 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3279 strict_aligned_p field.
3280 (nds32_expand_to_rtl_hook): New function.
3281 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3282 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3284 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3285 Chung-Ju Wu <jasonwucj@gmail.com>
3287 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3288 * config/nds32/nds32-n7.md: New file.
3289 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3290 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3292 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3293 * config/nds32/nds32.md (pipeline_model): Add n7.
3294 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3295 * config/nds32/pipelines.md: Include n7 settings.
3297 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3298 Chung-Ju Wu <jasonwucj@gmail.com>
3300 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3301 * config/nds32/nds32-e8.md: New file.
3302 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3303 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3305 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3306 * config/nds32/nds32.md (pipeline_model): Add e8.
3307 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3308 * config/nds32/pipelines.md: Include e8 settings.
3310 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3311 Chung-Ju Wu <jasonwucj@gmail.com>
3313 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3314 * config/nds32/nds32-n8.md: New file.
3315 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3316 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3318 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3319 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3320 * config/nds32/nds32.md (pipeline_model): Add n8.
3321 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3322 * config/nds32/pipelines.md: Include n8 settings.
3324 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3325 Chung-Ju Wu <jasonwucj@gmail.com>
3327 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3328 * config/nds32/nds32-n9-2r1w.md: New file.
3329 * config/nds32/nds32-n9-3r2w.md: New file.
3330 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3331 nds32_register_ports): New or modify for cpu n9.
3332 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3334 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3335 * config/nds32/nds32-utils.c: New file.
3336 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3337 TARGET_MUL_SLOW): Define.
3338 * config/nds32/nds32.md (pipeline_model): New attribute.
3339 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3340 New options that support cpu n9.
3341 * config/nds32/pipelines.md: Include n9 settings.
3342 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3344 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3346 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3347 information if necessary.
3348 (output_cond_branch_compare_zero): Likewise.
3349 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3350 (nds32_target_alignment): Refine for alignment.
3351 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3352 (FUNCTION_BOUNDARY): Modify.
3353 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3355 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3357 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3359 * config/nds32/constants.md (unspec_volatile_element): Add values for
3360 TLB operation and data prefetch.
3361 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3362 functions for TLB operation and data prefetch.
3363 * config/nds32/nds32-intrinsic.md: Likewise.
3364 * config/nds32/nds32_intrinsic.h: Likewise.
3365 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3366 (nds32_print_operand): Likewise.
3367 * config/nds32/nds32.h (nds32_builtins): Likewise.
3369 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3370 Andrew Pinski <pinsika@gcc.gnu.org>
3373 * match.pd: Use constant_boolean_node of correct type instead of
3374 boolean_true_node or boolean_false_node for simplifying
3375 pointer comparisons to zero.
3377 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3379 PR tree-optimization/80021
3380 * tree.c (verify_type_variant): Make error call in verify_variant_match
3381 translatable and remove final full stop.
3383 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3385 * config/nds32/constants.md (unspec_volatile_element): Add
3386 UNSPEC_VOLATILE_EH_RETURN.
3387 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3388 nds32_output_stack_pop): Support dwarf exception handling process.
3389 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3390 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3391 exception handling process.
3392 (nds32_compute_stack_frame): Likewise.
3393 (nds32_return_addr_rtx): Likewise.
3394 (nds32_initial_elimination_offset): Likewise.
3395 (nds32_expand_prologue): Likewise.
3396 (nds32_expand_epilogue): Likewise.
3397 (nds32_dynamic_chain_address): New function.
3398 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3400 (DYNAMIC_CHAIN_ADDRESS): Define.
3401 (EH_RETURN_DATA_REGNO): Define.
3402 (EH_RETURN_STACKADJ_RTX): Define.
3403 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3404 patterns for dwarf exception handling.
3406 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3408 * config/nds32/nds32.h: Clean up obsolete macros.
3410 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3412 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3413 Add enum values for particular instructions.
3414 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3415 particular intrinsic functions.
3416 * config/nds32/nds32-intrinsic.md: Likewise.
3417 * config/nds32/nds32_intrinsic.h: Likewise.
3418 * config/nds32/nds32.h (nds32_builtins): Likewise.
3419 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3420 (btst, ave): New patterns for particular instructions.
3422 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3424 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3425 Add enum values for atomic load/store and memory sync.
3426 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3428 * config/nds32/nds32-intrinsic.md: Likewise.
3429 * config/nds32/nds32_intrinsic.h: Likewise.
3430 * config/nds32/nds32.h (nds32_builtins): Likewise.
3432 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3434 PR tree-optimization/85257
3435 * fold-const.c (native_encode_vector): If not all elts could fit
3436 and off is -1, return 0 rather than offset.
3437 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3438 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3439 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3440 adjust buffer in native_interpret_expr call.
3442 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3444 * config/nds32/constants.md (unspec_volatile_element): Add cache
3445 control enum values.
3446 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3447 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3448 * config/nds32/nds32.c (nds32_cctl_names): New.
3449 (nds32_print_operand): Handle cache control register names.
3450 * config/nds32/nds32.h (nds32_builtins): New enum values.
3451 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3453 * config/nds32/nds32.md (type): Add mmu.
3454 * config/nds32/pipelines.md (simple_insn): Add mmu.
3456 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3458 * config/nds32/nds32.md (type): Remove call.
3459 * config/nds32/pipelines.md (simple_insn): Likewise.
3461 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3463 * config/nds32/constants.md (unspec_volatile_element): Add
3464 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3465 UNSPEC_VOLATILE_FMFCFG.
3466 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3467 description for fmfcfg and fmfcsr.
3468 (bdesc_1arg): Add fmtcsr.
3469 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3470 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3471 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3472 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3473 unspec_fmfcfg): New patterns.
3474 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3475 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3476 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3477 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3478 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3479 __nds32__fmfcfg): Define.
3481 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3483 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3484 intrinsic register names.
3485 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3486 intrinsic register enum values and macros.
3488 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3490 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3491 for load/store addressing form.
3492 (nds32_print_operand_address): Likewise.
3494 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3497 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3498 based on LABEL_REF. Remove useless assertion.
3499 (pic_address_needs_scratch): Fix formatting.
3500 (sparc_legitimize_pic_address): Minor tweaks.
3501 (sparc_delegitimize_address): Adjust assertion accordingly.
3502 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3503 into symbolic_operand.
3504 (movsi_high_pic_label_ref): Likewise.
3505 (movsi_lo_sum_pic_label_ref): Likewise.
3506 (movdi_pic_label_ref): Likewise.
3507 (movdi_high_pic_label_ref): Likewise.
3508 (movdi_lo_sum_pic_label_ref): Likewise.
3510 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3512 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3513 custom LIB_SPEC setup.
3515 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3516 Kito Cheng <kito.cheng@gmail.com>
3518 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3519 * config/riscv/freebsd.h: New.
3521 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3523 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3524 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3527 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3528 Kito Cheng <kito.cheng@gmail.com>
3530 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3531 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3532 * config/nds32/nds32-protos.h (nds32_output_call,
3533 nds32_output_return): Declare.
3534 * config/nds32/nds32.md: Refine all the call and return patterns.
3536 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3539 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3540 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3542 PR rtl-optimization/84872
3543 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3544 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3547 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3549 * expr.c (copy_blkmode_to_reg): Revert 254862.
3550 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3552 2018-04-06 Richard Biener <rguenther@suse.de>
3555 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3556 after seeing a component reference with an adjacent field. Treat
3557 refs to arrays at struct end of external decls similar to
3558 refs to unconstrained commons.
3560 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3563 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3564 look through SAVE_EXPRs with non-side-effects argument. Adjust
3566 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3569 2018-04-06 Richard Biener <rguenther@suse.de>
3572 * alias.c (find_base_term): New wrapper around find_base_term
3573 unwinding CSELIB_VAL_PTR changes.
3574 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3577 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3579 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3581 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3582 constant definitions.
3583 ("nop"): lr 0,0 -> nopr r0
3584 ("nop_lr0", "nop_lr1"): New insn definitions.
3586 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3588 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3589 NDS32_V3PUSH_AVAILABLE_P macro.
3591 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3592 Chung-Ju Wu <jasonwucj@gmail.com>
3594 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3595 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3596 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3597 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3598 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3599 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3600 * config/nds32/constraints.md: New constraints and checking for hard
3601 float configuration.
3602 * config/nds32/iterators.md: New mode iterator and attribute for hard
3603 float configuration.
3604 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3606 * config/nds32/nds32-fpu.md: New file.
3607 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3608 deal with hard float code generation.
3609 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3611 (abi_type, float_reg_number): New enum type.
3612 * config/nds32/nds32-predicates.c: New predicates for hard float.
3613 * config/nds32/nds32-protos.h: Declare functions for hard float.
3614 * config/nds32/nds32.c: Implementation for hard float configuration.
3615 * config/nds32/nds32.h: Definitions for hard float configuration.
3616 * config/nds32/nds32.md: Include hard float machine description and
3617 modify patterns for hard float configuration.
3618 * config/nds32/nds32.opt: New options for hard float configuration.
3619 * config/nds32/predicates.md: New predicates for hard float
3622 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3624 * common/config/nds32/nds32-common.c
3625 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3627 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3630 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3631 CONSTRUCTOR_ELT (ctor, ...)->value.
3633 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3636 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3638 2018-04-05 Tom de Vries <tom@codesourcery.com>
3641 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3644 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3645 Kito Cheng <kito.cheng@gmail.com>
3647 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3648 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3649 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3650 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3652 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3653 Kito Cheng <kito.cheng@gmail.com>
3655 * config/nds32/constraints.md (Ufe): New memory constraint.
3656 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3657 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3658 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3660 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3661 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3663 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3665 * config/nds32/nds32.md: Use optimize_size in the condition for
3666 alu-shift instructions.
3668 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3670 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3672 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3674 * config/nds32/nds32.md (negsi2): Refine pattern.
3676 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3677 Chung-Ju Wu <jasonwucj@gmail.com>
3679 * config/nds32/iterators.md (shift_rotate): New code iterator.
3680 (shift): New code attribute.
3681 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3682 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3683 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3684 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3685 bit-wise operations.
3686 (andsi3, *andsi3): Ditto.
3687 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3688 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3689 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3690 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3691 nds32_ior_operand, nds32_xor_operand): New predicates.
3693 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3695 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3696 (addsi3, subsi3): ... this.
3698 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3700 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3702 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3704 * config/nds32/nds32.md: Adjust indention.
3706 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3708 * config/nds32/nds32.md (feature): New attribute.
3710 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3712 * config/nds32/nds32.md (subtype): New attribute.
3714 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3717 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3718 expansion to perform a bitwise AND of the argument followed by a
3719 boolean negation of the result.
3721 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3723 PR rtl-optimization/84878
3724 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3725 the basic block. Assert the use reference is not artificial and that
3726 it has an associated insn.
3728 2018-04-04 Michael Matz <matz@suse.de>
3730 * builtins.c (compute_objsize): Pass correct operand
3731 to array_at_struct_end_p.
3733 2018-04-04 Richard Biener <rguenther@suse.de>
3736 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3737 from contexts for DINFO_LEVEL_TERSE and below.
3739 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3741 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3742 resiter_operand condition.
3743 * config/nds32/nds32.md (*move<mode>): Ditto.
3745 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3746 Monk Chiang <sh.chiang04@gmail.com>
3748 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3750 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3752 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3754 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3755 Kito Cheng <kito.cheng@gmail.com>
3757 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3758 nds32_cond_code_str, output_cond_branch,
3759 output_cond_branch_compare_zero, nds32_expand_cbranch,
3760 nds32_expand_cstore, nds32_expand_movcc,
3761 nds32_output_cbranchsi4_equality_zero,
3762 nds32_output_cbranchsi4_equality_reg,
3763 nds32_output_cbranchsi4_equality_reg_or_const_int,
3764 nds32_output_cbranchsi4_greater_less_zero: New functions.
3765 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3766 nds32_expand_cstore, nds32_expand_movcc,
3767 nds32_output_cbranchsi4_equality_zero,
3768 nds32_output_cbranchsi4_equality_reg,
3769 nds32_output_cbranchsi4_equality_reg_or_const_int,
3770 nds32_output_cbranchsi4_greater_less_zero): Declare.
3771 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3772 nds32_rimm11s_operand): New predicates.
3773 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3774 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3777 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3779 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3780 * config/nds32/nds32.md: Ditto.
3781 * config/nds32/pipelines.md: Ditto.
3783 2018-04-04 Richard Biener <rguenther@suse.de>
3785 PR tree-optimization/85168
3786 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3787 propagating abnormals.
3789 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3791 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3793 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3794 Kito Cheng <kito.cheng@gmail.com>
3796 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3797 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3798 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3799 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3800 * config/nds32/nds32.md (sibcall_internal): New.
3801 (sibcall_register): Remove.
3802 (sibcall_immediate): Remove.
3803 (sibcall_value_internal): New.
3804 (sibcall_value_register): Remove.
3805 (sibcall_value_immediate): Remove.
3806 * config/nds32/predicates.md (nds32_general_register_operand): New.
3807 (nds32_call_address_operand): New.
3809 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3811 PR rtl-optimization/85167
3812 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3813 bb_defs if *split_p, instead preinitialize it to NULL.
3815 PR tree-optimization/85156
3816 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3817 evaluating the argument multiple times.
3819 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3821 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3823 (_mm_cvtpd_ps): Likewise.
3824 (_mm_cvttpd_epi32): Likewise.
3825 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3826 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3827 vector, pixel, and bool following altivec.h include.
3829 2018-04-03 Martin Sebor <msebor@redhat.com>
3831 * doc/extend.texi (Common Function Attributes): Clarify.
3832 (const attribute): Likewise.
3833 (pure attribute): Likewise.
3835 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3838 * config/i386/i386.c (ix86_expand_vector_set): Use
3839 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3841 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3843 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3844 instructions when changing rounding bits to preserve precision bits
3845 in the x87 control word.
3847 2018-04-03 Martin Liska <mliska@suse.cz>
3849 PR tree-optimization/82491
3850 * rtl.h (strip_offset_and_add): Replace += suboffset with
3851 poly_uint64 () + suboffset.
3853 2018-03-29 Martin Liska <mliska@suse.cz>
3854 Martin Jambor <mjambor@suse.cz>
3857 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3858 param_type is not an integral or pointer type.
3860 2018-04-03 Richard Biener <rguenther@suse.de>
3862 * sese.h (recompute_all_dominators): Remove.
3864 2018-04-02 Martin Sebor <msebor@redhat.com>
3866 * doc/invoke.texi (-Wrestrict): Fix typos.
3868 2018-04-02 Jim Wilson <jimw@sifive.com>
3870 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3871 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3872 (<optab>di3, <optab>si3_extend): Likewise.
3873 (<optab>si3_mask, <optab>si3_mask_1): New.
3874 (<optab>di3_mask, <optab>di3_mask_1): New.
3875 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3876 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3877 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3879 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3881 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3884 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3886 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3887 (nds32_canonicalize_comparison): New function.
3889 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3890 Kito Cheng <kito.cheng@gmail.com>
3891 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3893 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3894 * config/nds32/constants.md (unspec_volatile_element): Add
3895 UNSPEC_VOLATILE_RELAX_GROUP.
3896 * config/nds32/nds32-relax-opt.c: New file.
3897 * config/nds32/nds32-predicates.c
3898 (nds32_symbol_load_store_p): New function.
3899 * config/nds32/nds32-protos.h
3900 (nds32_symbol_load_store_p): Declare function.
3901 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3902 * config/nds32/nds32.c
3903 (nds32_register_pass): New function to register pass.
3904 (nds32_register_passes): New function to register passes.
3905 * config/nds32/nds32.md (relax_group): New pattern.
3906 * config/nds32/nds32.opt (mrelax-hint): New option.
3907 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3909 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3911 * config/nds32/t-nds32: Modify files dependency.
3913 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3915 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3916 (PROFILE_HOOK): Define its implementation.
3918 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3920 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3921 type and 32-bit size.
3923 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3926 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3927 (V_128_256): New mode iterator.
3928 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3929 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3930 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3932 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3933 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3935 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3938 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3939 NaN inputs correctly.
3941 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3944 * config/rs6000/vsx.md (??r): New mode attribute.
3945 (*vsx_mov<mode>_64bit): Use it.
3946 (*vsx_mov<mode>_32bit): Likewise.
3948 2018-03-30 Martin Sebor <msebor@redhat.com>
3950 PR tree-optimization/84818
3951 * builtins.c (check_access): Use warning_n.
3953 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3956 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3958 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3961 2018-03-30 Julia Koval <julia.koval@intel.com>
3964 * x86-tune.def (movx, partial_reg_dependency): Enable for
3967 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3970 * lra-constraints.c (process_alt_operands): Move setting
3971 this_alternative_matches below.
3973 2018-03-29 Martin Liska <mliska@suse.cz>
3976 * doc/invoke.texi: Document how LTO works with debug info.
3977 Describe auto-load support of binutils. Mention 'x86-64'
3978 as valid option value of -march option.
3980 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3982 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3985 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3986 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3987 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3990 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3993 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3994 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3995 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3996 for RS6000_BTM_POWERPC64.
3997 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3998 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3999 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4004 2018-03-28 Carl Love <cel@us.ibm.com>
4007 2017-09-27 Carl Love <cel@us.ibm.com>
4009 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4010 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4011 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4014 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4016 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4017 instead of __vector bool.
4018 (_mm_max_pu8): Likewise.
4019 (_mm_min_pi16): Likewise.
4021 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4024 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4025 (DIVWEUO): Likewise.
4027 (DIVDEUO): Likewise.
4028 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4029 DIVWEUO and DIVDEUO.
4030 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4031 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4032 (div_extend): Likewise.
4033 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4035 (__builtin_divweuo): Likewise.
4036 (__builtin_divdeo): Likewise.
4037 (__builtin_divdeuo): Likewise.
4039 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4042 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4043 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4045 PR tree-optimization/82004
4046 * gimple-match-head.c (optimize_pow_to_exp): New function.
4047 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4048 Don't fold to exp if optimize_pow_to_exp is false.
4050 2018-03-28 Martin Liska <mliska@suse.cz>
4053 * calls.c (initialize_argument_information): Fix trailing space.
4054 * common.opt: Fix typo and provide better explanation for
4055 -fsanitize-coverage option.
4056 * config/i386/i386.opt: Fix typo.
4058 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4059 Martin Liska <mliska@suse.cz>
4062 * gimplify.c (asan_poison_variable): Don't do the check for
4063 gimplify_omp_ctxp here.
4064 (gimplify_decl_expr): Do it here.
4065 (gimplify_target_expr): Likewise.
4067 2018-03-28 Martin Liska <mliska@suse.cz>
4070 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4071 chkp_type_bounds_count if MPX is not enabled.
4073 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4075 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4077 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4080 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4081 function to create the function decl for complex long double
4082 multiply and divide for -mabi=ieeelongdouble.
4083 (init_float128_ieee): Call it.
4085 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4088 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4089 -fcf-protection=branch -mibt.
4090 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4092 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4095 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4097 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4100 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4101 extern array declarations.
4103 2018-03-27 Richard Biener <rguenther@suse.de>
4106 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4107 explicit single_use checks.
4109 2018-03-27 Richard Biener <rguenther@suse.de>
4111 PR tree-optimization/85082
4112 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4115 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4117 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4118 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4119 Turn on fasynchronous-unwind-tables and funwind-tables.
4121 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4124 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4125 (*bmi_blsr_<mode>_ccz): Ditto.
4127 2018-03-26 Tom de Vries <tom@codesourcery.com>
4129 PR tree-optimization/85063
4130 * omp-general.c (offloading_function_p): New function. Factor out
4132 * omp-offload.c (pass_omp_target_link::gate): ... here.
4133 * omp-general.h (offloading_function_p): Declare.
4134 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4135 with attribute omp declare target for offloading functions.
4137 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4139 PR tree-optimization/84005
4140 * tree-data-ref.h (get_base_for_alignment): Declare.
4141 * tree-data-ref.c (get_base_for_alignment_1): New function.
4142 (get_base_for_alignment): Likewise.
4143 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4144 get_base_for_alignment to find a suitable base object, instead
4145 of always using drb->base_address.
4147 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4150 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4151 known size by default.
4153 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4156 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4157 and non BLKmode operands.
4159 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4162 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4163 Clean up attributes.
4165 2018-03-23 Richard Biener <rguenther@suse.de>
4168 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4169 we are going to emit early debug for LTO.
4171 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4174 * function.c (match_asm_constraints_1): Don't optimize if input
4175 doesn't satisfy general_operand predicate for output's mode.
4178 * alias.c (write_dependence_p): Don't require for x_canonicalized
4179 non-VOIDmode if x has VOIDmode.
4182 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4183 just don't try to optimize it rather than assert it never happens.
4185 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4187 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4188 macro expansions for definition of ST_INTERNAL_<mode> and
4189 LD_INTERNAL_<mode> builtins.
4190 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4192 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4194 (altivec_expand_st_builtin): Likewise.
4195 (altivec_expand_builtin): Remove calls to deleted functions.
4196 (rs6000_address_for_altivec): Delete this function.
4197 * config/rs6000/vector.md: Remove expands for
4198 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4200 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4203 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4204 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4205 re-computing once computed.
4206 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4207 (arm_init_machine_status): Initialize
4208 machine->static_chain_stack_bytes.
4210 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4213 * doc/extend.texi: Add four new prototypes for vec_ld.
4214 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4215 definitions for more logical presentation.
4216 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4217 entries for V1TI variants of __builtin_altivec_ld builtin.
4218 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4219 handling of V1TI variant of LVX icode pattern.
4220 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4221 (rs6000_gimple_fold_builtin): Likewise.
4222 (altivec_init_builtins): Add code to define
4223 __builtin_altivec_lvx_v1ti function.
4225 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4228 * function.c (match_asm_constraints_1): Don't do the optimization
4229 if input isn't a REG, SUBREG, MEM or constant.
4231 2018-03-22 Tom de Vries <tom@codesourcery.com>
4233 PR tree-optimization/84956
4234 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4235 bb_has_abnormal_pred.
4237 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4240 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4241 DECL_INITIAL (decl) to decl at the end.
4242 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4245 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4247 * doc/extend.texi (__builtin_tgmath): Document when complex
4248 integer types are treated as _Complex _Float64.
4250 2018-03-21 Tom de Vries <tom@codesourcery.com>
4252 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4254 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4256 PR tree-optimization/84960
4257 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4258 if it is ENTRY block, move them into single succ of ENTRY in that case.
4260 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4262 PR tree-optimization/84811
4263 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4264 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4265 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4266 (known_subrange_p): Update accordingly. Cast each value involved
4267 in the size comparison, rather than casting the result of the
4270 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4272 PR tree-optimization/84982
4273 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4274 by flipping the least significant bit rather than all bits from
4275 bitpos to bitpos + bitsize - 1.
4277 2018-03-21 Nathan Sidwell <nathan@acm.org>
4279 * doc/extend.texi (Deprecated Features): Remove mention of
4280 long-deleted deprecations.
4282 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4285 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4286 * configure: Regenerate.
4288 2018-03-21 Tom de Vries <tom@codesourcery.com>
4290 PR tree-optimization/83126
4291 * tree-parloops.c (num_phis): New function.
4292 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4294 2018-03-21 Nathan Sidwell <nathan@acm.org>
4296 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4297 mention anon-struct/union members and trailing attributes.
4299 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4301 PR tree-optimization/84969
4302 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4303 builtin memset partitions if they set different rhs values.
4305 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4307 PR rtl-optimization/84989
4308 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4309 VEC_DUPLICATE with scalar result mode.
4311 2018-03-21 Martin Liska <mliska@suse.cz>
4314 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4315 not intended return statement.
4317 2018-03-21 Martin Liska <mliska@suse.cz>
4320 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4321 (chkp_find_bound_slots_1): Limit number of iterations.
4323 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4326 * Minor grammar fixes for x86 options.
4328 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4331 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4332 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4334 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4337 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4338 (altivec_lvx_<mode>_1op): Likewise.
4339 (altivec_stvx_<mode>_2op): Likewise.
4340 (altivec_stvx_<mode>_1op): Likewise.
4341 (altivec_lvx_<VM2:mode>): New define_expand.
4342 (altivec_stvx_<VM2:mode>): Likewise.
4343 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4344 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4345 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4346 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4347 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4348 (rs6000_gen_lvx): Likewise.
4349 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4350 (altivec_expand_stv_builtin): Likewise.
4351 (altivec_expand_builtin): Likewise.
4352 * config/rs6000/vector.md: Likewise.
4354 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4357 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4360 2018-03-20 Richard Biener <rguenther@suse.de>
4363 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4364 sign-conversions as zero, fall back to standard scalar_stmt
4367 2018-03-20 Martin Liska <mliska@suse.cz>
4370 * predict.c (rebuild_frequencies): Handle case when we have
4371 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4373 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4376 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4377 flag_section_anchors.
4378 * varasm.c (use_blocks_for_decl_p): Remove hack for
4379 dw2_force_const_mem.
4382 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4384 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4385 be created, use lowpart_subreg of operands[0] rather than operands[0]
4387 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4388 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4389 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4390 and n constraint instead of aarch64_shift_imm_di and Usd.
4391 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4392 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4394 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4397 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4398 to favor GPR over NEON registers.
4399 (<shift>di3_neon): Likewise.
4401 2018-03-20 Tom de Vries <tom@codesourcery.com>
4404 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4405 (nvptx_process_pars): Emit bar.sync asap and alap.
4407 2018-03-20 Tom de Vries <tom@codesourcery.com>
4410 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4411 seen_label if seen_label is already set.
4413 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4416 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4417 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4418 Use 1U instead of 1. Formatting fixes.
4421 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4422 instead of TREE_TYPE (s1) for the return value.
4424 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4426 PR tree-optimization/84946
4427 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4428 bitsize + bitsize in poly_uint64 rather than poly_int64.
4431 * dwarf2asm.c: Include fold-const.c.
4432 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4433 of decl rather than decl itself.
4435 PR rtl-optimization/84643
4436 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4438 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4441 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4442 calling assemble_variable.
4444 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4447 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4448 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4450 2018-03-19 Jim Wilson <jimw@sifive.com>
4453 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4454 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4455 (riscv_first_stack_step): Likewise.
4456 (riscv_option_override): Use STACK_BOUNDARY instead of
4458 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4460 (BIGGEST_ALIGNMENT): Set to 128.
4461 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4462 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4465 2018-03-19 Richard Biener <rguenther@suse.de>
4467 PR tree-optimization/84933
4468 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4469 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4471 2018-03-19 Richard Biener <rguenther@suse.de>
4473 PR tree-optimization/84859
4474 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4475 (cond_if_else_store_replacement): Perform sinking operation on
4476 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4477 Generalize what a BB with a single eligible store is.
4479 2018-03-19 Richard Biener <rguenther@suse.de>
4481 PR tree-optimization/84929
4482 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4483 chrec_is_positive against non-chrec arg.
4485 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4488 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4490 2018-03-18 Martin Liska <mliska@suse.cz>
4492 PR rtl-optimization/84635
4493 * regrename.c (build_def_use): Use matches_mode only when
4496 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4498 PR tree-optimization/84913
4499 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4500 vectorize chains of COND_EXPRs.
4502 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4504 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4506 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4508 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4510 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4512 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4514 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4515 Kito Cheng <kito.cheng@gmail.com>
4517 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4518 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4519 (nds32_adjust_reg_alloc_order): New function.
4520 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4522 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4524 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4525 nds32_print_operand, nds32_print_operand_address): Use
4526 HOST_WIDE_INT_PRINT_DEC instead.
4528 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4530 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4532 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4535 * config/i386/i386.c (initial_ix86_tune_features,
4536 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4538 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4539 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4540 rather than 1u << ix86_tune. Formatting fix.
4541 (ix86_option_override_internal): Change ix86_arch_mask from
4542 unsigned int to unsigned HOST_WIDE_INT, initialize to
4543 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4544 (ix86_function_specific_restore): Likewise.
4546 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4549 * postreload.c (reload_combine_recognize_pattern): Perform
4550 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4551 truncate_int_for_mode the result for the destination's mode.
4554 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4555 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4557 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4560 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4563 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4564 regno_allocno_class_array and sorted_pseudos.
4565 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4566 insns where regno is used.
4568 2018-03-16 Martin Liska <mliska@suse.cz>
4571 * multiple_target.c (create_dispatcher_calls): Redirect
4572 reference in the symbol table.
4574 2018-03-16 Martin Liska <mliska@suse.cz>
4577 * multiple_target.c (create_dispatcher_calls): Redirect also
4580 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4584 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4585 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4586 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4588 2018-03-16 Julia Koval <julia.koval@intel.com>
4590 * doc/invoke.texi (Skylake Server): Add CLWB.
4591 Cannonlake): Remove CLWB.
4593 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4595 PR tree-optimization/84841
4596 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4598 (FLOAT_ONE_CONST_TYPE): Define.
4599 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4600 (sort_by_operand_rank): Put entries with higher constant_type last
4601 rather than first to match comments.
4603 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4605 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4608 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4611 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4612 check and use address of target always.
4614 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4617 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4618 (indirect_thunk_bnd_needed): Likewise.
4619 (indirect_thunks_used): Likewise.
4620 (indirect_thunks_bnd_used): Likewise.
4621 (indirect_return_needed): New.
4622 (indirect_return_bnd_needed): Likewise.
4623 (output_indirect_thunk_function): Add a bool argument for
4625 (output_indirect_thunk_function): Don't generate alias for
4626 function return thunk.
4627 (ix86_code_end): Call output_indirect_thunk_function to generate
4628 function return thunks.
4629 (ix86_output_function_return): Set indirect_return_bnd_needed
4630 and indirect_return_needed instead of indirect_thunk_bnd_needed
4631 and indirect_thunk_needed.
4633 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4635 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4636 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4637 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4639 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4640 Paul Hua <paul.hua.gm@gmail.com>
4643 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4645 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4647 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4648 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4651 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4654 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4655 instead of GET_MODE_SIZE when comparing Units.
4657 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4660 * varasm.c (hash_section): Return an unchangeble hash value
4661 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4662 Return !aarch64_can_use_per_function_literal_pools_p ().
4664 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4667 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4668 rather than address of cmode as last argument to prepare_cmp_insn.
4670 2018-03-15 Julia Koval <julia.koval@intel.com>
4672 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4673 F_AVX512VNNI, F_AVX512BITALG): New.
4675 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4678 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4679 insn for floating-point loads and stores.
4681 2018-03-14 Carl Love <cel@us.ibm.com>
4683 * config/rs6000/rs6000-c.c: Add macro definitions for
4684 ALTIVEC_BUILTIN_VEC_PERMXOR.
4685 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4686 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4687 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4688 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4690 * config/doc/extend.texi: Add prototypes for vec_permxor.
4692 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4695 * diagnostic-show-locus.c (class layout_point): Convert m_line
4696 from int to linenum_type.
4697 (line_span::comparator): Use linenum "compare" function when
4698 comparing line numbers.
4699 (test_line_span): New function.
4700 (layout_range::contains_point): Convert param "row" from int to
4702 (layout_range::intersects_line_p): Likewise.
4703 (layout::will_show_line_p): Likewise.
4704 (layout::print_source_line): Likewise.
4705 (layout::should_print_annotation_line_p): Likewise.
4706 (layout::print_annotation_line): Likewise.
4707 (layout::print_leading_fixits): Likewise.
4708 (layout::annotation_line_showed_range_p): Likewise.
4709 (struct line_corrections): Likewise for field m_row.
4710 (line_corrections::line_corrections): Likewise for param "row".
4711 (layout::print_trailing_fixits): Likewise.
4712 (layout::get_state_at_point): Likewise.
4713 (layout::get_x_bound_for_row): Likewise.
4714 (layout::print_line): Likewise.
4715 (diagnostic_show_locus): Likewise for locals "last_line" and
4717 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4718 * input.c (selftest::test_linenum_comparisons): New function.
4719 (selftest::input_c_tests): Call it.
4720 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4721 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4722 * selftest.h (ASSERT_GT): New macro.
4723 (ASSERT_GT_AT): New macro.
4724 (ASSERT_LT): New macro.
4725 (ASSERT_LT_AT): New macro.
4727 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4729 PR rtl-optimization/84780
4730 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4732 2018-03-14 Martin Liska <mliska@suse.cz>
4734 * tree.c (record_node_allocation_statistics): Use
4735 get_stats_node_kind.
4736 (get_stats_node_kind): New function extracted from
4737 record_node_allocation_statistics.
4738 (free_node): Use get_stats_node_kind.
4740 2018-03-14 Richard Biener <rguenther@suse.de>
4742 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4743 that the value-set of ANTIC_IN doesn't grow.
4746 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4748 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4749 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4751 2018-03-14 Julia Koval <julia.koval@intel.com>
4753 * config.gcc (icelake-client, icelake-server): New.
4755 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4756 (initial_ix86_arch_features): Ditto.
4757 (PTA_SKYLAKE): Add SGX.
4758 (PTA_ICELAKE): Remove.
4759 (PTA_ICELAKE_CLIENT): New.
4760 (PTA_ICELAKE_SERVER): New.
4761 (ix86_option_override_internal): Split up icelake on icelake client and
4763 (get_builtin_code_for_version): Ditto.
4764 (fold_builtin_cpu): Ditto.
4765 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4766 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4767 * config/i386/i386.h (processor_type): Ditto.
4768 * doc/invoke.texi: Ditto.
4770 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4773 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4774 INTEGER_CST offset, add it together with bitpos / 8 and
4775 sign extend based on POINTER_SIZE.
4779 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4782 * config/i386/constraints.md (Yc): New register constraint.
4783 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4784 Use Yc constraint for alternative 2 of operand 0. Remove
4785 preferred_for_speed attribute.
4787 2018-03-14 Richard Biener <rguenther@suse.de>
4789 PR tree-optimization/84830
4790 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4791 with the old one to avoid oscillations.
4793 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4796 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4798 (assign_by_spills): Return a flag of reload assignment failure.
4799 Do not process the reload assignment failures. Do not spill other
4800 reload pseudos if they has the same reg class. Update n if
4802 (lra_assign): Add a return arg. Set up from the result of
4803 assign_by_spills call.
4804 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4805 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4806 usage_insns if it is not NULL.
4807 (spill_hard_reg_in_range): New function.
4808 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4809 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4810 function prototypes.
4811 (lra_assign): Change prototype.
4812 * lra.c (lra): Add code to deal with fails by splitting hard reg
4815 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4817 * config/riscv/riscv.opt (mrelax): New option.
4818 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4819 "norelax" when riscv_mrelax is disabled.
4820 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4822 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4825 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4826 reassociation for int modes.
4828 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4830 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4831 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4833 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4834 * config/aarch64/aarch64-sve.md
4835 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4836 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4837 (*extend<mode><Vwide>2): Rename to...
4838 (aarch64_sve_extend<mode><Vwide>2): ...this.
4839 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4840 renaming the old pattern to...
4841 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4843 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4844 define_expand, renaming the old pattern to...
4845 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4846 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4847 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4848 account when deciding which SVE instruction the optab should use.
4849 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4851 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4853 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4854 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4855 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4856 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4857 (tlsdesc_small_<mode>): Turn a define_expand and use
4858 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4859 (tlsdesc_small_advsimd_<mode>): ...this.
4860 (tlsdesc_small_sve_<mode>): New pattern.
4862 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4864 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4865 (UNSPEC_UMUL_HIGHPART): New constants.
4866 (MUL_HIGHPART): New int iteraor.
4867 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4868 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4870 (*<su>mul<mode>3_highpart): New define_insn.
4872 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4875 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4878 2018-03-13 Martin Liska <mliska@suse.cz>
4881 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4883 (sem_item_optimizer::~sem_item_optimizer): Release it.
4884 (sem_item_optimizer::merge_classes): Register variable aliases.
4885 (sem_item_optimizer::fixup_pt_set): New function.
4886 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4887 * ipa-icf.h: Declare new variables and functions.
4889 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4892 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4893 integer_pow2p@2 and test integer_pow2p in condition.
4894 (A < 0 ? C : 0): Similarly for @1.
4897 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4898 characters starting at p contain '\0' character, don't look beyond
4902 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4903 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4906 * reg-stack.c (change_stack): Change update_end var from int to
4907 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4908 also call set_block_for_insn on the newly added insns and rescan.
4911 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4912 on the last operand.
4915 * tree.c (stabilize_reference_1): Return save_expr (e) for
4916 STATEMENT_LIST even if it doesn't have side-effects.
4918 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4920 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4922 2018-03-12 Renlin Li <renlin.li@arm.com>
4924 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4925 aarch64_output_scalar_simd_mov_immediate.
4927 2018-03-12 Martin Sebor <msebor@redhat.com>
4929 PR tree-optimization/83456
4930 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4931 for perfectly overlapping calls to memcpy.
4932 (gimple_fold_builtin_memory_chk): Same.
4933 (gimple_fold_builtin_strcpy): Handle no-warning.
4934 (gimple_fold_builtin_stxcpy_chk): Same.
4935 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4937 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4939 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4940 parameter. Use it for SFmode.
4941 (rs6000_function_arg_advance_1): Adjust.
4942 (rs6000_function_arg): Adjust.
4943 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4945 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4947 PR rtl-optimization/84169
4948 PR rtl-optimization/84780
4949 * combine.c (can_combine_p): Check for a 2-insn combination whether
4950 the destination register is used between the two insns, too.
4952 2018-03-12 Richard Biener <rguenther@suse.de>
4954 PR tree-optimization/84803
4955 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4956 for refs DR analysis didn't process.
4958 2018-03-12 Richard Biener <rguenther@suse.de>
4960 PR tree-optimization/84777
4961 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4962 force-vectorize loops ignore whether we are optimizing for size.
4964 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4966 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4967 (TARGET_MD_ASM_ADJUST): Define.
4969 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4970 Kito Cheng <kito.cheng@gmail.com>
4971 Chung-Ju Wu <jasonwucj@gmail.com>
4973 * config/nds32/nds32.c (nds32_compute_stack_frame,
4974 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4975 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4976 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4977 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4978 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4979 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4980 * config/nds32/nds32.md (prologue, epilogue): Use macro
4981 NDS32_V3PUSH_AVAILABLE_P to do checking.
4983 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4986 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4987 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4988 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4989 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4990 addition of most attributes on !orig_type_die or the attribute not
4991 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4993 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4994 Chung-Ju Wu <jasonwucj@gmail.com>
4996 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4998 * config/nds32/nds32.opt (mvh): New option.
5000 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5001 Chung-Ju Wu <jasonwucj@gmail.com>
5003 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5005 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5006 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5009 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5010 Chung-Ju Wu <jasonwucj@gmail.com>
5012 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5014 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5015 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5017 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5018 Kito Cheng <kito.cheng@gmail.com>
5019 Chung-Ju Wu <jasonwucj@gmail.com>
5021 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5022 UNSPEC_FFMISM and UNSPEC_FLMISM.
5023 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5024 for ffb, ffmism and flmism.
5025 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5026 (unspec_ffmism): Ditto.
5027 (unspec_flmism): Ditto.
5028 (nds32_expand_builtin_impl): Check if string extension is available.
5029 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5030 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5032 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5035 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5038 * lra-assigns.c (assign_by_spills): Return a flag of reload
5039 assignment failure. Do not process the reload assignment
5040 failures. Do not spill other reload pseudos if they has the same
5042 (lra_assign): Add a return arg. Set up from the result of
5043 assign_by_spills call.
5044 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5045 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5046 usage_insns if it is not NULL.
5047 (spill_hard_reg_in_range): New function.
5048 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5049 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5050 function prototypes.
5051 (lra_assign): Change prototype.
5052 * lra.c (lra): Add code to deal with fails by splitting hard reg
5055 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5058 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5060 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5063 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5064 (dw_val_node): Add val_symbolic_view.
5065 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5066 (symview_upper_bound): New.
5067 (new_line_info_table): Initialize symviews_since_reset.
5068 (dwarf2out_source_line): Count symviews_since_reset and set
5069 symview_upper_bound.
5070 (dw_val_equal_p): Handle symview.
5071 (add_AT_symview): New.
5072 (print_dw_val): Handle symview.
5073 (attr_checksum, attr_checksum_ordered): Likewise.
5074 (same_dw_val_p, size_of_die): Likewise.
5075 (value_format, output_die): Likewise.
5076 (add_high_low_attributes): Use add_AT_symview for entry_view.
5077 (dwarf2out_finish): Reset symview_upper_bound, clear
5080 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5083 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5084 Add strict argument and use it.
5085 (rs6000_split_multireg_move): Update for new strict argument.
5086 (mem_operand_gpr): Disallow all non-offsettable addresses.
5087 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5089 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5092 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5093 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5094 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5097 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5098 decl, use remap_type if we want to use the type.
5100 2018-03-09 Martin Sebor <msebor@redhat.com>
5102 PR tree-optimization/84526
5103 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5105 (builtin_access::generic_overlap): Be prepared to handle non-array
5108 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5110 PR rtl-optimization/84682
5111 * lra-constraints.c (process_address_1): Check is_address flag
5112 for address constraints.
5113 (process_alt_operands): Likewise.
5114 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5115 preprocess_constraints.
5116 * recog.h (preprocess_constraints): Add oploc parameter.
5118 * recog.c (preprocess_constraints): Test address_operand for
5119 CT_ADDRESS constraints.
5121 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5124 * lra-assigns.c (assign_by_spills): Return a flag of reload
5125 assignment failure. Do not process the reload assignment
5126 failures. Do not spill other reload pseudos if they has the same
5128 (lra_assign): Add a return arg. Set up from the result of
5129 assign_by_spills call.
5130 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5131 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5132 usage_insns if it is not NULL.
5133 (spill_hard_reg_in_range): New function.
5134 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5135 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5136 function prototypes.
5137 (lra_assign): Change prototype.
5138 * lra.c (lra): Add code to deal with fails by splitting hard reg
5141 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5144 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5145 Accept complain bool parameter. Only emit errors if it is true.
5146 (arm_parse_cpu_option_name): Likewise.
5147 (arm_target_thumb_only): Adjust callers of the above.
5148 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5149 prototype to take a default true bool parameter.
5150 (arm_parse_arch_option_name): Likewise.
5152 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5153 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5157 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5158 * configure: Regenerate.
5159 * configure.ac ("linker --version-script option"): New.
5160 ("linker soname option"): New.
5162 2018-03-09 Richard Biener <rguenther@suse.de>
5164 PR tree-optimization/84775
5165 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5166 immediate uses of predicate stmts and mark them modified.
5169 PR tree-optimization/84178
5170 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5172 (version_loop_for_if_conversion): Delay update_ssa call.
5173 (tree_if_conversion): Delay update_ssa until after predicate
5176 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5179 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5180 when the function accesses prior frames.
5182 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5185 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5186 gen_llsym, otherwise call maybe_gen_llsym.
5189 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5190 has ',' character inside of it.
5192 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5195 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5196 as clobbering CC_REGNUM.
5198 2018-03-08 Richard Biener <rguenther@suse.de>
5201 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5202 (follow_copies_to_constant): Do not follow SSA names registered
5205 2018-03-08 Richard Biener <rguenther@suse.de>
5207 PR tree-optimization/84178
5208 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5210 (version_loop_for_if_conversion): Delay update_ssa call.
5211 (tree_if_conversion): Delay update_ssa until after predicate
5214 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5216 PR tree-optimization/84178
5217 * tree-if-conv.c (release_bb_predicate): Remove the
5218 the assertion that the stmts have NULL use_ops.
5219 Discard the statements, asserting that they haven't
5220 yet been added to a BB.
5222 2018-03-08 Richard Biener <rguenther@suse.de>
5224 PR tree-optimization/84746
5225 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5226 (phi_translate): Pass in destination ANTIC_OUT set.
5227 (phi_translate_1): Likewise. For a simplified result lookup
5228 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5229 (phi_translate_set): Adjust.
5230 (do_pre_regular_insertion): Likewise.
5231 (do_pre_partial_partial_insertion): Likewise.
5233 2018-03-08 Martin Liska <mliska@suse.cz>
5235 PR gcov-profile/84735
5236 * doc/gcov.texi: Document usage of profile files.
5237 * gcov-io.h: Document changes in the format.
5239 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5243 * dwarf2out.c (struct dw_line_info_table): Update comments for
5245 (FORCE_RESET_NEXT_VIEW): New.
5246 (FORCE_RESETTING_VIEW_P): New.
5247 (RESETTING_VIEW_P): Check for -1 too.
5248 (ZERO_VIEW_P): Likewise.
5249 (new_line_info_table): Force-reset next view.
5250 (dwarf2out_begin_function): Likewise.
5251 (dwarf2out_source_line): Simplify zero_view_p initialization.
5252 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5253 view directly. Omit view when omitting .loc at line 0.
5255 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5257 PR tree-optimization/84740
5258 * tree-switch-conversion.c (process_switch): Call build_constructors
5259 only if info.phi_count is non-zero.
5261 PR tree-optimization/84739
5262 * tree-tailcall.c (find_tail_calls): Check call arguments against
5263 DECL_ARGUMENTS (current_function_decl) rather than
5264 DECL_ARGUMENTS (func) when checking for tail recursion.
5266 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5268 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5269 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5270 Volker Reichelt's entry and add entries for people that perform
5271 GCC fuzzy testing and report numerous bugs.
5273 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5276 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5277 readonly data in sdata, if that is disabled.
5278 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5279 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5280 -mreadonly-in-sdata option.
5282 2018-03-07 Martin Sebor <msebor@redhat.com>
5284 PR tree-optimization/84468
5285 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5286 basic block when looking for nul assignment.
5288 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5291 * except.h (output_function_exception_table): Adjust prototype.
5292 * except.c (output_function_exception_table): Remove FNNAME parameter
5293 and add SECTION parameter. Ouput one part of the table at a time.
5294 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5295 the first part of the exception table and emit unwind directives.
5296 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5297 (i386_pe_seh_cold_init): Likewise.
5298 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5299 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5300 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5301 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5302 * config/i386/winnt.c: Include except.h.
5303 (struct seh_frame_state): Add reg_offset, after_prologue and
5304 in_cold_section fields.
5305 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5306 (i386_pe_seh_cold_init): New function.
5307 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5308 to seh->in_cold_section.
5309 (seh_emit_push): Record the offset of the push.
5310 (seh_emit_save): Record the offet of the save.
5311 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5312 Test seh->after_prologue to disregard the epilogue.
5313 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5314 (i386_pe_end_cold_function): New function.
5316 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5319 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5320 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5323 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5325 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5326 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5329 * multiple_target.c: Include tree-inline.h and intl.h.
5330 (expand_target_clones): Diagnose and fail if node->definition and
5331 !tree_versionable_function_p (node->decl).
5333 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5335 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5337 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5338 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5339 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5341 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5344 * combine.c (try_combine): Use reg_or_subregno instead of handling
5345 just paradoxical SUBREGs and REGs.
5347 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5349 * config/arc/arc.c (arc_finalize_pic): Remove function.
5350 (arc_must_save_register): We use single base PIC register, remove
5351 checks to save/restore the PIC register.
5352 (arc_expand_prologue): Likewise.
5353 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5355 (arc_verify_short): Likewise.
5356 (arc_attr_type): Likewise.
5357 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5358 (walk_stores): Likewise.
5359 (arc_address_cost): Make it static.
5360 (arc_verify_short): Likewise.
5361 (branch_dest): Likewise.
5362 (arc_attr_type): Likewise.
5363 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5364 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5365 (arc_final_prescan_insn): Remove inserting the nops due to
5366 hardware hazards. It is done in reorg step.
5367 (insn_length_variant_t): Remove.
5368 (insn_length_parameters_t): Likewise.
5369 (arc_insn_length_parameters): Likewise.
5370 (arc_get_insn_variants): Likewise.
5371 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5373 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5376 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5379 PR tree-optimization/84687
5380 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5382 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5384 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5386 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5387 Rename to ppc_speculation_barrier.
5388 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5389 __builtin_ppc_speculation_barrier.
5391 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5394 * combine.c (combine_simplify_rtx): Don't try to simplify if
5395 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5398 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5400 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5401 to 32 bytes when compiling for POWER9.
5403 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5406 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5407 regparm >= 3 with no arg reg available also for calls with
5408 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5411 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5413 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5415 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5418 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5420 2018-03-05 Richard Biener <rguenther@suse.de>
5422 PR tree-optimization/84486
5423 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5424 When inserting a __builtin_assume_aligned call set the LHS
5425 SSA name alignment info accordingly.
5427 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5429 PR tree-optimization/84114
5430 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5431 Avoid reassociation of FLOAT_MODE addition.
5433 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5435 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5436 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5437 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5438 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5439 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5440 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5441 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5443 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5444 __builtin_ia32_wbinvd): New builtins.
5445 (SPECIAL_ARGS2): New.
5446 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5447 (SPECIAL_ARGS2): New.
5448 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5449 (ix86_valid_target_attribute_inner_p): Ditto.
5450 (ix86_init_mmx_sse_builtins): Add special_args2.
5451 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5452 TARGET_WBNOINVD_P): New.
5453 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5454 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5455 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5456 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5457 * config/i386/pconfigintrin.h: New file.
5458 * config/i386/wbnoinvdintrin.h: Ditto.
5459 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5460 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5462 2018-03-05 Richard Biener <rguenther@suse.de>
5464 PR tree-optimization/84670
5465 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5467 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5468 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5469 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5470 doesn't grow if all successors (recursively) were visited at least
5473 2018-03-05 Richard Biener <rguenther@suse.de>
5475 PR tree-optimization/84650
5476 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5477 if executed in the loop pipeline.
5479 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5481 * doc/configfiles.texi (Configuration Files): Move info about
5482 conditionalizing $target-protos.h to...
5483 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5484 differs from $target-protos.h.
5486 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5487 Chung-Ju Wu <jasonwucj@gmail.com>
5489 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5490 * config/nds32/nds32-multiple.md (setmemsi): Define.
5491 * config/nds32/nds32-memory-manipulation.c
5492 (nds32_gen_dup_4_byte_to_word_value): New.
5493 (emit_setmem_word_loop): New.
5494 (emit_setmem_byte_loop): New.
5495 (nds32_expand_setmem_loop): New.
5496 (nds32_expand_setmem_loop_v3m): New.
5497 (nds32_expand_setmem_unroll): New.
5498 (nds32_expand_setmem): New.
5500 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5501 Chung-Ju Wu <jasonwucj@gmail.com>
5503 * config/nds32/nds32-memory-manipulation.c
5504 (nds32_emit_load_store): New.
5505 (nds32_emit_post_inc_load_store): New.
5506 (nds32_emit_mem_move): New.
5507 (nds32_emit_mem_move_block): New.
5508 (nds32_expand_movmemsi_loop_unknown_size): New.
5509 (nds32_expand_movmemsi_loop_known_size): New.
5510 (nds32_expand_movmemsi_loop): New.
5511 (nds32_expand_movmemsi_unroll): New.
5512 (nds32_expand_movmemqi): Rename ...
5513 (nds32_expand_movmemsi): ... to this.
5514 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5515 (movmemsi): ... to this.
5516 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5517 (nds32_expand_movmemsi): ... to this.
5519 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5520 Monk Chiang <sh.chiang04@gmail.com>
5521 Chung-Ju Wu <jasonwucj@gmail.com>
5523 * config/nds32/nds32-protos.h
5524 (nds32_expand_load_multiple): New arguments.
5525 (nds32_expand_store_multiple): Ditto.
5526 (nds32_valid_multiple_load_store): Rename ...
5527 (nds32_valid_multiple_load_store_p): ... to this.
5528 * config/nds32/nds32-memory-manipulation.c
5529 (nds32_expand_load_multiple): Refine implementation.
5530 (nds32_expand_store_multiple): Ditto.
5531 * config/nds32/nds32-multiple.md
5532 (load_multiple): Update nds32_expand_load_multiple interface.
5533 (store_multiple): Update nds32_expand_store_multiple interface.
5534 * config/nds32/nds32-predicates.c
5535 (nds32_valid_multiple_load_store): Rename ...
5536 (nds32_valid_multiple_load_store_p): ... to this and refine
5538 * config/nds32/predicates.md
5539 (nds32_load_multiple_and_update_address_operation): New predicate.
5540 (nds32_store_multiple_and_update_address_operation): New predicate.
5542 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5543 Chung-Ju Wu <jasonwucj@gmail.com>
5545 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5546 (combo): New attribute.
5547 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5549 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5551 * config/nds32/nds32.opt: Change -mcmodel= default value.
5553 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5554 Monk Chiang <sh.chiang04@gmail.com>
5555 Chung-Ju Wu <jasonwucj@gmail.com>
5557 * config/nds32/constants.md (unspec_element): New enum.
5558 * config/nds32/constraints.md (Umw): New constraint.
5559 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5560 * config/nds32/nds32-intrinsic.md: Likewise.
5561 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5562 (nds32_valid_smw_lwm_base_p): New.
5563 (nds32_output_smw_single_word): New.
5564 (nds32_output_lmw_single_word): New.
5565 (nds32_expand_unaligned_load): New.
5566 (nds32_expand_unaligned_store): New.
5567 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5568 (nds32_output_smw_single_word): Declare.
5569 (nds32_output_lmw_single_word): Declare.
5570 (nds32_expand_unaligned_load): Declare.
5571 (nds32_expand_unaligned_store): Declare.
5572 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5573 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5574 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5575 NDS32_BUILTIN_UASTORE_DW.
5576 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5579 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5580 Kito Cheng <kito.cheng@gmail.com>
5581 Chung-Ju Wu <jasonwucj@gmail.com>
5583 * config/nds32/nds32-intrinsic.c
5584 (nds32_expand_builtin_null_ftype_reg): Delete.
5585 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5586 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5587 (nds32_read_argument): New.
5588 (nds32_legitimize_target): Ditto.
5589 (nds32_legitimize_argument): Ditto.
5590 (nds32_check_constant_argument): Ditto.
5591 (nds32_expand_unop_builtin): Ditto.
5592 (nds32_expand_unopimm_builtin): Ditto.
5593 (nds32_expand_binop_builtin): Ditto.
5594 (nds32_builtin_decl_impl): Ditto.
5595 (builtin_description): Ditto.
5596 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5597 (nds32_init_builtins_impl): Ditto.
5598 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5599 (nds32_builtin_decl): New.
5600 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5601 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5603 2018-03-02 Jeff Law <law@redhat.com>
5605 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5606 (redundant_insn, fill_simple_delay_slots): Likewise.
5607 (fill_slots_from_thread): Likewise.
5608 * resource.c (mark_referenced_resources): Likewise.
5609 (mark_set_resources, find_dead_or_set_registers): Likewise.
5611 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5613 * substring-locations.h (format_warning_va): Formatting fix for
5615 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5617 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5618 * substring-locations.c: Include intl.h.
5619 (format_warning_va): Turned into small wrapper around
5620 format_warning_n_va, renamed to ...
5621 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5622 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5624 (format_warning_at_substring_n): New function.
5625 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5626 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5627 format_warning_at_substring with just a shorter name instead of
5628 const function pointer.
5629 (fmtwarn_n): New function.
5630 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5631 appropriate, get rid of all the fmtstr temporaries, move conditionals
5632 with G_() wrapped string literals directly into fmtwarn arguments,
5633 cast dir.len to (int), formatting fixes.
5635 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5637 * doc/invoke.texi: Remove "Cilk Plus" references.
5639 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5640 Richard Biener <rguenther@suse.de>
5643 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5644 for error or warning attributes if CALL_FROM_THUNK_P is set.
5647 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5650 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5651 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5654 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5655 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5657 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5658 instead of -1U in last predictors element's probability member.
5660 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5663 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5664 arguments if they are comparable.
5666 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5668 PR tree-optimization/84634
5669 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5670 masks and masked_loop_p with a single loop_masks, making sure it's
5671 null for bb vectorization.
5673 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5675 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5676 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5677 loop->force_vectorize to check whether there is no alias.
5679 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5682 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5684 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5686 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5687 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5688 instead of a loop around prev_real_insn.
5689 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5693 * config/i386/i386.c (ix86_print_operand): Use conditional
5694 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5697 2018-03-02 Richard Biener <rguenther@suse.de>
5699 PR tree-optimization/84427
5700 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5701 (bitmap_set_subtract_values): Rewrite to handle multiple
5704 (prune_clobbered_mems): Likewise.
5705 (phi_translate): Take edge instead of pred/phiblock.
5706 (phi_translate_1): Likewise.
5707 (phi_translate_set): Likewise. Insert all translated
5708 exprs for a value into the set, keeping possibly multiple
5709 expressions per value.
5710 (compute_antic_aux): Adjust for phi_translate changes.
5711 When intersecting union the expressions and prune those
5712 not in the final value set, keeping possibly multiple
5713 expressions per value. Do not use value-insertion
5714 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5715 all expressions. Add verification that the value-sets
5716 only shrink during iteration.
5717 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5718 (do_pre_regular_insertion): Likewise.
5719 (do_pre_partial_partial_insertion): Likewise.
5721 2018-03-02 Richard Biener <rguenther@suse.de>
5724 * config/darwin.c (saved_debug_info_level): New static global.
5725 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5726 (darwin_asm_lto_end): Restore debug info generation settings.
5728 2018-03-01 Martin Liska <mliska@suse.cz>
5731 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5734 2018-03-01 Richard Biener <rguenther@suse.de>
5737 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5738 type annotation with LTO when debug was disabled at compile-time.
5740 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5742 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5744 (mips_final_postscan_insn): Likewise.
5746 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5748 PR rtl-optimization/84528
5749 * alias.c (init_alias_target): Add commentary.
5750 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5751 a unique base value if the frame pointer is not eliminated
5752 to the stack pointer.
5754 2018-03-01 Tom de Vries <tom@codesourcery.com>
5756 PR rtl-optimization/83327
5757 * lra-int.h (hard_regs_spilled_into): Declare.
5758 * lra.c (hard_regs_spilled_into): Define.
5759 (init_reg_info): Init hard_regs_spilled_into.
5760 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5761 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5762 (process_bb_lives): Handle hard_regs_spilled_into.
5763 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5764 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5766 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5768 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5769 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5770 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5771 * config/rs6000/aix72.h: New file.
5773 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5775 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5776 instead of warning_at with conditional singular and plural messages
5780 * stor-layout.c (update_alignment_for_field): For
5781 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5782 && !DECL_PACKED (field), do the alignment update, just use
5783 only desired_align instead of MAX (type_align, desired_align)
5785 (place_field): Don't do known_align < desired_align handling
5786 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5787 is non-NULL, instead do it after rli->prev_field handling and
5788 only if not within a bitfield word. For DECL_PACKED (field)
5789 use type_align of BITS_PER_UNIT.
5791 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5793 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5794 superfluous parentheses and trailing spaces.
5796 2018-02-28 Richard Biener <rguenther@suse.de>
5798 PR tree-optimization/84584
5799 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5800 SCoPs with fake exit edge.
5802 2018-02-28 Martin Liska <mliska@suse.cz>
5805 * timevar.c (timer::print): Fix format to properly print 100%
5808 2018-02-28 Richard Biener <rguenther@suse.de>
5811 * genmatch.c (capture_info::walk_match): Do not mark
5812 captured expressions without operands as expr_p given
5813 they act more like predicates and should be subject to
5814 "lost tail" side-effect preserving.
5816 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5818 PR rtl-optimization/81611
5819 * auto-inc-dec.c (attempt_change): Move dead note from
5820 mem_insn if it's the next use of regno
5821 (find_address): Take address use of reg holding
5822 non-incremented value. Add parm to limit search to the named
5824 (merge_in_block): Attempt to use a mem insn that is the next
5825 use of the original regno.
5827 2018-02-27 Martin Sebor <msebor@redhat.com>
5830 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5831 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5833 2018-02-27 Martin Sebor <msebor@redhat.com>
5835 PR translation/84207
5836 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5837 n argument to unsigned HOST_WIDE_INT.
5838 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5839 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5840 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5841 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5843 2018-02-27 Richard Biener <rguenther@suse.de>
5845 PR tree-optimization/84512
5846 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5847 Do not use the estimate returned from record_stmt_cost for
5848 the scalar iteration cost but sum properly using add_stmt_cost.
5850 2018-02-27 Richard Biener <rguenther@suse.de>
5852 PR tree-optimization/84466
5853 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5854 Adjust last change to less strictly validate use operands.
5856 2018-02-27 Martin Liska <mliska@suse.cz>
5858 PR gcov-profile/84548
5859 * gcov.c (process_file): Allow partial overlap and consider it
5860 also as group functions.
5861 (output_lines): Properly calculate range of lines for a group.
5863 2018-02-27 Martin Liska <mliska@suse.cz>
5865 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5866 'ggc' suffixes. Change first column width.
5867 (timer::print): Fix formatting of the column.
5869 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5871 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5872 preserve inline entry blocks for the sake of debug inline
5873 entry point markers alone.
5874 (remove_unused_locals): Suggest in comments a better place to
5875 force the preservation of inline entry blocks that are
5876 otherwise unused, but do not preserve them.
5878 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5880 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5882 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5885 * config/i386/constraints.md (Bs): Replace
5886 ix86_indirect_branch_register with
5887 TARGET_INDIRECT_BRANCH_REGISTER.
5889 * config/i386/i386.md (indirect_jump): Likewise.
5890 (tablejump): Likewise.
5891 (*sibcall_memory): Likewise.
5892 (*sibcall_value_memory): Likewise.
5893 Peepholes of indirect call and jump via memory: Likewise.
5894 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5895 (*sibcall_value_GOT_32): Likewise.
5896 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5897 (GOT_memory_operand): Likewise.
5898 (call_insn_operand): Likewise.
5899 (sibcall_insn_operand): Likewise.
5900 (GOT32_symbol_operand): Likewise.
5901 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5903 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5905 PR rtl-optimization/83496
5906 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5907 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5908 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5909 redundant insn, if any.
5910 (relax_delay_slots): Likewise.
5911 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5913 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5915 PR tree-optimization/83965
5916 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5917 that grouped statements are part of a reduction chain. Return
5918 true if the statement is not marked as a reduction itself but
5920 (vect_recog_dot_prod_pattern): Don't check whether the statement
5921 is part of a group here.
5922 (vect_recog_sad_pattern): Likewise.
5923 (vect_recog_widen_sum_pattern): Likewise.
5925 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5928 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5930 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5933 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5935 (ix86_output_indirect_function_return): New prototype.
5936 (ix86_split_simple_return_pop_internal): Likewise.
5937 * config/i386/i386.c (indirect_return_via_cx): New.
5938 (indirect_return_via_cx_bnd): Likewise.
5939 (indirect_thunk_name): Handle return va CX_REG.
5940 (output_indirect_thunk_function): Create alias for
5941 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5942 (ix86_output_indirect_jmp): Remove the bool argument.
5943 (ix86_output_indirect_function_return): New function.
5944 (ix86_split_simple_return_pop_internal): Likewise.
5945 * config/i386/i386.md (*indirect_jump): Don't pass false
5946 to ix86_output_indirect_jmp.
5947 (*tablejump_1): Likewise.
5948 (simple_return_pop_internal): Change it to define_insn_and_split.
5949 Call ix86_split_simple_return_pop_internal to split it for
5951 (simple_return_indirect_internal): Call
5952 ix86_output_indirect_function_return instead of
5953 ix86_output_indirect_jmp.
5955 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5958 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5959 memset and value initialization afterwards.
5961 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5963 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5965 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5968 * common/config/aarch64/aarch64-common.c
5969 (aarch_option_optimization_table[]): Switch
5970 off fomit-frame-pointer
5972 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5973 Chung-Ju Wu <jasonwucj@gmail.com>
5975 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5977 (store_multiple): Ditto.
5979 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5981 * config.gcc: Add --with-cpu support for nds32 target.
5982 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5983 * config/nds32/nds32.opt: Add -mcpu= option.
5985 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5987 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5988 isel=yes): Warn for these deprecated options.
5990 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5992 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5993 ISA_2_5_MASKS_EMBEDDED.
5995 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5997 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5998 p->max as pointers rather than using iterative_hash_expr.
6000 2018-02-23 Carl Love <cel@us.ibm.com>
6002 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6003 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6005 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6006 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6007 P8V_BUILTIN_VEC_VUNSIGNED2.
6009 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6012 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6013 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6015 * lra-constraints.c (curr_insn_transform): Set up
6016 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6018 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6019 * lra-eliminations.c (spill_pseudos): Ditto.
6020 (process_insn_for_elimination): Ditto.
6021 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6022 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6023 LRA_NON_CLOBBERED_ALT.
6025 2018-02-22 Martin Sebor <msebor@redhat.com>
6027 PR tree-optimization/84480
6028 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6029 to maybe_diag_stxncpy_trunc. Call it.
6030 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6031 from gimple_fold_builtin_strcpy. Print inlining stack.
6032 (handle_builtin_stxncpy): Print inlining stack.
6033 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6035 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6038 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6039 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6040 and -fcheck-pointer-bounds are used together.
6041 (indirect_thunk_prefix): New enum.
6042 (indirect_thunk_need_prefix): New function.
6043 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6044 "_nt" instead of "_bnd" for NOTRACK prefix.
6045 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6046 (output_indirect_thunk_function): Likewise.
6048 (ix86_code_end): Update output_indirect_thunk_function calls.
6049 (ix86_output_indirect_branch_via_reg): Replace
6050 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6051 (ix86_output_indirect_branch_via_push): Likewise.
6052 (ix86_output_function_return): Likewise.
6053 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6054 incompatible with -fcf-protection=branch and
6055 -fcheck-pointer-bounds.
6057 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6060 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6061 Change gcc_assert call to output_operand_lossage.
6063 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6065 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6067 2018-02-22 DJ Delorie <dj@redhat.com>
6068 Sebastian Perta <sebastian.perta@renesas.com>
6069 Oleg Endo <olegendo@gcc.gnu.org>
6071 * config/rx/rx.c (rx_rtx_costs): New function.
6072 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6074 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6076 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6078 2018-02-22 Martin Liska <mliska@suse.cz>
6081 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6082 Add "native" as a possible value.
6084 2018-02-22 Martin Liska <mliska@suse.cz>
6087 * config/i386/i386.c (ix86_option_override_internal):
6088 Add "native" as a possible value for -march and -mtune.
6090 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6093 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6094 to all type variants.
6096 PR tree-optimization/84503
6097 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6098 width as info->bitpos + info->bitsize - start.
6099 (merged_store_group::merge_overlapping): Simplify width computation.
6100 (check_no_overlap): New function.
6101 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6102 start + width and last_order of the group, fail if check_no_overlap
6104 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6105 to group if check_no_overlap fails.
6107 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6109 * config/rs6000/altivec.md: Delete contraint arguments to
6110 define_expand, define_split, and define_peephole2, and in
6111 define_insn_and_split if always unused.
6112 * config/rs6000/darwin.md: Ditto.
6113 * config/rs6000/dfp.md: Ditto.
6114 * config/rs6000/rs6000.md: Ditto.
6115 * config/rs6000/sync.md: Ditto.
6116 * config/rs6000/vector.md: Ditto.
6117 * config/rs6000/vsx.md: Ditto.
6119 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6121 * config/rs6000/altivec.md: Write output control strings as braced
6122 blocks instead of double-quoted strings.
6123 * config/rs6000/darwin.md: Ditto.
6124 * config/rs6000/rs6000.md: Ditto.
6125 * config/rs6000/vector.md: Ditto.
6126 * config/rs6000/vsx.md: Ditto.
6128 2018-02-21 Jason Merrill <jason@redhat.com>
6130 PR c++/84314 - ICE with templates and fastcall attribute.
6131 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6133 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6135 * ipa-cp.c (determine_versionability): Fix comment typos.
6137 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6140 * ipa-cp.c (determine_versionability): Do not version functions caling
6143 2018-02-21 Martin Liska <mliska@suse.cz>
6146 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6147 Add "native" as a possible value.
6148 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6149 the macro when native cpu detection is available.
6151 2018-02-21 Martin Liska <mliska@suse.cz>
6154 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6155 Add "native" as a possible value.
6156 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6157 when native cpu detection is available.
6159 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6160 Martin Sebor <msebor@redhat.com>
6162 PR tree-optimization/84478
6163 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6165 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6166 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6167 support which is conservatively correct, for 2 only stay conservative
6168 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6169 argument to the 2 argument get_range_strlen, adjust 6 arg
6170 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6172 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6173 (gimple_fold_builtin_strlen): Pass true as last argument to
6176 2018-02-20 Martin Sebor <msebor@redhat.com>
6179 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6180 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6181 (builtin_memref::builtin_memref): Factor out parts into
6182 set_base_and_offset and call it.
6184 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6187 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6188 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6189 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6190 search at the associated MODE_INT.
6192 2018-02-20 Jeff Law <law@redhat.com>
6195 PR tree-optimization/81592
6197 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6198 for range data rather than using global data.
6199 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6200 range data rather than using global data.
6201 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6202 pass it to children as needed.
6203 (struct directive::fmtresult): Similarly.
6204 (struct directive::set_width): Similarly.
6205 (struct directive::set_precision): Similarly.
6206 (format_integer, format_directive, parse_directive): Similarly.
6207 (format_none): Accept unnamed vr_values parameter.
6208 (format_percent, format_floating, format_character): Similarly.
6209 (format_string, format_plain): Similarly.
6210 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6211 the EVRP range analyzer for range data rather than using global data.
6212 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6213 gimple-ssa-evrp-analyze.h
6214 (class sprintf_dom_walker): Add after_dom_children member function.
6215 Add evrp_range_analyzer member.
6216 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6217 range analyzer as needed.
6218 (sprintf_dom_walker::after_dom_children): New member function.
6219 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6221 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6222 (evrp_range_analyzer::pop_to_marker): Likewise.
6224 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6226 PR tree-optimization/84419
6227 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6228 with the required type if its current type is compatible but
6231 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6234 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6235 after vectorization.
6237 2018-02-20 Martin Liska <mliska@suse.cz>
6240 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6241 possible values if we don't have a hint.
6243 2018-02-20 Martin Liska <mliska@suse.cz>
6247 * final.c (shorten_branches): Build align_tab array with one
6249 * opts.c (finish_options): Add alignment option limit check.
6250 (MAX_CODE_ALIGN): Likewise.
6251 (MAX_CODE_ALIGN_VALUE): Likewise.
6252 * doc/invoke.texi: Document maximum allowed option value for
6253 all -falign-* options.
6255 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6258 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6259 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6260 * var-tracking.c (emit_note_insn_var_location): Remove all references
6261 to NOTE_INSN_CALL_ARG_LOCATION.
6262 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6263 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6264 Use copy_rtx_if_shared.
6265 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6266 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6267 (dwarf2out_var_location): Remove handling of
6268 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6270 * final.c (final_scan_insn): Remove all references to
6271 NOTE_INSN_CALL_ARG_LOCATION.
6272 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6273 before dumping final insns.
6274 * except.c (emit_note_eh_region_end): Remove all references to
6275 NOTE_INSN_CALL_ARG_LOCATION.
6276 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6277 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6278 * config/arc/arc.c (hwloop_optimize): Likewise.
6279 * config/arm/arm.c (create_fix_barrier): Likewise.
6280 * config/s390/s390.c (s390_chunkify_start): Likewise.
6281 * config/sh/sh.c (find_barrier): Likewise.
6282 * config/i386/i386.c (rest_of_insert_endbranch,
6283 ix86_seh_fixup_eh_fallthru): Likewise.
6284 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6285 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6286 * config/frv/frv.c (frv_function_prologue): Likewise.
6287 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6289 (note_outside_basic_block_p): Remove all references to
6290 NOTE_INSN_CALL_ARG_LOCATION.
6291 * gengtype.c (adjust_field_rtx_def): Likewise.
6292 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6294 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6295 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6298 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6301 PR tree-optimization/84452
6302 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6303 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6306 2018-02-19 Martin Liska <mliska@suse.cz>
6309 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6311 2018-02-19 Martin Liska <mliska@suse.cz>
6312 Richard Sandiford <richard.sandiford@linaro.org>
6314 PR tree-optimization/82491
6315 * gimple-fold.c (get_base_constructor): Make earlier bail out
6318 2018-02-19 Carl Love <cel@us.ibm.com>
6320 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6321 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6323 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6324 P8V_BUILTIN_VEC_NEG.
6326 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6328 * config/rl78/rl78.md (movdf): New define expand.
6330 2018-02-19 Martin Liska <mliska@suse.cz>
6333 * doc/invoke.texi: Fix typo.
6334 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6336 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6338 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6339 handle rs6000_single_float and rs6000_double_float specially for
6342 2018-02-16 Jeff Law <law@redhat.com>
6344 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6347 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6350 * ipa-inline.c (inline_small_functions): Fix a typo.
6352 2018-02-16 Nathan Sidwell <nathan@acm.org>
6354 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6356 2018-02-16 Carl Love <cel@us.ibm.com>
6358 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6359 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6360 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6361 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6362 expansion to P8V_BUILTIN_VEC_FLOAT2.
6364 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6366 PR rtl-optimization/70023
6367 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6368 src_regno into account.
6370 2018-02-16 Carl Love <cel@us.ibm.com>
6372 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6373 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6374 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6375 * config/rs6000/rs6000.c: Remove case statements for
6376 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6377 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6378 and P9V_BUILTIN_VEC_VINSERT4B.
6379 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6380 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6381 * config/rs6000/vsx.md:
6382 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6385 2018-02-16 Carl Love <cel@us.ibm.com>
6387 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6389 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6391 * config/rs6000/rs6000-c.c: Add the definitions for
6392 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6393 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6394 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6395 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6396 definition for insert4b and define insn *insert3b_internal.
6397 * doc/extend.texi: Add documentation for vec_extract4b.
6399 2018-02-16 Nathan Sidwell <nathan@acm.org>
6401 * doc/extend.texi (Backwards Compatibility): Mention friend
6402 injection. Note for-scope is deprecated.
6403 * doc/invoke.texi (-ffriend-injection): Deprecate.
6405 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6407 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6408 that moved to I2, also allow destinations that are a paradoxical
6409 subreg (instead of a normal reg).
6411 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6414 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6417 2018-02-16 Richard Biener <rguenther@suse.de>
6419 PR tree-optimization/84037
6420 PR tree-optimization/84016
6422 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6423 Adjust vec_construct for the fact we need additional higher latency
6424 128bit inserts for AVX256 and AVX512 vector builds.
6425 (ix86_add_stmt_cost): Scale vector construction cost for
6428 2018-02-16 Richard Biener <rguenther@suse.de>
6430 PR tree-optimization/84417
6431 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6432 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6433 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6435 2018-02-16 Martin Liska <mliska@suse.cz>
6438 * internal-fn.def (ASAN_CHECK): Set proper flags.
6439 (ASAN_MARK): Likewise.
6441 2018-02-16 Julia Koval <julia.koval@intel.com>
6443 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6444 from PTA_CANNONLAKE.
6446 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6449 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6450 Use ++iter rather than iter++ for std::list iterators.
6451 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6452 defer deleting them until all nodes in the forest are processed. Do
6453 free even leaf nodes. Change to_process into auto_vec.
6456 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6457 * vec.h (vec_default_construct): Use memset instead of placement new
6458 if BROKEN_VALUE_INITIALIZATION is defined.
6459 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6460 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6463 PR rtl-optimization/83723
6464 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6465 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6466 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6468 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6470 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6472 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6474 PR rtl-optimization/81443
6475 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6476 from inner REGs to paradoxical SUBREGs.
6478 2018-02-16 Richard Biener <rguenther@suse.de>
6480 PR tree-optimization/84399
6481 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6482 For operands we can analyze at their definition make sure we can
6483 analyze them at each use as well.
6485 2018-02-16 Richard Biener <rguenther@suse.de>
6487 PR tree-optimization/84190
6488 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6489 volatile accesses if the decl isn't volatile.
6491 2018-02-15 Jason Merrill <jason@redhat.com>
6493 PR c++/84314 - ICE with templates and fastcall attribute.
6494 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6495 TYPE_CANONICAL on an existing type.
6497 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6499 PR tree-optimization/84383
6500 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6501 dstoff nor call operand_equal_p if dstbase is NULL.
6503 PR tree-optimization/84334
6504 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6505 also a CONSTANT_CLASS_P, punt.
6507 2018-02-14 Jim Wilson <jimw@sifive.com>
6509 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6510 first SMALL_OPERAND check. New local min_second_step. Move assert
6511 to where locals are set. Add TARGET_RVC support.
6512 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6514 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6516 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6518 2018-02-14 Martin Sebor <msebor@redhat.com>
6520 PR tree-optimization/83698
6521 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6522 arrays constrain the offset range to their bounds.
6523 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6524 (builtin_access::overlap): Avoid setting the size of overlap if it's
6526 (maybe_diag_overlap): Also consider arrays when deciding what values
6527 of offsets to include in diagnostics.
6529 2018-02-14 Martin Sebor <msebor@redhat.com>
6532 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6533 that correspond to the kind of a declaration.
6535 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6538 * config/pa/pa.md: Load address of PIC label using the linkage table
6539 if the label is nonlocal.
6541 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6543 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6544 warning message if user requests -maltivec=be.
6545 * doc/invoke.texi: Document deprecation of -maltivec=be.
6547 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6550 * config/rs6000/rs6000-c.c: Update definitions for
6551 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6552 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6554 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6557 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6558 add _get_ssp intrinsics. Remove argument from
6559 __builtin_ia32_rdssp[d|q].
6560 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6561 * config/i386/i386-builtin.def: Remove argument from
6562 __builtin_ia32_rdssp[d|q].
6563 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6564 ix86_expand_special_args_builtin for _rdssp[d|q].
6565 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6566 Clear register before usage.
6567 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6568 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6570 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6572 PR tree-optimization/84357
6573 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6574 operand 1 of an ARRAY_REF too.
6576 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6579 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6580 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6582 (set_of_reg): New struct.
6583 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6584 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6585 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6587 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6588 Split into bitclr, bitset, bitinvert patterns if appropriate.
6589 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6590 use rx_fuse_in_memory_bitop.
6591 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6592 to named insn, correct maximum insn length.
6594 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6597 * machmode.def: Define a complex mode for PARTIAL_INT.
6598 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6600 * doc/rtl.texi: Document CSPImode.
6601 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6603 (msp430_hard_regno_nregs_with_padding): Likewise.
6605 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6608 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6610 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6612 PR rtl-optimization/84169
6613 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6614 we generated a parallel as new i3 and we split that to new i2 and i3
6615 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6616 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6617 those to i2, not i1. Partially rewrite this scan code.
6619 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6622 * stor-layout.c (place_field): For variable length fields, adjust
6623 offset_align afterwards not just based on the field's alignment,
6624 but also on the size.
6627 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6628 of exps and logs in the use_exp2 case.
6630 2018-02-13 Jeff Law <law@redhat.com>
6632 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6635 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6638 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6642 * common.opt (gas-loc-support, gas-locview-support): New.
6643 (ginline-points, ginternal-reset-location-views): New.
6644 * doc/invoke.texi: Document them. Use @itemx where intended.
6645 (gvariable-location-views): Adjust.
6646 * target.def (reset_location_view): New.
6647 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6648 (TARGET_RESET_LOCATION_VIEW): New.
6649 * doc/tm.texi: Rebuilt.
6650 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6651 (dwarf2out_default_as_locview_support): New.
6652 (output_asm_line_debug_info): Use option variables.
6653 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6654 (output_loc_list): Likewise.
6655 (add_high_low_attributes): Check option variables.
6656 Don't output entry view attribute in strict mode.
6657 (gen_inlined_subroutine_die): Check option variables.
6658 (dwarf2out_inline_entry): Likewise.
6659 (init_sections_and_labels): Likewise.
6660 (dwarf2out_early_finish): Likewise.
6661 (maybe_reset_location_view): New, from...
6662 (dwarf2out_var_location): ... here. Call it.
6663 * debug.h (dwarf2out_default_as_loc_support): Declare.
6664 (dwarf2out_default_as_locview_support): Declare.
6665 * hooks.c (hook_int_rtx_insn_0): New.
6666 * hooks.h (hook_int_rtx_insn_0): Declare.
6667 * toplev.c (process_options): Take -gas-loc-support and
6668 -gas-locview-support from dwarf2out. Enable
6669 -gvariable-location-views by default only with locview
6670 assembler support. Enable -ginternal-reset-location-views by
6671 default only if the target defines the corresponding hook.
6672 Enable -ginline-points by default if location views are
6673 enabled; force it disabled if statement frontiers are
6675 * tree-inline.c (expand_call_inline): Check option variables.
6676 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6678 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6680 PR tree-optimization/84321
6681 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6682 handling. Also check whether the anti-range contains any values
6683 that satisfy the mask; switch to a VR_RANGE if not.
6685 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6688 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6690 2018-02-13 Martin Jambor <mjambor@suse.cz>
6693 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6694 of call statements, also set location of a load to a temporary.
6696 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6698 * config/rl78/rl78.c (add_vector_labels): New function.
6699 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6700 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6701 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6702 which checks that no arguments are passed.
6703 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6704 * doc/extend.texi: Documentation for the new attribute.
6706 2018-02-13 Andreas Schwab <schwab@suse.de>
6708 * config/riscv/linux.h (CPP_SPEC): Define.
6710 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6713 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6714 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6715 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6716 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6717 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6718 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6719 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6720 temporarily for AES and PCLMUL builtins.
6722 PR tree-optimization/84339
6723 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6724 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6728 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6729 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6730 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6732 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6734 * omp-simd-clone.h: New file.
6735 * omp-simd-clone.c: Include omp-simd-clone.h.
6736 (expand_simd_clones): No longer static.
6737 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6738 cgraph.h and omp-simd-clone.h.
6739 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6740 (vect_recog_widen_shift_pattern): Formatting fix.
6741 (vect_pattern_recog_1): Don't check optab for calls.
6744 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6745 operands[2] into a REG before using gen_lowpart on it.
6747 2018-02-12 Jeff Law <law@redhat.com>
6750 * config/sh/sh.c (find_barrier): Consider a sibling call
6753 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6754 successfully back substituting a reg.
6756 2018-02-12 Richard Biener <rguenther@suse.de>
6758 PR tree-optimization/84037
6759 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6760 parameter, move visited init to caller.
6761 (vect_slp_analyze_operations): Separate cost from validity
6762 check, initialize visited once for all instances.
6763 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6765 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6767 (vect_model_promotion_demotion_cost): Likewise.
6768 (vectorizable_bswap): Guard cost modeling with !slp_node
6769 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6771 (vectorizable_call): Likewise.
6772 (vectorizable_conversion): Likewise.
6773 (vectorizable_assignment): Likewise.
6774 (vectorizable_shift): Likewise.
6775 (vectorizable_operation): Likewise.
6776 (vectorizable_store): Likewise.
6777 (vectorizable_load): Likewise.
6778 (vectorizable_condition): Likewise.
6779 (vectorizable_comparison): Likewise.
6781 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6784 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6785 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6786 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6788 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6791 * params.def (inline-min-speedup): Increase from 8 to 15.
6792 (max-inline-insns-auto): Decrease from 40 to 30.
6793 * ipa-split.c (consider_split): Add some buffer for function to
6794 be considered inlining candidate.
6795 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6798 2018-02-12 Richard Biener <rguenther@suse.de>
6800 PR tree-optimization/84037
6801 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6802 matched stmts if we cannot swap the non-matched ones.
6804 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6806 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6807 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6808 _mm_maskz_scalef_round_ss): New intrinsics.
6809 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6810 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6811 __builtin_ia32_scalefss_round): Remove.
6812 (__builtin_ia32_scalefsd_mask_round,
6813 __builtin_ia32_scalefss_mask_round): New intrinsics.
6814 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6815 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6816 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6817 "<round_constraint>")): Changed to ...
6818 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6819 "<round_scalar_constraint>")): ... this.
6820 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6821 %0, %1, %2<round_op3>}"): Changed to ...
6822 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6823 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6824 %2<round_scalar_mask_op3>}"): ... this.
6825 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6827 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6829 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6830 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6831 (_mm_maskz_sqrt_round_ss): New intrinsics.
6832 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6833 (__builtin_ia32_sqrtsd_mask_round)
6834 (__builtin_ia32_sqrtss_mask_round): New builtins.
6835 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6836 (__builtin_ia32_sqrtss_round): Remove.
6837 (__builtin_ia32_sqrtsd_mask_round)
6838 (__builtin_ia32_sqrtss_mask_round): New builtins.
6839 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6840 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6841 ((match_operand:VF_128 1 "vector_operand"
6842 "xBm,<round_constraint>")): Changed to ...
6843 ((match_operand:VF_128 1 "vector_operand"
6844 "xBm,<round_scalar_constraint>")): ... this.
6845 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6846 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6847 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6848 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6849 %<iptr>1<round_scalar_mask_op3>}): ... this.
6850 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6851 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6853 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6856 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6857 Cast vec_cmpeq result to correct type.
6858 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6859 Cast vec_cmpgt result to correct type.
6861 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6863 * final.c (final_scan_insn_1): Renamed from...
6864 (final_scan_insn): ... this. New wrapper, to recover
6865 seen from the outermost call in recursive ones.
6866 * config/sparc/sparc.c (output_return): Drop seen from call.
6867 (output_sibcall): Likewise.
6868 * config/visium/visium.c (output_branch): Likewise.
6870 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6872 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6875 2018-02-10 Alan Modra <amodra@gmail.com>
6878 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6879 Specify LR as an input.
6881 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6884 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6885 remove_member_access_dummy_vars): New functions.
6886 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6887 lower_omp_1, execute_lower_omp): Use them.
6889 PR rtl-optimization/84308
6890 * shrink-wrap.c (spread_components): Release todo vector.
6892 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6894 PR rtl-optimization/57193
6895 * ira-color.c (struct allocno_color_data): Add member
6896 conflict_allocno_hard_prefs.
6897 (update_conflict_allocno_hard_prefs): New.
6898 (bucket_allocno_compare_func): Add a preference based on
6899 conflict_allocno_hard_prefs.
6900 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6901 (color_allocnos): Remove a dead code. Initiate
6902 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6904 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6907 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6908 constraint from =wa to wa. Avoid a subreg on the output operand,
6909 instead use a pseudo and subreg it in a move.
6910 (p9_xxbrd_<mode>): Changed to ...
6911 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6912 (p9_xxbrd_v2df): New expander.
6913 (p9_xxbrw_<mode>): Changed to ...
6914 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6915 (p9_xxbrw_v4sf): New expander.
6917 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6919 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6921 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6924 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6925 multiply in 32-bit mode.
6926 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6927 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6930 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6932 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6933 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6934 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6935 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6937 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6940 * dwarf2out.c (is_trivial_indirect_ref): New function.
6941 (dwarf2out_late_global_decl): Do not generate a location
6942 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6943 and that are not defined in the current unit.
6945 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6947 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6948 instead of a libcall for UNORDERED.
6950 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6953 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6954 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6956 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6959 * config/s390/s390.c (s390_set_current_function): Invoke
6960 s390_indirect_branch_settings also if fndecl didn't change.
6962 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6964 * config/rs6000/rs6000.md (blockage): Set length to zero.
6966 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6968 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6970 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6973 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6974 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6978 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6979 PARALLEL incoming that failed vt_get_decl_and_offset check.
6982 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6983 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6985 (get_variable_section): For decls in named .bss* sections pass true as
6986 second argument to bss_initializer_p.
6988 2018-02-09 Marek Polacek <polacek@redhat.com>
6989 Jakub Jelinek <jakub@redhat.com>
6992 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6993 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6994 Sync some changes from cxx_fold_indirect_ref.
6996 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6998 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7000 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7001 (BLOCK_INLINE_ENTRY_LABEL): New.
7002 (dwarf2out_var_location): Disregard inline entry markers.
7003 (inline_entry_data): New struct.
7004 (inline_entry_data_hasher): New hashtable type.
7005 (inline_entry_data_hasher::hash): New.
7006 (inline_entry_data_hasher::equal): New.
7007 (inline_entry_data_table): New variable.
7008 (add_high_low_attributes): Add DW_AT_entry_pc and
7009 DW_AT_GNU_entry_view attributes if a pending entry is found
7010 in inline_entry_data_table. Add old entry_pc attribute only
7011 if debug nonbinding markers are disabled.
7012 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7013 markers are enabled.
7014 (block_within_block_p, dwarf2out_inline_entry): New.
7015 (dwarf2out_finish): Check that no entries remained in
7016 inline_entry_data_table.
7017 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7018 (final_scan_insn, notice_source_line): Likewise.
7019 (rest_of_clean_state): Skip inline entry markers.
7020 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7022 * gimple.c (gimple_build_debug_inline_entry): New.
7023 * gimple.h (enum gimple_debug_subcode): Add
7024 GIMPLE_DEBUG_INLINE_ENTRY.
7025 (gimple_build_debug_inline_entry): Declare.
7026 (gimple_debug_inline_entry_p): New.
7027 (gimple_debug_nonbind_marker_p): Adjust.
7028 * insn-notes.def (INLINE_ENTRY): New.
7029 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7030 inline entry marker notes.
7031 (print_insn): Likewise.
7032 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7033 (INSN_DEBUG_MARKER_KIND): Likewise.
7034 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7035 * tree-inline.c (expand_call_inline): Build and insert
7036 debug_inline_entry stmt.
7037 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7038 inline entry blocks early, if nonbind markers are enabled.
7039 (dump_scope_block): Dump fragment info.
7040 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7041 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7042 (gimple_build_debug_inline_entry): New.
7043 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7044 Enable/disable inline entry points too.
7045 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7046 (DEBUG_INSN): Describe inline entry markers.
7048 * common.opt (gvariable-location-views): New.
7049 (gvariable-location-views=incompat5): New.
7050 * config.in: Rebuilt.
7051 * configure: Rebuilt.
7052 * configure.ac: Test assembler for view support.
7053 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7054 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7055 * dwarf2out.c (var_loc_view): New typedef.
7056 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7057 (dwarf2out_locviews_in_attribute): New.
7058 (dwarf2out_locviews_in_loclist): New.
7059 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7060 (enum dw_line_info_opcode): Add LI_adv_address.
7061 (struct dw_line_info_table): Add view.
7062 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7063 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7064 (zero_view_p): New variable.
7065 (ZERO_VIEW_P): New macro.
7066 (output_asm_line_debug_info): New.
7067 (struct var_loc_node): Add view.
7068 (add_AT_view_list, AT_loc_list): New.
7069 (add_var_loc_to_decl): Add view param. Test it against last.
7070 (new_loc_list): Add view params. Record them.
7071 (AT_loc_list_ptr): Handle loc and view lists.
7072 (view_list_to_loc_list_val_node): New.
7073 (print_dw_val): Handle dw_val_class_view_list.
7074 (size_of_die): Likewise.
7075 (value_format): Likewise.
7076 (loc_list_has_views): New.
7077 (gen_llsym): Set vl_symbol too.
7078 (maybe_gen_llsym, skip_loc_list_entry): New.
7079 (dwarf2out_maybe_output_loclist_view_pair): New.
7080 (output_loc_list): Output view list or entries too.
7081 (output_view_list_offset): New.
7082 (output_die): Handle dw_val_class_view_list.
7083 (output_dwarf_version): New.
7084 (output_compilation_unit_header): Use it.
7085 (output_skeleton_debug_sections): Likewise.
7086 (output_rnglists, output_line_info): Likewise.
7087 (output_pubnames, output_aranges): Update version comments.
7088 (output_one_line_info_table): Output view numbers in asm comments.
7089 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7090 Call maybe_gen_llsym.
7091 (loc_list_from_tree_1): Adjust.
7092 (add_AT_location_description): Create view list attribute if
7093 needed, check it's absent otherwise.
7094 (convert_cfa_to_fb_loc_list): Adjust.
7095 (maybe_emit_file): Call output_asm_line_debug_info for test.
7096 (dwarf2out_var_location): Reset views as needed. Precompute
7097 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7098 attribute. Set view.
7099 (new_line_info_table): Reset next view.
7100 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7101 (dwarf2out_source_line): Likewise. Output view resets and labels to
7102 the assembler, or select appropriate line info opcodes.
7103 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7104 (optimize_string_length): Catch it. Adjust.
7105 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7106 dw_val_class_view_list, and remove it if no longer needed.
7107 (hash_loc_list): Hash view numbers.
7108 (loc_list_hasher::equal): Compare them.
7109 (optimize_location_lists): Check whether a view list symbol is
7110 needed, and whether the locview attribute is present, and
7111 whether they match. Remove the locview attribute if no longer
7113 (index_location_lists): Call skip_loc_list_entry for test.
7114 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7115 Use output_dwarf_version.
7116 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7117 (struct dw_val_node): Add val_view_list.
7118 * final.c (SEEN_NEXT_VIEW): New.
7119 (set_next_view_needed): New.
7120 (clear_next_view_needed): New.
7121 (maybe_output_next_view): New.
7122 (final_start_function): Rename to...
7123 (final_start_function_1): ... this. Take pointer to FIRST,
7124 add SEEN parameter. Emit param bindings in the initial view.
7125 (final_start_function): Reintroduce SEEN-less interface.
7126 (final): Rename to...
7127 (final_1): ... this. Take SEEN parameter. Output final pending
7128 next view at the end.
7129 (final): Reintroduce seen-less interface.
7130 (final_scan_insn): Output pending next view before switching
7131 sections or ending a block. Mark the next view as needed when
7132 outputting variable locations. Notify debug backend of section
7133 changes, and of location view changes.
7134 (rest_of_handle_final): Adjust.
7135 * toplev.c (process_options): Autodetect value for debug variable
7136 location views option. Warn on incompat5 without -gdwarf-5.
7137 * doc/invoke.texi (gvariable-location-views): New.
7138 (gvariable-location-views=incompat5): New.
7139 (gno-variable-location-views): New.
7141 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7143 PR tree-optimization/84136
7144 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7145 that the result of find_edge is non-NULL.
7147 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7150 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7151 storing integer register in SImode. Fix cost of 256 and 512
7152 byte aligned SSE register store.
7154 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7156 * config/i386/i386.c (ix86_multiplication_cost): Fix
7157 multiplication cost for TARGET_AVX512DQ.
7159 2018-02-08 Marek Polacek <polacek@redhat.com>
7161 PR tree-optimization/84238
7162 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7165 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7167 PR tree-optimization/84265
7168 * tree-vect-stmts.c (vectorizable_store): Don't treat
7169 VMAT_CONTIGUOUS accesses as grouped.
7170 (vectorizable_load): Likewise.
7172 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7174 PR tree-optimization/81635
7175 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7176 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7177 (test_round_for_mask): New functions.
7178 (wide_int_cc_tests): Call test_round_for_mask.
7179 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7180 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7181 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7182 range returned by get_range_info.
7184 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7187 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7188 * symtab.c: Include builtins.h
7189 (symtab_node::output_to_lto_symbol_table_p): Move here
7190 from lto-streamer-out.c:output_symbol_p.
7191 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7192 (output_symbol_p): Move all logic to symtab.c
7193 (produce_symtab): Update.
7195 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7197 * config/s390/s390-opts.h (enum indirect_branch): Define.
7198 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7199 (s390_indirect_branch_via_thunk)
7200 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7201 (enum s390_indirect_branch_type): Define.
7202 * config/s390/s390.c (struct s390_frame_layout, struct
7203 machine_function): Remove.
7204 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7205 (indirect_branch_table_label_no, indirect_branch_table_name):
7207 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7208 (enum s390_indirect_branch_option): Define.
7209 (s390_return_addr_from_memory): New function.
7210 (s390_handle_string_attribute): New function.
7211 (s390_attribute_table): Add new attribute handler.
7212 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7213 (s390_indirect_branch_via_thunk): New function.
7214 (s390_indirect_branch_via_inline_thunk): New function.
7215 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7216 sibling call optimization for non z10 compiles.
7217 (s390_emit_call): Force indirect branch target to be a single
7218 register. Add r1 clobber for non-z10 compiles.
7219 (s390_emit_epilogue): Emit return jump via return_use expander.
7220 (s390_reorg): Handle JUMP_INSNs as execute targets.
7221 (s390_option_override_internal): Perform validity checks for the
7222 new command line options.
7223 (s390_indirect_branch_attrvalue): New function.
7224 (s390_indirect_branch_settings): New function.
7225 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7226 (s390_output_indirect_thunk_function): New function.
7227 (s390_code_end): Implement target hook.
7228 (s390_case_values_threshold): Implement target hook.
7229 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7231 * config/s390/s390.h (struct s390_frame_layout)
7232 (struct machine_function): Move here from s390.c.
7233 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7234 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7235 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7236 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7237 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7238 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7239 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7240 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7241 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7242 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7243 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7244 (mnemonic attribute): Add values which aren't recognized
7246 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7247 pattern for branch conversion. Fix mnemonic attribute.
7248 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7249 indirect branch via thunk if requested.
7250 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7251 ("*indirect_jump"): Disable for branch conversion using out of
7253 ("indirect_jump_via_thunk<mode>_z10")
7254 ("indirect_jump_via_thunk<mode>")
7255 ("indirect_jump_via_inlinethunk<mode>_z10")
7256 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7257 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7258 ("casesi_jump_via_inlinethunk<mode>_z10")
7259 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7260 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7261 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7262 ("*indirect2_jump"): Disable for branch conversion.
7263 ("casesi_jump"): Turn into expander and expand patterns for branch
7265 ("return_use"): New expander.
7266 ("*return"): Emit return via thunk and rename it to ...
7267 ("*return<mode>"): ... this one.
7268 * config/s390/s390.opt: Add new options and and enum for the
7271 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7273 * lra-constraints.c (match_reload): Unconditionally use
7274 gen_lowpart_SUBREG, rather than selecting between that
7275 and equivalent gen_rtx_SUBREG code.
7277 2018-02-08 Richard Biener <rguenther@suse.de>
7279 PR tree-optimization/84233
7280 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7281 changed flag instead of boguously re-using phi_inserted.
7283 2018-02-08 Martin Jambor <mjambor@suse.cz>
7285 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7286 static local variables.
7288 2018-02-08 Richard Biener <rguenther@suse.de>
7290 PR tree-optimization/84278
7291 * tree-vect-stmts.c (vectorizable_store): When looking for
7292 smaller vector types to perform grouped strided loads/stores
7293 make sure the mode is supported by the target.
7294 (vectorizable_load): Likewise.
7296 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7298 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7299 Increase LDP/STP opportunities by adding adjacent callee-saves.
7301 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7303 PR rtl-optimization/84068
7304 PR rtl-optimization/83459
7305 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7307 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7309 PR tree-optimization/84224
7310 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7311 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7314 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7317 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7318 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7319 count, remove one USE.
7321 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7323 * doc/install.texi (Configuration): Document the
7324 --with-long-double-format={ibm,ieee} PowerPC configuration
7328 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7329 Convert from define_expand to be define_insn_and_split. Rework
7330 float/double/_Float128 conversions to QI/HI/SImode to work with
7331 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7332 conversions to QI/HImode types did a store and then a load to
7333 truncate the value. For conversions to VSX registers, don't split
7334 the insn, instead emit the code directly. Use the code iterator
7335 any_fix to combine signed and unsigned conversions.
7336 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7337 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7338 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7339 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7340 (fix_<mode>di2_hw): Likewise.
7341 (fixuns_<mode>di2_hw): Likewise.
7342 (fix_<mode>si2_hw): Likewise.
7343 (fixuns_<mode>si2_hw): Likewise.
7344 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7345 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7346 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7347 fix<uns>_trunc<SFDF:mode>si2_p8.
7348 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7350 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7351 (fix<uns>_<mode>_mem): Likewise.
7352 (fctiw<u>z_<mode>_mem): Likewise.
7353 (fix<uns>_<mode>_mem): Likewise.
7354 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7355 the register allocator from doing a direct move to the GPRs to do
7356 a store, and instead use the ISA 3.0 store byte/half-word from
7357 vector register instruction. For IEEE 128-bit floating point,
7358 also optimize stores of 32-bit ints.
7359 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7361 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7363 * genextract.c (push_pathstr_operand): New function to support
7365 (walk_rtx): Call push_pathstr_operand.
7366 (print_path): Support [a-zA-Z].
7368 2018-02-07 Richard Biener <rguenther@suse.de>
7370 PR tree-optimization/84037
7371 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7372 (cse_and_gimplify_to_preheader): Declare.
7373 (vect_get_place_in_interleaving_chain): Likewise.
7374 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7376 (_loop_vec_info::~_loop_vec_info): Delete it.
7377 (cse_and_gimplify_to_preheader): New function.
7378 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7379 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7380 (vectorizable_load): Likewise. For grouped stores always base
7381 the IV on the first element.
7382 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7383 condition before gimplifying.
7385 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7387 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7388 *DIV_EXPR and *MOD_EXPR.
7390 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7393 * config/i386/i386.c (ix86_option_override_internal): Mask out
7394 the CF_SET bit when checking -fcf-protection.
7396 2018-02-07 Tom de Vries <tom@codesourcery.com>
7399 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7402 2018-02-07 Richard Biener <rguenther@suse.de>
7404 PR tree-optimization/84204
7405 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7408 PR tree-optimization/84205
7409 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7410 special-case isl_ast_op_zdiv_r.
7412 PR tree-optimization/84223
7413 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7414 Only add conditions from within the region.
7415 (gather_bbs::after_dom_children): Adjust.
7417 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7420 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7421 * config/avr/avr.md: Only post-reload split REG-REG moves if
7422 either register is GENERAL_REG_P.
7424 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7426 PR tree-optimization/84235
7427 * tree-ssa-scopedtables.c
7428 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7429 if the subtraction is performed in floating point type where NaNs are
7430 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7431 build 1. Formatting fix.
7433 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7436 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7437 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7438 and skip it regardless of bb boundaries. Use CALL_P macro,
7439 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7440 unnecessarily, formatting fix.
7442 2018-02-06 Michael Collison <michael.collison@arm.com>
7444 * config/arm/thumb2.md:
7445 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7446 (*thumb_mov_notscc): Ditto.
7448 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7451 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7452 unsigned_fix, not "s".
7454 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7456 * configure.ac (gcc_fn_eh_frame_ro): New function.
7457 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7458 correct .eh_frame permissions.
7459 * configure: Regenerate.
7461 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7463 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7466 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7468 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7469 Display warning message for -mno-speculate-indirect-jumps.
7471 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7473 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7475 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7477 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7479 PR tree-optimization/84225
7480 * tree-eh.c (find_trapping_overflow): Only call
7481 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7483 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7486 * config/i386/i386.c: Reimplement the check of possible options
7487 -mibt/-mshstk conbination. Change error messages.
7488 * doc/invoke.texi: Fix a typo: remove extra '='.
7490 2018-02-06 Marek Polacek <polacek@redhat.com>
7492 PR tree-optimization/84228
7493 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7495 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7498 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7499 emitted arch directives.
7500 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7501 __ARM_FEATURE_COPROC before changing architectures.
7503 2018-02-06 Richard Biener <rguenther@suse.de>
7505 * config/i386/i386.c (print_reg): Fix typo.
7506 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7508 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7510 * configure: Regenerate.
7512 2018-02-05 Martin Sebor <msebor@redhat.com>
7514 PR tree-optimization/83369
7515 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7518 2018-02-05 Martin Liska <mliska@suse.cz>
7520 * doc/invoke.texi: Cherry-pick upstream r323995.
7522 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7524 * ira.c (ira_init_register_move_cost): Adjust comment.
7526 2018-02-05 Martin Liska <mliska@suse.cz>
7528 PR gcov-profile/84137
7529 * doc/gcov.texi: Fix typo in documentation.
7531 2018-02-05 Martin Liska <mliska@suse.cz>
7533 PR gcov-profile/83879
7534 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7535 using dlopen functionality.
7537 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7539 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7540 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7541 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7542 _mm_maskz_range_round_ss): New intrinsics.
7543 (__builtin_ia32_rangesd128_round)
7544 (__builtin_ia32_rangess128_round): Remove.
7545 (__builtin_ia32_rangesd128_mask_round,
7546 __builtin_ia32_rangess128_mask_round): New builtins.
7547 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7548 __builtin_ia32_rangess128_round): Remove.
7549 (__builtin_ia32_rangesd128_mask_round,
7550 __builtin_ia32_rangess128_mask_round): New builtins.
7551 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7552 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7553 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7554 "<round_saeonly_constraint>")): Changed to ...
7555 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7556 "<round_saeonly_scalar_constraint>")): ... this.
7557 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7558 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7559 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7560 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7561 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7563 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7565 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7567 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7568 Remove all values except native, 8540 and 8548.
7570 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7572 * config/i386/i386.c (ix86_output_function_return): Pass
7573 INVALID_REGNUM, instead of -1, as invalid register number to
7574 indirect_thunk_name and output_indirect_thunk.
7576 2018-02-02 Julia Koval <julia.koval@intel.com>
7578 * config.gcc: Add -march=icelake.
7579 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7580 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7581 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7582 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7583 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7584 (processor_target_table): Add icelake.
7585 (ix86_option_override_internal): Handle new PTAs.
7586 (get_builtin_code_for_version): Handle icelake.
7587 (M_INTEL_COREI7_ICELAKE): New.
7588 (fold_builtin_cpu): Handle icelake.
7589 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7590 * doc/invoke.texi: Add -march=icelake.
7592 2018-02-02 Julia Koval <julia.koval@intel.com>
7594 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7595 to wide_int_bitmask.
7596 * wide-int-bitmask.h: New.
7598 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7601 * config/i386/i386.md: Replace Pmode with word_mode in
7602 builtin_setjmp_setup and builtin_longjmp to support x32.
7604 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7608 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7610 (rs6000_supported_cpu_names): New static variable.
7611 (linux_cpu_translation_table): Likewise.
7612 (elf_platform) <cpu>: Define new static variable and use it.
7613 Translate kernel AT_PLATFORM name to canonical name if needed.
7614 Error if platform name is unknown.
7616 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7619 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7621 2018-02-01 Jeff Law <law@redhat.com>
7624 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7625 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7626 the scratch if RELEASE_VIA_POP is false.
7627 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7628 If we have to save a temporary register, decrement SIZE appropriately.
7629 Pass new arguments to release_scratch_register_on_entry.
7630 (ix86_adjust_stack_and_probe): Likewise.
7631 (ix86_emit_probe_stack_range): Pass new arguments to
7632 release_scratch_register_on_entry.
7634 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7636 PR rtl-optimization/84157
7637 * combine.c (change_zero_ext): Use REG_P predicate in
7638 front of HARD_REGISTER_P predicate.
7640 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7642 * config/avr/avr.c (avr_option_override): Move disabling of
7643 -fdelete-null-pointer-checks to...
7644 * common/config/avr/avr-common.c (avr_option_optimization_table):
7647 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7649 PR tree-optimization/81635
7650 * tree-data-ref.c (split_constant_offset_1): For types that
7651 wrap on overflow, try to use range info to prove that wrapping
7654 2018-02-01 Renlin Li <renlin.li@arm.com>
7657 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7659 (aarch64_register_move_cost): Likewise.
7660 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7662 (REG_CLASS_NAMES): Likewise.
7663 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7664 TAILCALL_ADDR_REGS. Remove IP registers.
7665 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7667 2018-02-01 Richard Biener <rguenther@suse.de>
7669 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7670 for specifying RPO order and allow NULL for that.
7671 * domwalk.c (dom_walker::dom_walker): Likewise.
7672 (dom_walker::walk): Handle NULL RPO order.
7673 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7675 (rewrite_update_dom_walker): Likewise.
7676 (mark_def_dom_walker): Likewise.
7678 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7680 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7681 (aarch64_maybe_expand_sve_subreg_move): Declare.
7682 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7683 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7685 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7686 that are semantically a reverse operation.
7687 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7688 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7689 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7691 (aarch64_can_change_mode_class): For big-endian, forbid changes
7692 between two SVE modes if they have different element sizes.
7694 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7696 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7697 the TImode handling for big-endian targets.
7699 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7701 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7702 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7704 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7705 Remove BSWAP handing for big-endian targets and use the form of
7706 LD1RQ appropariate for the mode.
7708 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7710 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7711 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7714 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7717 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7718 check for operands that need to go through aarch64_sve_reload_be.
7720 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7722 PR tree-optimization/81661
7723 PR tree-optimization/84117
7724 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7725 * tree-eh.c: Include gimplify.h.
7726 (find_trapping_overflow, replace_trapping_overflow,
7727 rewrite_to_non_trapping_overflow): New functions.
7728 * tree-vect-loop.c: Include tree-eh.h.
7729 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7730 * tree-data-ref.c: Include tree-eh.h.
7731 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7733 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7735 PR rtl-optimization/84123
7736 * combine.c (change_zero_ext): Check if hard register satisfies
7737 can_change_dest_mode before calling gen_lowpart_SUBREG.
7739 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7742 * ira.c (ira_init_register_move_cost): Remove assert.
7744 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7746 PR rtl-optimization/84071
7747 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7748 * doc/tm.texi: Regenerate.
7750 2018-01-31 Richard Biener <rguenther@suse.de>
7752 PR tree-optimization/84132
7753 * tree-data-ref.c (analyze_miv_subscript): Properly
7754 check whether evolution_function_is_affine_multivariate_p
7755 before calling gcd_of_steps_may_divide_p.
7757 2018-01-31 Julia Koval <julia.koval@intel.com>
7760 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7761 * config/i386/i386.md (rdpid_rex64) New.
7762 (rdpid): Make 32bit only.
7764 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7767 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7768 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7770 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7773 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7775 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7777 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7779 PR rtl-optimization/84071
7780 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7781 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7783 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7785 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7786 (arc_attribute_table): Add 'aux' attribute.
7787 (arc_in_small_data_p): Consider aux like variables.
7788 (arc_is_aux_reg_p): New function.
7789 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7790 (arc_get_aux_arg): New function.
7791 (prepare_move_operands): Handle aux-register access.
7792 (arc_handle_aux_attribute): New function.
7793 * doc/extend.texi (ARC Variable attributes): Add subsection.
7795 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7797 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7798 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7799 (arc_attribute_table): Add 'uncached' attribute.
7800 (arc_print_operand): Print '.di' flag for uncached memory
7802 (arc_in_small_data_p): Do not consider for small data the uncached
7804 (arc_is_uncached_mem_p): New function.
7805 * config/arc/predicates.md (compact_store_memory_operand): Check
7806 for uncached memory accesses.
7807 (nonvol_nonimm_operand): Likewise.
7808 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7810 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7813 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7814 falign-loops=): Add Optimization flag.
7816 2018-01-30 Jeff Law <law@redhat.com>
7819 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7820 INT_REGISTERS_SAVED. Check it prior to calling
7821 get_scratch_register_on_entry.
7822 (ix86_adjust_stack_and_probe): Similarly.
7823 (ix86_emit_probe_stack_range): Similarly.
7824 (ix86_expand_prologue): Corresponding changes.
7826 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7829 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7830 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7832 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7835 * lra-constraints.c (curr_insn_transform): Process AND in the
7838 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7840 PR rtl-optimization/83986
7841 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7842 dependence against last_pending_memory_flush in addition to
7845 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7847 PR tree-optimization/81611
7848 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7851 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7854 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7857 2018-01-30 Richard Biener <rguenther@suse.de>
7858 Jakub Jelinek <jakub@redhat.com>
7860 PR tree-optimization/84111
7861 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7862 inner loops added during recursion, as they don't have up-to-date
7865 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7868 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7869 (can_inline_edge_by_limits_p): ... here.
7870 (can_early_inline_edge_p, check_callers,
7871 update_caller_keys, update_callee_keys, recursive_inlining,
7872 add_new_edges_to_heap, speculation_useful_p,
7873 inline_small_functions,
7874 inline_small_functions, flatten_function,
7875 inline_to_all_callers_1): Update.
7877 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7879 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7882 2018-01-30 Richard Biener <rguenther@suse.de>
7884 PR tree-optimization/83008
7885 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7886 invariant and constant vector uses in stmts when they need
7889 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7892 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7893 * configure: Regenerate.
7895 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7897 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7899 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7900 Use gen_rtx_REG rather than gen_lowpart.
7902 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7904 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7905 rather than 0 when creating partial subregs.
7907 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7909 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7912 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7915 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7916 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7917 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7918 flags. This restores the settings used before the 2017-07-24.
7919 Turning off pre increment/decrement/modify allows IVOPTS to
7920 optimize DF/SF loops where the index is an int.
7922 2018-01-29 Richard Biener <rguenther@suse.de>
7923 Kelvin Nilsen <kelvin@gcc.gnu.org>
7926 * tree-vect-stmts.c (vectorizable_call): Don't call
7927 targetm.vectorize_builtin_md_vectorized_function if callee is
7930 2018-01-22 Carl Love <cel@us.ibm.com>
7932 * doc/extend.tex: Fix typo in second arg in
7933 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7935 2018-01-29 Richard Biener <rguenther@suse.de>
7937 PR tree-optimization/84086
7938 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7939 (flush_ssaname_freelist): When SSA names were released reset
7940 the SCEV hash table.
7942 2018-01-29 Richard Biener <rguenther@suse.de>
7944 PR tree-optimization/84057
7945 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7946 removed paths when removing edges.
7948 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7950 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7951 -mfunction-return=@var{choice}.
7953 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7956 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7958 (layout::print_source_line): Likewise.
7959 (test_get_line_width_without_trailing_whitespace): Add test cases.
7961 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7964 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7967 2018-01-26 Jim Wilson <jimw@sifive.com>
7969 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7971 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7974 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7976 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7977 and CMP + SUB-immediate -> SUBS.
7979 2018-01-26 Martin Sebor <msebor@redhat.com>
7981 PR tree-optimization/83896
7982 * tree-ssa-strlen.c (get_string_len): Rename...
7983 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7984 Avoid assuming length is constant.
7985 (handle_char_store): Use HOST_WIDE_INT for string length.
7987 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7990 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7991 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7993 2018-01-26 Richard Biener <rguenther@suse.de>
7995 PR rtl-optimization/84003
7996 * dse.c (record_store): Only record redundant stores when
7997 the earlier store aliases at least all accesses the later one does.
7999 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8001 PR rtl-optimization/83985
8002 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8003 REG_CFA_RESTORE insns.
8004 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8005 REG_CFA_RESTORE insns here.
8008 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8009 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8011 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8013 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8014 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8015 (arc_init): Likewise.
8016 (arc_override_options): Likewise.
8017 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8019 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8021 * config/arc/arc.h (TARGET_DBNZ): Define.
8022 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8023 properly set the tune attribute.
8024 (dbnz): Use TARGET_DBNZ guard.
8025 * config/arc/arc.opt (mtune): Add core3 option.
8027 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8029 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8030 recognize new pic like addresses.
8031 (arc_delegitimize_address): Clean up.
8033 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8035 * config/arc/arc-arches.def: Option mrf16 valid for all
8037 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8038 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8039 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8040 * config/arc/arc-tables.opt: Regenerate.
8041 * config/arc/arc.c (arc_conditional_register_usage): Handle
8042 reduced register file case.
8043 (arc_file_start): Set must have build attributes.
8044 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8046 * config/arc/arc.opt (mrf16): Add new option.
8047 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8048 * config/arc/genmultilib.awk: Handle new mrf16 option.
8049 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8050 * config/arc/t-multilib: Regenerate.
8051 * doc/invoke.texi (ARC Options): Document mrf16 option.
8053 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8055 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8056 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8057 (arc_attribute_table): Add 'secure_call' attribute.
8058 (arc_print_operand): Print secure call operand.
8059 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8061 (arc_is_secure_call_p): New function. * config/arc/arc.md
8062 (call_i): Add support for sjli instruction.
8063 (call_value_i): Likewise.
8064 * config/arc/constraints.md (Csc): New constraint.
8066 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8067 John Eric Martin <John.Martin@emmicro-us.com>
8069 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8070 * config/arc/arc.c (_arc_jli_section): New struct.
8071 (arc_jli_section): New type.
8072 (rc_jli_sections): New static variable.
8073 (arc_handle_jli_attribute): New function.
8074 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8075 (arc_file_end): New function.
8076 (TARGET_ASM_FILE_END): Define.
8077 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8078 (arc_add_jli_section): New function.
8079 (jli_call_scan): Likewise.
8080 (arc_reorg): Call jli_call_scan.
8081 (arc_output_addsi): Remove 'S' from printing asm operand.
8082 (arc_is_jli_call_p): New function.
8083 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8085 (movhi_insn): Likewise.
8086 (movsi_insn): Likewise.
8087 (movsi_set_cc_insn): Likewise.
8088 (loadqi_update): Likewise.
8089 (load_zeroextendqisi_update): Likewise.
8090 (load_signextendqisi_update): Likewise.
8091 (loadhi_update): Likewise.
8092 (load_zeroextendhisi_update): Likewise.
8093 (load_signextendhisi_update): Likewise.
8094 (loadsi_update): Likewise.
8095 (loadsf_update): Likewise.
8096 (movsicc_insn): Likewise.
8097 (bset_insn): Likewise.
8098 (bxor_insn): Likewise.
8099 (bclr_insn): Likewise.
8100 (bmsk_insn): Likewise.
8101 (bicsi3_insn): Likewise.
8102 (cmpsi_cc_c_insn): Likewise.
8103 (movsi_ne): Likewise.
8104 (movsi_cond_exec): Likewise.
8105 (clrsbsi2): Likewise.
8116 (call_i): Remove 'S' asm letter, add jli instruction.
8117 (call_value_i): Likewise.
8118 * config/arc/arc.op (mjli-always): New option.
8119 * config/arc/constraints.md (Cji): New constraint.
8120 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8122 (subsf3_fpx): Likewise.
8123 (mulsf3_fpx): Likewise.
8124 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8126 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8128 * doc/invoke.texi (ARC): Document mjli-always option.
8130 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8132 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8133 avoid addition with 0 and use incw and decw where possible.
8135 2018-01-26 Richard Biener <rguenther@suse.de>
8137 PR tree-optimization/81082
8138 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8139 association if it requires casting to unsigned.
8140 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8141 from fold_plusminus_mult_expr to catch important cases late when
8142 range info is available.
8144 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8146 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8147 * configure.ac (hidden_linkonce): New test.
8148 * configure: Regenerate.
8149 * config.in: Regenerate.
8151 2018-01-26 Julia Koval <julia.koval@intel.com>
8153 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8154 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8155 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8156 _mm_mask_bitshuffle_epi64_mask): Fix type.
8157 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8158 USI_FTYPE_V4DI_V4DI_USI): Remove.
8159 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8160 __builtin_ia32_vpshufbitqmb256_mask,
8161 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8162 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8163 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8165 2018-01-26 Alan Modra <amodra@gmail.com>
8168 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8169 UNSPEC_VBPERMQ. Sort other unspecs.
8171 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8173 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8175 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8178 * predict.c (drop_profile): Do not push/pop cfun; update also
8180 (handle_missing_profiles): Fix logic looking for zero profiles.
8182 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8185 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8186 on functions with #pragma omp declare simd or functions with simd
8188 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8189 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8190 Remove trailing \n from warning_at calls.
8192 2018-01-25 Tom de Vries <tom@codesourcery.com>
8195 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8196 for neutered workers.
8198 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8201 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8202 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8204 2018-01-24 Jeff Law <law@redhat.com>
8207 * i386.c (get_probe_interval): Move to earlier point.
8208 (ix86_compute_frame_layout): If -fstack-clash-protection and
8209 the frame is larger than the probe interval, then use pushes
8210 to save registers rather than reg->mem moves.
8211 (ix86_expand_prologue): Remove conditional for int_registers_saved
8214 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8217 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8218 min/max for never referenced object.
8220 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8223 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8225 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8226 attributes from DECL_ATTRIBUTES (decl) without affecting
8227 DECL_ATTRIBUTES (current_function_decl).
8228 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8229 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8231 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8233 PR tree-optimization/83979
8234 * fold-const.c (fold_comparison): Use constant_boolean_node
8235 instead of boolean_{true,false}_node.
8237 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8239 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8242 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8244 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8245 Simplify the clause that sets the length attribute.
8246 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8247 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8248 clause that sets the length attribute.
8249 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8251 2018-01-24 Tom de Vries <tom@codesourcery.com>
8254 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8255 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8256 Add strict parameter.
8257 (prevent_branch_around_nothing): Insert dummy insn between branch to
8258 label and label with no ptx insn inbetween.
8259 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8261 2018-01-24 Tom de Vries <tom@codesourcery.com>
8264 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8265 for neutered threads in warp.
8266 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8268 2018-01-24 Richard Biener <rguenther@suse.de>
8270 PR tree-optimization/83176
8271 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8274 2018-01-24 Richard Biener <rguenther@suse.de>
8276 PR tree-optimization/82819
8277 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8278 code generating pluses that are no-ops in the target precision.
8280 2018-01-24 Richard Biener <rguenther@suse.de>
8283 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8285 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8287 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8288 to merge probabilities.
8289 * predict.c (probably_never_executed): Also mark as cold functions
8290 with global 0 profile and guessed local profile.
8291 * profile-count.c (profile_probability::combine_with_count): New
8293 * profile-count.h (profile_probability::operator*,
8294 profile_probability::operator*=, profile_probability::operator/,
8295 profile_probability::operator/=): Reduce precision to adjusted
8296 and set value to guessed on contradictory divisions.
8297 (profile_probability::combine_with_freq): Remove.
8298 (profile_probability::combine_wiht_count): Declare.
8299 (profile_count::force_nonzero):: Set to adjusted.
8300 (profile_count::probability_in):: Set quality to adjusted.
8301 * tree-ssa-tail-merge.c (replace_block_by): Use
8304 2018-01-23 Andrew Waterman <andrew@sifive.com>
8305 Jim Wilson <jimw@sifive.com>
8307 * config/riscv/riscv.c (riscv_stack_boundary): New.
8308 (riscv_option_override): Set riscv_stack_boundary. Handle
8309 riscv_preferred_stack_boundary_arg.
8310 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8311 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8312 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8313 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8314 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8315 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8317 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8320 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8321 of struct ix86_frame.
8322 (ix86_expand_epilogue): Likewise. Add a local variable for
8323 the reg_save_offset field in struct ix86_frame.
8325 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8327 PR tree-optimization/82604
8328 * tree-loop-distribution.c (enum partition_kind): New enum item
8329 PKIND_PARTIAL_MEMSET.
8330 (partition_builtin_p): Support above new enum item.
8331 (generate_code_for_partition): Ditto.
8332 (compute_access_range): Differentiate cases that equality can be
8333 proven at all loops, the innermost loops or no loops.
8334 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8335 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8336 (finalize_partitions, distribute_loop): Don't fuse partition of
8337 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8338 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8341 2018-01-23 Martin Liska <mliska@suse.cz>
8343 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8344 order to ignore the predictor.
8345 (PRED_POLYMORPHIC_CALL): Likewise.
8346 (PRED_RECURSIVE_CALL): Likewise.
8348 2018-01-23 Martin Liska <mliska@suse.cz>
8350 * tree-profile.c (tree_profiling): Print function header to
8351 aware reader which function we are working on.
8352 * value-prof.c (gimple_find_values_to_profile): Do not print
8353 not interesting value histograms.
8355 2018-01-23 Martin Liska <mliska@suse.cz>
8357 * profile-count.h (enum profile_quality): Add
8358 profile_uninitialized as the first value. Do not number values
8359 as they are zero based.
8360 (profile_count::verify): Update sanity check.
8361 (profile_probability::verify): Likewise.
8363 2018-01-23 Nathan Sidwell <nathan@acm.org>
8365 * doc/invoke.texi (ffor-scope): Deprecate.
8367 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8369 PR tree-optimization/83510
8370 * domwalk.c (set_all_edges_as_executable): New function.
8371 (dom_walker::dom_walker): Convert bool param
8372 "skip_unreachable_blocks" to enum reachability. Move setup of
8373 edge flags to set_all_edges_as_executable and only do it when
8374 reachability is REACHABLE_BLOCKS.
8375 * domwalk.h (enum dom_walker::reachability): New enum.
8376 (dom_walker::dom_walker): Convert bool param
8377 "skip_unreachable_blocks" to enum reachability.
8378 (set_all_edges_as_executable): New decl.
8379 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8380 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8382 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8383 but converting true to REACHABLE_BLOCKS.
8384 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8386 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8387 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8388 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8390 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8391 if check_all_array_refs will be called.
8393 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8395 * tree.c (selftest::test_location_wrappers): Add more test
8398 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8400 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8401 (selftest::test_bit_in_range): Likewise.
8403 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8406 * doc/sourcebuild.texi (vect_float): Say that the selector
8407 only describes the situation when -funsafe-math-optimizations is on.
8408 (vect_float_strict): Document.
8410 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8412 PR tree-optimization/83965
8413 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8414 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8415 instead of checking only for a reduction.
8416 (vect_recog_widen_sum_pattern): Likewise.
8418 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8420 * predict.c (probably_never_executed): Only use precise profile info.
8421 (compute_function_frequency): Skip after inlining hack since we now
8422 have quality checking.
8424 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8426 * profile-count.h (profile_probability::very_unlikely,
8427 profile_probability::unlikely, profile_probability::even): Set
8428 precision to guessed.
8430 2018-01-23 Richard Biener <rguenther@suse.de>
8432 PR tree-optimization/83963
8433 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8434 Properly terminate dominator walk when crossing the exit edge not
8435 when visiting its source block.
8437 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8440 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8441 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8443 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8445 PR tree-optimization/83957
8446 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8447 semicolon after for body surrounded by braces.
8449 PR tree-optimization/83081
8450 * profile-count.h (profile_probability::split): New method.
8451 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8452 Use profile_probability::split.
8453 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8454 when splitting a single conditional jump into 2.
8456 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8458 PR tree-optimization/69452
8459 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8462 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8464 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8465 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8466 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8468 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8470 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8471 * config/rl78/rl78.md (movdi): New define_expand.
8472 * config/rl78/rl78.c (rl78_split_movdi): New function.
8474 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8477 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8479 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8480 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8481 128-bit to produce an UNSPEC move to get the double word with the
8482 signbit and then a shift directly to do signbit.
8483 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8484 implementation with a new version that just does either a direct
8485 move or a regular move. Move memory interface to separate insns.
8486 Move insns so they are next to the expander.
8487 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8488 with signbit move. Split big and little endian case.
8489 (signbit<mode>2_dm_mem_le): Likewise.
8490 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8491 (signbit<mode>2_dm2): Likewise.
8493 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8495 * config/rl78/rl78.md (anddi3): New define_expand.
8497 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8499 * config/rl78/rl78.md (umindi3): New define_expand.
8501 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8503 * config/rl78/rl78.md (smindi3): New define_expand.
8505 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8507 * config/rl78/rl78.md (smaxdi3): New define_expand.
8509 2018-01-22 Carl Love <cel@us.ibm.com>
8511 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8512 LVX_V1TI): Add macro expansion.
8513 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8514 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8515 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8516 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8517 Change check to determine if the instruction is a byte reversing
8518 entry. Fix typo in comment.
8519 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8520 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8521 Add def_builtin calls for new builtins.
8522 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8523 Add define_insn expansion.
8525 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8527 * config/rl78/rl78.md (umaxdi3): New define_expand.
8529 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8531 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8532 for non-QImode registers.
8534 2018-01-22 Richard Biener <rguenther@suse.de>
8536 PR tree-optimization/83963
8537 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8538 including the loop exit block.
8539 (scop_detection::merge_sese): Likewise.
8540 (scop_detection::add_scop): Do it here instead.
8542 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8544 * doc/sourcebuild.texi (arm_softfloat): Document.
8546 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8549 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8550 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8551 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8553 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8554 David Edelsohn <dje.gcc@gmail.com>
8557 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8558 Change "crset eq" to "crset 2".
8559 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8560 (*call_indirect_aix<mode>_nospec): Likewise.
8561 (*call_value_indirect_aix<mode>_nospec): Likewise.
8562 (*call_indirect_elfv2<mode>_nospec): Likewise.
8563 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8564 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8565 change assembly output from . to $.
8566 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8567 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8568 (*tablejump<mode>_internal1_nospec): Likewise.
8570 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8573 * config/sh/sh_optimize_sett_clrt.cc:
8574 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8576 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8578 PR tree-optimization/83940
8579 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8580 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8581 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8582 use it to pass back the definition type.
8583 (vect_check_store_rhs): Likewise rhs_dt_out.
8584 (vect_build_gather_load_calls): Add a mask_dt argument and use
8585 it instead of a call to vect_is_simple_use.
8586 (vectorizable_store): Update calls to vect_check_load_store_mask
8587 and vect_check_store_rhs. Use the dt returned by the latter instead
8588 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8589 instead of calls to vect_is_simple_use. Pass the scalar rather
8590 than the vector operand to vect_is_simple_use when handling
8591 second and subsequent copies of an rhs value.
8592 (vectorizable_load): Update calls to vect_check_load_store_mask
8593 and vect_build_gather_load_calls. Use the cached mask_dt and
8594 gs_info.offset_dt instead of calls to vect_is_simple_use.
8596 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8599 * tree-emutls.c: Include gimplify.h.
8600 (lower_emutls_2): New function.
8601 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8602 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8603 it before further processing.
8606 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8607 UINTVAL (trueop1) instead of INTVAL (op1).
8609 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8613 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8614 INCOMING_FRAME_SP_OFFSET if not defined.
8615 (scan_trace): Add ENTRY argument. If true and
8616 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8617 emit a note to adjust the CFA offset.
8618 (create_cfi_notes): Adjust scan_trace callers.
8619 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8620 INCOMING_FRAME_SP_OFFSET in the CIE.
8621 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8622 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8624 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8625 * doc/tm.texi: Regenerated.
8627 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8629 PR rtl-optimization/83147
8630 * lra-constraints.c (remove_inheritance_pseudos): Use
8631 lra_substitute_pseudo_within_insn.
8633 2018-01-19 Tom de Vries <tom@codesourcery.com>
8634 Cesar Philippidis <cesar@codesourcery.com>
8637 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8639 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8642 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8643 spaces for function labels.
8645 2018-01-19 Martin Liska <mliska@suse.cz>
8647 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8648 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8649 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8650 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8651 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8652 (PRED_CONST_RETURN): Change from 69 to 65.
8653 (PRED_NULL_RETURN): Change from 91 to 71.
8654 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8655 (PRED_LOOP_GUARD): Change from 66 to 73.
8657 2018-01-19 Martin Liska <mliska@suse.cz>
8659 * predict.c (predict_insn_def): Add new assert.
8660 (struct branch_predictor): Change type to signed integer.
8661 (test_prediction_value_range): Amend test to cover
8663 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8664 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8665 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8666 (PRED_LOOP_IV_COMPARE): Likewise.
8667 * predict.h (PROB_UNINITIALIZED): Define new constant.
8669 2018-01-19 Martin Liska <mliska@suse.cz>
8671 * predict.c (dump_prediction): Add new format for
8672 analyze_brprob.py script which is enabled with -details
8674 * profile-count.h (precise_p): New function.
8676 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8678 PR tree-optimization/83922
8679 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8680 there are no statements that need masking.
8681 (vect_active_double_reduction_p): New function.
8682 (vect_analyze_loop_operations): Use it when handling phis that
8683 are not in the loop header.
8685 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8687 PR tree-optimization/83914
8688 * tree-vect-loop.c (vectorizable_induction): Don't convert
8689 init_expr or apply the peeling adjustment for inductions
8690 that are nested within the vectorized loop.
8692 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8694 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8697 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8701 * function.h (gimplify_parameters): Add gimple_seq * argument.
8702 * function.c: Include gimple.h and options.h.
8703 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8704 for the added local temporaries if needed.
8705 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8706 if there are any parameter cleanups, wrap whole body into a
8707 try/finally with the cleanups.
8709 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8712 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8713 Use GET_MODE_CLASS for scalar floating point.
8715 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8719 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8720 Fix call of call_cgraph_insertion_hooks.
8722 2018-01-18 Martin Sebor <msebor@redhat.com>
8724 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8726 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8729 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8732 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8735 * common.opt: (-ffile-prefix-map): New option.
8736 * opts.c (common_handle_option): Defer it.
8737 * opts-global.c (handle_common_deferred_options): Handle it.
8738 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8739 * file-prefix-map.h: New file.
8740 (remap_debug_filename, add_debug_prefix_map): ...here.
8741 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8742 * final.c (debug_prefix_map, add_debug_prefix_map
8743 remap_debug_filename): Move to...
8744 * file-prefix-map.c: New file.
8745 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8746 generalize, get rid of alloca(), use strrchr() instead of strchr().
8747 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8748 Implement in terms of add_prefix_map().
8749 (remap_macro_filename, remap_debug_filename): Implement in term of
8751 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8752 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8753 * dbxout.c: Include file-prefix-map.h.
8754 * varasm.c: Likewise.
8755 * vmsdbgout.c: Likewise.
8756 * xcoffout.c: Likewise.
8757 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8758 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8759 * doc/invoke.texi (-ffile-prefix-map): Document.
8760 (-fdebug-prefix-map): Update description.
8762 2018-01-18 Martin Liska <mliska@suse.cz>
8764 * config/i386/i386.c (indirect_thunk_name): Document that also
8766 (output_indirect_thunk): Document why both instructions
8767 (pause and lfence) are generated.
8769 2018-01-18 Richard Biener <rguenther@suse.de>
8771 PR tree-optimization/83887
8772 * graphite-scop-detection.c
8773 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8774 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8775 (scop_detection::merge_sese): Re-implement with a flood-fill
8776 algorithm that properly finds a SESE region if it exists.
8778 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8781 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8782 pointer_diff optimizations use view_convert instead of convert.
8784 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8786 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8787 Generate different code for -mno-speculate-indirect-jumps.
8788 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8789 (*call_indirect_aix<mode>): Disable for
8790 -mno-speculate-indirect-jumps.
8791 (*call_indirect_aix<mode>_nospec): New define_insn.
8792 (*call_value_indirect_aix<mode>): Disable for
8793 -mno-speculate-indirect-jumps.
8794 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8795 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8796 -mno-speculate-indirect-jumps.
8797 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8799 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8801 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8802 long double type, set the flags for noting the default long double
8803 type, even if we don't pass or return a long double type.
8805 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8808 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8811 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8813 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8814 support for merge[hl].
8815 (fold_mergehl_helper): New helper function.
8816 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8817 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8818 (altivec_vmrglw_direct): Add xxmrglw insn.
8820 2018-01-17 Andrew Waterman <andrew@sifive.com>
8822 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8823 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8825 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8828 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8829 call the lto_location_cache before reading the
8830 DECL_SOURCE_LOCATION of the types.
8832 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8833 Richard Sandiford <richard.sandiford@linaro.org>
8835 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8836 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8837 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8838 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8839 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8841 * config/aarch64/constraints.md (aarch64_movti_operand):
8843 * config/aarch64/predicates.md (Uti): Add new constraint.
8845 2018-01-17 Carl Love <cel@us.ibm.com>
8847 * config/rs6000/vsx.md (define_expand xl_len_r,
8848 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8849 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8851 (define_expand, define_insn): Move the shift left from the
8852 define_insn to the define_expand for lxvl and stxvl instructions.
8853 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8854 and XL_LEN_R definitions to PURE.
8856 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8858 * config/i386/i386.c (indirect_thunk_name): Declare regno
8859 as unsigned int. Compare regno with INVALID_REGNUM.
8860 (output_indirect_thunk): Ditto.
8861 (output_indirect_thunk_function): Ditto.
8862 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8863 in the call to output_indirect_thunk_function.
8865 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8868 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8869 rather than the size of inner_type to determine the stack slot size
8870 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8872 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8875 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8878 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8880 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8881 endian Linux systems to optionally enable multilibs for selecting
8882 the long double type if the user configured an explicit type.
8883 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8884 have no long double multilibs if not defined.
8885 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8886 warn if the user used -mabi={ieee,ibm}longdouble and we built
8887 multilibs for long double.
8888 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8889 appropriate multilib option.
8890 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8892 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8893 for building long double multilibs.
8894 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8896 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8898 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8901 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8903 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8906 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8909 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8912 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8914 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8915 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8917 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8919 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8920 different rtl trees depending on TARGET_64BIT.
8921 (rs6000_gen_lvx): Likewise.
8923 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8925 * config/visium/visium.md (nop): Tweak comment.
8926 (hazard_nop): Likewise.
8928 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8930 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8931 -mspeculate-indirect-jumps.
8932 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8933 for -mno-speculate-indirect-jumps.
8934 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8935 (*call_value_indirect_elfv2<mode>): Disable for
8936 -mno-speculate-indirect-jumps.
8937 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8938 (indirect_jump): Emit different RTL for
8939 -mno-speculate-indirect-jumps.
8940 (*indirect_jump<mode>): Disable for
8941 -mno-speculate-indirect-jumps.
8942 (*indirect_jump<mode>_nospec): New define_insn.
8943 (tablejump): Emit different RTL for
8944 -mno-speculate-indirect-jumps.
8945 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8946 (tablejumpsi_nospec): New define_expand.
8947 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8948 (tablejumpdi_nospec): New define_expand.
8949 (*tablejump<mode>_internal1): Disable for
8950 -mno-speculate-indirect-jumps.
8951 (*tablejump<mode>_internal1_nospec): New define_insn.
8952 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8955 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8957 * caller-save.c (insert_save): Drop unnecessary parameter. All
8960 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8961 Richard Biener <rguenth@suse.de>
8964 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8965 return early, inline manually is_gimple_sizepos. Make sure if we
8966 call gimplify_expr we don't end up with a gimple constant.
8967 * tree.c (variably_modified_type_p): Don't return true for
8968 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8969 * gimplify.h (is_gimple_sizepos): Remove.
8971 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8973 PR tree-optimization/83857
8974 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8975 vectorizable_live_operation for pure SLP statements.
8976 (vectorizable_live_operation): Handle PHIs.
8978 2018-01-16 Richard Biener <rguenther@suse.de>
8980 PR tree-optimization/83867
8981 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8982 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8984 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8987 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8988 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8989 If off is not INTEGER_CST, issue a may not be aligned warning
8990 rather than isn't aligned. Use isn%'t rather than isn't.
8991 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8993 <case MULT_EXPR>: Improve the case when bottom and one of the
8994 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8995 operand, in that case check if the other operand is multiple of
8996 bottom divided by the INTEGER_CST operand.
8998 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9001 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9002 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9003 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9004 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9005 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9006 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9007 (pa_function_arg_size): New function.
9009 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9011 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9012 in a separate statement.
9014 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9016 PR tree-optimization/83847
9017 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9018 group gathers and scatters.
9020 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9022 PR rtl-optimization/86620
9023 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9025 PR rtl-optimization/83213
9026 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9027 to last if both are JUMP_INSNs.
9029 PR tree-optimization/83843
9030 * gimple-ssa-store-merging.c
9031 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9032 store_immediate_info for bswap/nop orig_stores.
9034 2018-01-15 Andrew Waterman <andrew@sifive.com>
9036 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9038 <UDIV>: Increase cost if !TARGET_DIV.
9040 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9042 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9043 (define_attr "cr_logical_3op"): New.
9044 (cceq_ior_compare): Adjust.
9045 (cceq_ior_compare_complement): Adjust.
9046 (*cceq_rev_compare): Adjust.
9047 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9048 (is_cracked_insn): Adjust.
9049 (insn_must_be_first_in_group): Adjust.
9050 * config/rs6000/40x.md: Adjust.
9051 * config/rs6000/440.md: Adjust.
9052 * config/rs6000/476.md: Adjust.
9053 * config/rs6000/601.md: Adjust.
9054 * config/rs6000/603.md: Adjust.
9055 * config/rs6000/6xx.md: Adjust.
9056 * config/rs6000/7450.md: Adjust.
9057 * config/rs6000/7xx.md: Adjust.
9058 * config/rs6000/8540.md: Adjust.
9059 * config/rs6000/cell.md: Adjust.
9060 * config/rs6000/e300c2c3.md: Adjust.
9061 * config/rs6000/e500mc.md: Adjust.
9062 * config/rs6000/e500mc64.md: Adjust.
9063 * config/rs6000/e5500.md: Adjust.
9064 * config/rs6000/e6500.md: Adjust.
9065 * config/rs6000/mpc.md: Adjust.
9066 * config/rs6000/power4.md: Adjust.
9067 * config/rs6000/power5.md: Adjust.
9068 * config/rs6000/power6.md: Adjust.
9069 * config/rs6000/power7.md: Adjust.
9070 * config/rs6000/power8.md: Adjust.
9071 * config/rs6000/power9.md: Adjust.
9072 * config/rs6000/rs64.md: Adjust.
9073 * config/rs6000/titan.md: Adjust.
9075 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9077 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9078 ix86_indirect_branch_register logic.
9080 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9082 * config/i386/constraints.md (Bs): Update
9083 ix86_indirect_branch_register check. Don't check
9084 ix86_indirect_branch_register with GOT_memory_operand.
9086 * config/i386/predicates.md (GOT_memory_operand): Don't check
9087 ix86_indirect_branch_register here.
9088 (GOT32_symbol_operand): Likewise.
9090 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9092 * config/i386/predicates.md (constant_call_address_operand):
9093 Rewrite ix86_indirect_branch_register logic.
9094 (sibcall_insn_operand): Likewise.
9096 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9098 * config/i386/constraints.md (Bs): Replace
9099 ix86_indirect_branch_thunk_register with
9100 ix86_indirect_branch_register.
9102 * config/i386/i386.md (indirect_jump): Likewise.
9103 (tablejump): Likewise.
9104 (*sibcall_memory): Likewise.
9105 (*sibcall_value_memory): Likewise.
9106 Peepholes of indirect call and jump via memory: Likewise.
9107 * config/i386/i386.opt: Likewise.
9108 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9109 (GOT_memory_operand): Likewise.
9110 (call_insn_operand): Likewise.
9111 (sibcall_insn_operand): Likewise.
9112 (GOT32_symbol_operand): Likewise.
9114 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9117 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9118 type rather than type addr's type points to.
9119 (expand_omp_atomic_mutex): Likewise.
9120 (expand_omp_atomic): Likewise.
9122 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9125 * config/i386/i386.c (output_indirect_thunk_function): Use
9126 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9127 for __x86_return_thunk.
9129 2018-01-15 Richard Biener <rguenther@suse.de>
9132 * expmed.c (extract_bit_field_1): Fix typo.
9134 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9137 * config/arm/iterators.md (VF): New mode iterator.
9138 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9139 Remove integer-related logic from pattern.
9140 (neon_vabd<mode>_3): Likewise.
9142 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9145 * common.opt (fstrict-overflow): No longer an alias.
9146 (fwrapv-pointer): New option.
9147 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9148 also for pointer types based on flag_wrapv_pointer.
9149 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9150 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9151 opts->x_flag_wrapv got set.
9152 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9153 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9154 POINTER_TYPE_OVERFLOW_UNDEFINED.
9155 * match.pd: Likewise in address comparison pattern.
9156 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9158 2018-01-15 Richard Biener <rguenther@suse.de>
9161 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9162 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9163 Reset type names to their identifier if their TYPE_DECL doesn't
9164 have linkage (and thus is used for ODR and devirt).
9165 (save_debug_info_for_decl): Remove.
9166 (save_debug_info_for_type): Likewise.
9167 (add_tree_to_fld_list): Adjust.
9168 * tree-pretty-print.c (dump_generic_node): Make dumping of
9169 type names more robust.
9171 2018-01-15 Richard Biener <rguenther@suse.de>
9173 * BASE-VER: Bump to 8.0.1.
9175 2018-01-14 Martin Sebor <msebor@redhat.com>
9178 * builtins.c (check_access): Avoid warning when the no-warning bit
9181 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9183 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9184 * ira-color (allocno_hard_regs_compare): Likewise.
9186 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9189 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9190 Use .pushsection/.popsection.
9192 2018-01-14 Martin Sebor <msebor@redhat.com>
9195 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9197 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9199 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9200 entry from extra_headers.
9201 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9202 extra_headers, make the list bitwise identical to the i?86-*-* one.
9204 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9206 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9207 -mcmodel=large with -mindirect-branch=thunk,
9208 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9209 -mfunction-return=thunk-extern.
9210 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9211 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9212 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9214 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9216 * config/i386/i386.c (print_reg): Print the name of the full
9217 integer register without '%'.
9218 (ix86_print_operand): Handle 'V'.
9219 * doc/extend.texi: Document 'V' modifier.
9221 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9223 * config/i386/constraints.md (Bs): Disallow memory operand for
9224 -mindirect-branch-register.
9226 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9227 (GOT_memory_operand): Likewise.
9228 (call_insn_operand): Likewise.
9229 (sibcall_insn_operand): Likewise.
9230 (GOT32_symbol_operand): Likewise.
9231 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9232 for -mindirect-branch-register.
9233 (tablejump): Likewise.
9234 (*sibcall_memory): Likewise.
9235 (*sibcall_value_memory): Likewise.
9236 Disallow peepholes of indirect call and jump via memory for
9237 -mindirect-branch-register.
9238 (*call_pop): Replace m with Bw.
9239 (*call_value_pop): Likewise.
9240 (*sibcall_pop_memory): Replace m with Bs.
9241 * config/i386/i386.opt (mindirect-branch-register): New option.
9242 * doc/invoke.texi: Document -mindirect-branch-register option.
9244 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9246 * config/i386/i386-protos.h (ix86_output_function_return): New.
9247 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9248 set function_return_type.
9249 (indirect_thunk_name): Add ret_p to indicate thunk for function
9251 (output_indirect_thunk_function): Pass false to
9252 indirect_thunk_name.
9253 (ix86_output_indirect_branch_via_reg): Likewise.
9254 (ix86_output_indirect_branch_via_push): Likewise.
9255 (output_indirect_thunk_function): Create alias for function
9256 return thunk if regno < 0.
9257 (ix86_output_function_return): New function.
9258 (ix86_handle_fndecl_attribute): Handle function_return.
9259 (ix86_attribute_table): Add function_return.
9260 * config/i386/i386.h (machine_function): Add
9261 function_return_type.
9262 * config/i386/i386.md (simple_return_internal): Use
9263 ix86_output_function_return.
9264 (simple_return_internal_long): Likewise.
9265 * config/i386/i386.opt (mfunction-return=): New option.
9266 (indirect_branch): Mention -mfunction-return=.
9267 * doc/extend.texi: Document function_return function attribute.
9268 * doc/invoke.texi: Document -mfunction-return= option.
9270 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9272 * config/i386/i386-opts.h (indirect_branch): New.
9273 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9274 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9275 with local indirect jump when converting indirect call and jump.
9276 (ix86_set_indirect_branch_type): New.
9277 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9278 (indirectlabelno): New.
9279 (indirect_thunk_needed): Likewise.
9280 (indirect_thunk_bnd_needed): Likewise.
9281 (indirect_thunks_used): Likewise.
9282 (indirect_thunks_bnd_used): Likewise.
9283 (INDIRECT_LABEL): Likewise.
9284 (indirect_thunk_name): Likewise.
9285 (output_indirect_thunk): Likewise.
9286 (output_indirect_thunk_function): Likewise.
9287 (ix86_output_indirect_branch_via_reg): Likewise.
9288 (ix86_output_indirect_branch_via_push): Likewise.
9289 (ix86_output_indirect_branch): Likewise.
9290 (ix86_output_indirect_jmp): Likewise.
9291 (ix86_code_end): Call output_indirect_thunk_function if needed.
9292 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9294 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9295 (ix86_attribute_table): Add indirect_branch.
9296 * config/i386/i386.h (machine_function): Add indirect_branch_type
9297 and has_local_indirect_jump.
9298 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9300 (tablejump): Likewise.
9301 (*indirect_jump): Use ix86_output_indirect_jmp.
9302 (*tablejump_1): Likewise.
9303 (simple_return_indirect_internal): Likewise.
9304 * config/i386/i386.opt (mindirect-branch=): New option.
9305 (indirect_branch): New.
9308 (thunk-inline): Likewise.
9309 (thunk-extern): Likewise.
9310 * doc/extend.texi: Document indirect_branch function attribute.
9311 * doc/invoke.texi: Document -mindirect-branch= option.
9313 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9316 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9318 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9320 * ipa-inline.c (want_inline_small_function_p): Return false if
9321 inlining has already failed with CIF_FINAL_ERROR.
9322 (update_caller_keys): Call want_inline_small_function_p before
9324 (update_callee_keys): Likewise.
9326 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9328 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9330 (rs6000_quadword_masked_address_p): Likewise.
9331 (quad_aligned_load_p): Likewise.
9332 (quad_aligned_store_p): Likewise.
9333 (const_load_sequence_p): Add comment to describe the outer-most loop.
9334 (mimic_memory_attributes_and_flags): New function.
9335 (rs6000_gen_stvx): Likewise.
9336 (replace_swapped_aligned_store): Likewise.
9337 (rs6000_gen_lvx): Likewise.
9338 (replace_swapped_aligned_load): Likewise.
9339 (replace_swapped_load_constant): Capitalize argument name in
9340 comment describing this function.
9341 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9342 and stores that access quad-word aligned addresses and replace
9343 with stvx or lvx instructions when appropriate.
9344 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9345 New function prototype.
9346 (rs6000_quadword_masked_address_p): Likewise.
9347 (rs6000_gen_lvx): Likewise.
9348 (rs6000_gen_stvx): Likewise.
9349 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9350 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9351 when memory address is aligned.
9352 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9353 this split to select lvx instruction when memory address is aligned.
9354 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9355 instruction when memory address is aligned.
9356 (*vsx_le_perm_load_v16qi): Likewise.
9357 (four unnamed splitters): Modify to select the stvx instruction
9358 when memory is aligned.
9360 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9362 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9363 which appears in the queue multiple times.
9365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9366 Alan Hayward <alan.hayward@arm.com>
9367 David Sherwood <david.sherwood@arm.com>
9369 * tree-vectorizer.h (vec_lower_bound): New structure.
9370 (_loop_vec_info): Add check_nonzero and lower_bounds.
9371 (LOOP_VINFO_CHECK_NONZERO): New macro.
9372 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9373 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9374 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9375 fields. Make seg_len the distance travelled, not including the
9377 (dr_direction_indicator): Declare.
9378 (dr_zero_step_indicator): Likewise.
9379 (dr_known_forward_stride_p): Likewise.
9380 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9382 (runtime_alias_check_p): Allow runtime alias checks with
9384 (operator ==): Compare access_size and align.
9385 (prune_runtime_alias_test_list): Rework for new distinction between
9386 the access_size and seg_len.
9387 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9389 (get_segment_min_max): New function.
9390 (create_intersect_range_checks): Use it.
9391 (dr_step_indicator): New function.
9392 (dr_direction_indicator): Likewise.
9393 (dr_zero_step_indicator): Likewise.
9394 (dr_known_forward_stride_p): Likewise.
9395 * tree-loop-distribution.c (data_ref_segment_size): Return
9396 DR_STEP * (niters - 1).
9397 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9399 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9400 (vect_preserves_scalar_order_p): New function, split out from...
9401 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9402 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9403 (vect_vfa_access_size): New function.
9404 (vect_vfa_align): Likewise.
9405 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9406 (dump_lower_bound): New function.
9407 (vect_check_lower_bound): Likewise.
9408 (vect_small_gap_p): Likewise.
9409 (vectorizable_with_step_bound_p): Likewise.
9410 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9411 depencies if the vectorization factor is 1. Convert the checks
9412 for nonzero steps into checks on the bounds of DR_STEP. Try using
9413 a bunds check for variable steps if the minimum required step is
9414 relatively small. Update calls to the dr_with_seg_len
9415 constructor and to vect_compile_time_alias.
9416 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9418 (vect_loop_versioning): Call it.
9419 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9421 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9424 Alan Hayward <alan.hayward@arm.com>
9425 David Sherwood <david.sherwood@arm.com>
9427 * doc/sourcebuild.texi (vect_scatter_store): Document.
9428 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9430 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9432 * genopinit.c (main): Add supports_vec_scatter_store and
9433 supports_vec_scatter_store_cached to target_optabs.
9434 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9435 IFN_MASK_SCATTER_STORE.
9436 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9438 * internal-fn.h (internal_store_fn_p): Declare.
9439 (internal_fn_stored_value_index): Likewise.
9440 * internal-fn.c (scatter_store_direct): New macro.
9441 (expand_scatter_store_optab_fn): New function.
9442 (direct_scatter_store_optab_supported_p): New macro.
9443 (internal_store_fn_p): New function.
9444 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9445 IFN_MASK_SCATTER_STORE.
9446 (internal_fn_mask_index): Likewise.
9447 (internal_fn_stored_value_index): New function.
9448 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9450 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9451 * optabs-query.c (supports_vec_scatter_store_p): New function.
9452 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9453 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9454 true for scatter stores.
9455 (vect_gather_scatter_fn_p): Handle scatter stores too.
9456 (vect_check_gather_scatter): Consider using scatter stores if
9457 supports_vec_scatter_store_p.
9458 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9460 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9461 internal_fn_stored_value_index.
9462 (check_load_store_masking): Handle scatter stores too.
9463 (vect_get_store_rhs): Make public.
9464 (vectorizable_call): Use internal_store_fn_p.
9465 (vectorizable_store): Handle scatter store internal functions.
9466 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9467 when deciding whether the end of the group has been reached.
9468 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9469 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9470 (mask_scatter_store<mode>): New insns.
9472 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9473 Alan Hayward <alan.hayward@arm.com>
9474 David Sherwood <david.sherwood@arm.com>
9476 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9477 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9478 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9480 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9481 Use vect_truncate_gather_scatter_offset if we can't treat the
9482 operation as a normal gather load or scatter store.
9483 (get_group_load_store_type): Take the gather_scatter_info
9484 as argument. Try using a gather load or scatter store for
9485 single-element groups.
9486 (get_load_store_type): Update calls to get_group_load_store_type
9487 and vect_use_strided_gather_scatters_p.
9489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9490 Alan Hayward <alan.hayward@arm.com>
9491 David Sherwood <david.sherwood@arm.com>
9493 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9494 optional tree argument.
9495 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9497 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9498 but continue to use the current value as a fallback.
9499 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9500 to compare the updates.
9501 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9502 (get_load_store_type): Use it when handling a strided access.
9503 (vect_get_strided_load_store_ops): New function.
9504 (vect_get_data_ptr_increment): Likewise.
9505 (vectorizable_load): Handle strided gather loads. Always pass
9506 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9509 Alan Hayward <alan.hayward@arm.com>
9510 David Sherwood <david.sherwood@arm.com>
9512 * doc/md.texi (gather_load@var{m}): Document.
9513 (mask_gather_load@var{m}): Likewise.
9514 * genopinit.c (main): Add supports_vec_gather_load and
9515 supports_vec_gather_load_cached to target_optabs.
9516 * optabs-tree.c (init_tree_optimization_optabs): Use
9517 ggc_cleared_alloc to allocate target_optabs.
9518 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9519 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9521 * internal-fn.h (internal_load_fn_p): Declare.
9522 (internal_gather_scatter_fn_p): Likewise.
9523 (internal_fn_mask_index): Likewise.
9524 (internal_gather_scatter_fn_supported_p): Likewise.
9525 * internal-fn.c (gather_load_direct): New macro.
9526 (expand_gather_load_optab_fn): New function.
9527 (direct_gather_load_optab_supported_p): New macro.
9528 (direct_internal_fn_optab): New function.
9529 (internal_load_fn_p): Likewise.
9530 (internal_gather_scatter_fn_p): Likewise.
9531 (internal_fn_mask_index): Likewise.
9532 (internal_gather_scatter_fn_supported_p): Likewise.
9533 * optabs-query.c (supports_at_least_one_mode_p): New function.
9534 (supports_vec_gather_load_p): Likewise.
9535 * optabs-query.h (supports_vec_gather_load_p): Declare.
9536 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9537 and memory_type field.
9538 (NUM_PATTERNS): Bump to 15.
9539 * tree-vect-data-refs.c: Include internal-fn.h.
9540 (vect_gather_scatter_fn_p): New function.
9541 (vect_describe_gather_scatter_call): Likewise.
9542 (vect_check_gather_scatter): Try using internal functions for
9543 gather loads. Recognize existing calls to a gather load function.
9544 (vect_analyze_data_refs): Consider using gather loads if
9545 supports_vec_gather_load_p.
9546 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9547 (vect_get_gather_scatter_offset_type): Likewise.
9548 (vect_convert_mask_for_vectype): Likewise.
9549 (vect_add_conversion_to_patterm): Likewise.
9550 (vect_try_gather_scatter_pattern): Likewise.
9551 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9552 (vect_vect_recog_func_ptrs): Add it.
9553 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9554 internal_fn_mask_index and internal_gather_scatter_fn_p.
9555 (check_load_store_masking): Take the gather_scatter_info as an
9556 argument and handle gather loads.
9557 (vect_get_gather_scatter_ops): New function.
9558 (vectorizable_call): Check internal_load_fn_p.
9559 (vectorizable_load): Likewise. Handle gather load internal
9561 (vectorizable_store): Update call to check_load_store_masking.
9562 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9563 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9564 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9565 (aarch64_gather_scale_operand_d): New predicates.
9566 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9567 (mask_gather_load<mode>): New insns.
9569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9570 Alan Hayward <alan.hayward@arm.com>
9571 David Sherwood <david.sherwood@arm.com>
9573 * optabs.def (fold_left_plus_optab): New optab.
9574 * doc/md.texi (fold_left_plus_@var{m}): Document.
9575 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9576 * internal-fn.c (fold_left_direct): Define.
9577 (expand_fold_left_optab_fn): Likewise.
9578 (direct_fold_left_optab_supported_p): Likewise.
9579 * fold-const-call.c (fold_const_fold_left): New function.
9580 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9581 * tree-parloops.c (valid_reduction_p): New function.
9582 (gather_scalar_reductions): Use it.
9583 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9584 (vect_finish_replace_stmt): Declare.
9585 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9586 (needs_fold_left_reduction_p): New function, split out from...
9587 (vect_is_simple_reduction): ...here. Accept reductions that
9588 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9589 (vect_force_simple_reduction): Also store the reduction type in
9590 the assignment's STMT_VINFO_REDUC_TYPE.
9591 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9592 (merge_with_identity): New function.
9593 (vect_expand_fold_left): Likewise.
9594 (vectorize_fold_left_reduction): Likewise.
9595 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9596 scalar phi in place for it. Check for target support and reject
9597 cases that would reassociate the operation. Defer the transform
9598 phase to vectorize_fold_left_reduction.
9599 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9600 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9601 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9605 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9606 call to ifc_temp_var.
9608 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9609 Alan Hayward <alan.hayward@arm.com>
9610 David Sherwood <david.sherwood@arm.com>
9612 * target.def (legitimize_address_displacement): Take the original
9613 offset as a poly_int.
9614 * targhooks.h (default_legitimize_address_displacement): Update
9616 * targhooks.c (default_legitimize_address_displacement): Likewise.
9617 * doc/tm.texi: Regenerate.
9618 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9619 as an argument, moving assert of ad->disp == ad->disp_term to...
9620 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9621 Try calling targetm.legitimize_address_displacement before expanding
9622 the address rather than afterwards, and adjust for the new interface.
9623 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9624 Match the new hook interface. Handle SVE addresses.
9625 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9630 * Makefile.in (OBJS): Add early-remat.o.
9631 * target.def (select_early_remat_modes): New hook.
9632 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9633 * doc/tm.texi: Regenerate.
9634 * targhooks.h (default_select_early_remat_modes): Declare.
9635 * targhooks.c (default_select_early_remat_modes): New function.
9636 * timevar.def (TV_EARLY_REMAT): New timevar.
9637 * passes.def (pass_early_remat): New pass.
9638 * tree-pass.h (make_pass_early_remat): Declare.
9639 * early-remat.c: New file.
9640 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9642 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9645 Alan Hayward <alan.hayward@arm.com>
9646 David Sherwood <david.sherwood@arm.com>
9648 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9649 vfm1 with a bound_epilog parameter.
9650 (vect_do_peeling): Update calls accordingly, and move the prologue
9651 call earlier in the function. Treat the base bound_epilog as 0 for
9652 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9653 this base when peeling for gaps.
9654 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9655 with fully-masked loops.
9656 (vect_estimate_min_profitable_iters): Handle the single peeled
9657 iteration in that case.
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 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9664 single-element interleaving even if the size is not a power of 2.
9665 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9666 accesses for single-element interleaving if the group size is
9669 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9670 Alan Hayward <alan.hayward@arm.com>
9671 David Sherwood <david.sherwood@arm.com>
9673 * doc/md.texi (fold_extract_last_@var{m}): Document.
9674 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9675 * optabs.def (fold_extract_last_optab): New optab.
9676 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9677 * internal-fn.c (fold_extract_direct): New macro.
9678 (expand_fold_extract_optab_fn): Likewise.
9679 (direct_fold_extract_optab_supported_p): Likewise.
9680 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9681 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9682 EXTRACT_LAST_REDUCTION.
9683 (get_initial_def_for_reduction): Do not create an initial vector
9684 for EXTRACT_LAST_REDUCTION reductions.
9685 (vectorizable_reduction): Leave the scalar phi in place for
9686 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9687 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9688 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9689 transform phase to vectorizable_condition.
9690 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9692 (vect_finish_stmt_generation): ...here.
9693 (vect_finish_replace_stmt): New function.
9694 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9695 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9697 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9700 Alan Hayward <alan.hayward@arm.com>
9701 David Sherwood <david.sherwood@arm.com>
9703 * doc/md.texi (extract_last_@var{m}): Document.
9704 * optabs.def (extract_last_optab): New optab.
9705 * internal-fn.def (EXTRACT_LAST): New internal function.
9706 * internal-fn.c (cond_unary_direct): New macro.
9707 (expand_cond_unary_optab_fn): Likewise.
9708 (direct_cond_unary_optab_supported_p): Likewise.
9709 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9710 loops using EXTRACT_LAST.
9711 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9712 (extract_last_<mode>): ...this optab.
9713 (vec_extract<mode><Vel>): Update accordingly.
9715 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9716 Alan Hayward <alan.hayward@arm.com>
9717 David Sherwood <david.sherwood@arm.com>
9719 * target.def (empty_mask_is_expensive): New hook.
9720 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9721 * doc/tm.texi: Regenerate.
9722 * targhooks.h (default_empty_mask_is_expensive): Declare.
9723 * targhooks.c (default_empty_mask_is_expensive): New function.
9724 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9725 if the target says that empty masks are expensive.
9726 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9728 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9731 Alan Hayward <alan.hayward@arm.com>
9732 David Sherwood <david.sherwood@arm.com>
9734 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9735 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9736 (vect_use_loop_mask_for_alignment_p): New function.
9737 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9738 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9739 niters_skip argument. Make sure that the first niters_skip elements
9740 of the first iteration are inactive.
9741 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9742 Update call to vect_set_loop_masks_directly.
9743 (get_misalign_in_elems): New function, split out from...
9744 (vect_gen_prolog_loop_niters): ...here.
9745 (vect_update_init_of_dr): Take a code argument that specifies whether
9746 the adjustment should be added or subtracted.
9747 (vect_update_init_of_drs): Likewise.
9748 (vect_prepare_for_masked_peels): New function.
9749 (vect_do_peeling): Skip prologue peeling if we're using a mask
9750 instead. Update call to vect_update_inits_of_drs.
9751 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9753 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9754 alignment. Do not include the number of peeled iterations in
9755 the minimum threshold in that case.
9756 (vectorizable_induction): Adjust the start value down by
9757 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9758 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9759 Take the number of skipped iterations into account when calculating
9761 * tree-vect-stmts.c (vect_gen_while_not): New function.
9763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9764 Alan Hayward <alan.hayward@arm.com>
9765 David Sherwood <david.sherwood@arm.com>
9767 * doc/sourcebuild.texi (vect_fully_masked): Document.
9768 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9770 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9772 (vect_analyze_loop_2): ...here. Don't check the vectorization
9773 factor against the number of loop iterations if the loop is
9776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9777 Alan Hayward <alan.hayward@arm.com>
9778 David Sherwood <david.sherwood@arm.com>
9780 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9781 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9782 (dump_groups): Update accordingly.
9783 (iv_use::mem_type): New member variable.
9784 (address_p): New function.
9785 (record_use): Add a mem_type argument and initialize the new
9787 (record_group_use): Add a mem_type argument. Use address_p.
9788 Remove obsolete null checks of base_object. Update call to record_use.
9789 (find_interesting_uses_op): Update call to record_group_use.
9790 (find_interesting_uses_cond): Likewise.
9791 (find_interesting_uses_address): Likewise.
9792 (get_mem_type_for_internal_fn): New function.
9793 (find_address_like_use): Likewise.
9794 (find_interesting_uses_stmt): Try find_address_like_use before
9795 calling find_interesting_uses_op.
9796 (addr_offset_valid_p): Use the iv mem_type field as the type
9797 of the addressed memory.
9798 (add_autoinc_candidates): Likewise.
9799 (get_address_cost): Likewise.
9800 (split_small_address_groups_p): Use address_p.
9801 (split_address_groups): Likewise.
9802 (add_iv_candidate_for_use): Likewise.
9803 (autoinc_possible_for_pair): Likewise.
9804 (rewrite_groups): Likewise.
9805 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9806 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9807 (get_alias_ptr_type_for_ptr_address): New function.
9808 (rewrite_use_address): Rewrite address uses in calls that were
9809 identified by find_address_like_use.
9811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9812 Alan Hayward <alan.hayward@arm.com>
9813 David Sherwood <david.sherwood@arm.com>
9815 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9817 * gimple-expr.h (is_gimple_addressable: Likewise.
9818 * gimple-expr.c (is_gimple_address): Likewise.
9819 * internal-fn.c (expand_call_mem_ref): New function.
9820 (expand_mask_load_optab_fn): Use it.
9821 (expand_mask_store_optab_fn): Likewise.
9823 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9824 Alan Hayward <alan.hayward@arm.com>
9825 David Sherwood <david.sherwood@arm.com>
9827 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9828 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9829 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9830 (cond_umax@var{mode}): Document.
9831 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9832 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9833 (cond_umin_optab, cond_umax_optab): New optabs.
9834 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9835 (COND_IOR, COND_XOR): New internal functions.
9836 * internal-fn.h (get_conditional_internal_fn): Declare.
9837 * internal-fn.c (cond_binary_direct): New macro.
9838 (expand_cond_binary_optab_fn): Likewise.
9839 (direct_cond_binary_optab_supported_p): Likewise.
9840 (get_conditional_internal_fn): New function.
9841 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9842 Cope with reduction statements that are vectorized as calls rather
9844 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9845 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9846 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9847 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9848 (UNSPEC_COND_EOR): New unspecs.
9849 (optab): Add mappings for them.
9850 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9851 (sve_int_op, sve_fp_op): New int attributes.
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 * optabs.def (while_ult_optab): New optab.
9858 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9859 * internal-fn.def (WHILE_ULT): New internal function.
9860 * internal-fn.h (direct_internal_fn_supported_p): New override
9861 that takes two types as argument.
9862 * internal-fn.c (while_direct): New macro.
9863 (expand_while_optab_fn): New function.
9864 (convert_optab_supported_p): Likewise.
9865 (direct_while_optab_supported_p): New macro.
9866 * wide-int.h (wi::udiv_ceil): New function.
9867 * tree-vectorizer.h (rgroup_masks): New structure.
9868 (vec_loop_masks): New typedef.
9869 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9871 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9872 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9873 (vect_max_vf): New function.
9874 (slpeel_make_loop_iterate_ntimes): Delete.
9875 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9876 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9877 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9878 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9879 internal-fn.h, stor-layout.h and optabs-query.h.
9880 (vect_set_loop_mask): New function.
9881 (add_preheader_seq): Likewise.
9882 (add_header_seq): Likewise.
9883 (interleave_supported_p): Likewise.
9884 (vect_maybe_permute_loop_masks): Likewise.
9885 (vect_set_loop_masks_directly): Likewise.
9886 (vect_set_loop_condition_masked): Likewise.
9887 (vect_set_loop_condition_unmasked): New function, split out from
9888 slpeel_make_loop_iterate_ntimes.
9889 (slpeel_make_loop_iterate_ntimes): Rename to..
9890 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9891 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9892 (vect_do_peeling): Update call accordingly.
9893 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9895 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9896 mask_compare_type, can_fully_mask_p and fully_masked_p.
9897 (release_vec_loop_masks): New function.
9898 (_loop_vec_info): Use it to free the loop masks.
9899 (can_produce_all_loop_masks_p): New function.
9900 (vect_get_max_nscalars_per_iter): Likewise.
9901 (vect_verify_full_masking): Likewise.
9902 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9903 retries, and free the mask rgroups before retrying. Check loop-wide
9904 reasons for disallowing fully-masked loops. Make the final decision
9905 about whether use a fully-masked loop or not.
9906 (vect_estimate_min_profitable_iters): Do not assume that peeling
9907 for the number of iterations will be needed for fully-masked loops.
9908 (vectorizable_reduction): Disable fully-masked loops.
9909 (vectorizable_live_operation): Likewise.
9910 (vect_halve_mask_nunits): New function.
9911 (vect_double_mask_nunits): Likewise.
9912 (vect_record_loop_mask): Likewise.
9913 (vect_get_loop_mask): Likewise.
9914 (vect_transform_loop): Handle the case in which the final loop
9915 iteration might handle a partial vector. Call vect_set_loop_condition
9916 instead of slpeel_make_loop_iterate_ntimes.
9917 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9918 (check_load_store_masking): New function.
9919 (prepare_load_store_mask): Likewise.
9920 (vectorizable_store): Handle fully-masked loops.
9921 (vectorizable_load): Likewise.
9922 (supportable_widening_operation): Use vect_halve_mask_nunits for
9924 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9925 (vect_gen_while): New function.
9926 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9927 (aarch64_uqdec<mode>): New insn.
9929 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9930 Alan Hayward <alan.hayward@arm.com>
9931 David Sherwood <david.sherwood@arm.com>
9933 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9934 (reduc_xor_scal_optab): New optabs.
9935 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9936 (reduc_xor_scal_@var{m}): Document.
9937 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9938 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9940 * fold-const-call.c (fold_const_call): Handle them.
9941 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9942 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9943 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9944 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9945 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9946 (UNSPEC_XORV): New unspecs.
9947 (optab): Add entries for them.
9948 (BITWISEV): New int iterator.
9949 (bit_reduc_op): New int attributes.
9951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9952 Alan Hayward <alan.hayward@arm.com>
9953 David Sherwood <david.sherwood@arm.com>
9955 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9956 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9957 * optabs.def (vec_shl_insert_optab): New optab.
9958 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9959 (duplicate_and_interleave): Likewise.
9960 * tree-vect-loop.c: Include internal-fn.h.
9961 (neutral_op_for_slp_reduction): New function, split out from
9962 get_initial_defs_for_reduction.
9963 (get_initial_def_for_reduction): Handle option 2 for variable-length
9964 vectors by loading the neutral value into a vector and then shifting
9965 the initial value into element 0.
9966 (get_initial_defs_for_reduction): Replace the code argument with
9967 the neutral value calculated by neutral_op_for_slp_reduction.
9968 Use gimple_build_vector for constant-length vectors.
9969 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9970 but the first group_size elements have a neutral value.
9971 Use duplicate_and_interleave otherwise.
9972 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9973 Update call to get_initial_defs_for_reduction. Handle SLP
9974 reductions for variable-length vectors by creating one vector
9975 result for each scalar result, with the elements associated
9976 with other scalar results stubbed out with the neutral value.
9977 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9978 Require IFN_VEC_SHL_INSERT for double reductions on
9979 variable-length vectors, or SLP reductions that have
9980 a neutral value. Require can_duplicate_and_interleave_p
9981 support for variable-length unchained SLP reductions if there
9982 is no neutral value, such as for MIN/MAX reductions. Also require
9983 the number of vector elements to be a multiple of the number of
9984 SLP statements when doing variable-length unchained SLP reductions.
9985 Update call to vect_create_epilog_for_reduction.
9986 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9987 and remove initial values.
9988 (duplicate_and_interleave): Make public.
9989 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9990 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9993 Alan Hayward <alan.hayward@arm.com>
9994 David Sherwood <david.sherwood@arm.com>
9996 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9997 (can_duplicate_and_interleave_p): New function.
9998 (vect_get_and_check_slp_defs): Take the vector of statements
9999 rather than just the current one. Remove excess parentheses.
10000 Restriction rejectinon of vect_constant_def and vect_external_def
10001 for variable-length vectors to boolean types, or types for which
10002 can_duplicate_and_interleave_p is false.
10003 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10004 (duplicate_and_interleave): New function.
10005 (vect_get_constant_vectors): Use gimple_build_vector for
10006 constant-length vectors and suitable variable-length constant
10007 vectors. Use duplicate_and_interleave for other variable-length
10008 vectors. Don't defer the update when inserting new statements.
10010 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10011 Alan Hayward <alan.hayward@arm.com>
10012 David Sherwood <david.sherwood@arm.com>
10014 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10015 min_profitable_iters doesn't go negative.
10017 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10018 Alan Hayward <alan.hayward@arm.com>
10019 David Sherwood <david.sherwood@arm.com>
10021 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10022 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10023 * optabs.def (vec_mask_load_lanes_optab): New optab.
10024 (vec_mask_store_lanes_optab): Likewise.
10025 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10026 (MASK_STORE_LANES): Likewise.
10027 * internal-fn.c (mask_load_lanes_direct): New macro.
10028 (mask_store_lanes_direct): Likewise.
10029 (expand_mask_load_optab_fn): Handle masked operations.
10030 (expand_mask_load_lanes_optab_fn): New macro.
10031 (expand_mask_store_optab_fn): Handle masked operations.
10032 (expand_mask_store_lanes_optab_fn): New macro.
10033 (direct_mask_load_lanes_optab_supported_p): Likewise.
10034 (direct_mask_store_lanes_optab_supported_p): Likewise.
10035 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10037 (vect_load_lanes_supported): Likewise.
10038 * tree-vect-data-refs.c (strip_conversion): New function.
10039 (can_group_stmts_p): Likewise.
10040 (vect_analyze_data_ref_accesses): Use it instead of checking
10041 for a pair of assignments.
10042 (vect_store_lanes_supported): Take a masked_p parameter.
10043 (vect_load_lanes_supported): Likewise.
10044 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10045 vect_store_lanes_supported and vect_load_lanes_supported.
10046 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10047 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10048 parameter. Don't allow gaps for masked accesses.
10049 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10050 and vect_load_lanes_supported.
10051 (get_load_store_type): Take a masked_p parameter and update
10052 call to get_group_load_store_type.
10053 (vectorizable_store): Update call to get_load_store_type.
10054 Handle IFN_MASK_STORE_LANES.
10055 (vectorizable_load): Update call to get_load_store_type.
10056 Handle IFN_MASK_LOAD_LANES.
10058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10059 Alan Hayward <alan.hayward@arm.com>
10060 David Sherwood <david.sherwood@arm.com>
10062 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10064 * config/aarch64/aarch64-protos.h
10065 (aarch64_sve_struct_memory_operand_p): Declare.
10066 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10067 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10068 (VPRED, vpred): Handle SVE structure modes.
10069 * config/aarch64/constraints.md (Utx): New constraint.
10070 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10071 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10072 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10073 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10074 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10075 structure modes. Split into pieces after RA.
10076 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10077 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10079 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10080 SVE structure modes.
10081 (aarch64_classify_address): Likewise.
10082 (sizetochar): Move earlier in file.
10083 (aarch64_print_operand): Handle SVE register lists.
10084 (aarch64_array_mode): New function.
10085 (aarch64_sve_struct_memory_operand_p): Likewise.
10086 (TARGET_ARRAY_MODE): Redefine.
10088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10089 Alan Hayward <alan.hayward@arm.com>
10090 David Sherwood <david.sherwood@arm.com>
10092 * target.def (array_mode): New target hook.
10093 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10094 * doc/tm.texi: Regenerate.
10095 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10096 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10097 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10098 targetm.array_mode.
10099 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10103 Alan Hayward <alan.hayward@arm.com>
10104 David Sherwood <david.sherwood@arm.com>
10106 * fold-const.c (fold_binary_loc): Check the argument types
10107 rather than the result type when testing for a vector operation.
10109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10111 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10112 * doc/tm.texi: Regenerate.
10114 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10115 Alan Hayward <alan.hayward@arm.com>
10116 David Sherwood <david.sherwood@arm.com>
10118 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10119 (sve): Document new AArch64 extension.
10120 * doc/md.texi (w): Extend the description of the AArch64
10121 constraint to include SVE vectors.
10122 (Upl, Upa): Document new AArch64 predicate constraints.
10123 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10125 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10126 (msve-vector-bits=): New option.
10127 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10128 SVE when these are disabled.
10129 (sve): New extension.
10130 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10131 modes. Adjust their number of units based on aarch64_sve_vg.
10132 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10133 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10134 aarch64_addr_query_type.
10135 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10136 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10137 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10138 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10139 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10140 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10141 (aarch64_simd_imm_zero_p): Delete.
10142 (aarch64_check_zero_based_sve_index_immediate): Declare.
10143 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10144 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10145 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10146 (aarch64_sve_float_mul_immediate_p): Likewise.
10147 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10148 rather than an rtx.
10149 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10150 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10151 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10152 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10153 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10154 (aarch64_regmode_natural_size): Likewise.
10155 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10156 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10158 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10159 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10160 for VG and the SVE predicate registers.
10161 (V_ALIASES): Add a "z"-prefixed alias.
10162 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10163 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10164 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10165 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10166 (REG_CLASS_NAMES): Add entries for them.
10167 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10168 and the predicate registers.
10169 (aarch64_sve_vg): Declare.
10170 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10171 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10172 (REGMODE_NATURAL_SIZE): Define.
10173 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10175 * config/aarch64/aarch64.c: Include cfgrtl.h.
10176 (simd_immediate_info): Add a constructor for series vectors,
10177 and an associated step field.
10178 (aarch64_sve_vg): New variable.
10179 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10180 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10181 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10182 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10183 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10184 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10185 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10186 (aarch64_get_mask_mode): New functions.
10187 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10188 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10189 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10190 predicate modes and predicate registers. Explicitly restrict
10191 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10192 to store a vector mode if it is recognized by
10193 aarch64_classify_vector_mode.
10194 (aarch64_regmode_natural_size): New function.
10195 (aarch64_hard_regno_caller_save_mode): Return the original mode
10197 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10198 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10199 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10200 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10202 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10203 does not overlap dest if the function is frame-related. Handle
10205 (aarch64_split_add_offset): New function.
10206 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10207 them aarch64_add_offset.
10208 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10209 and update call to aarch64_sub_sp.
10210 (aarch64_add_cfa_expression): New function.
10211 (aarch64_expand_prologue): Pass extra temporary registers to the
10212 functions above. Handle the case in which we need to emit new
10213 DW_CFA_expressions for registers that were originally saved
10214 relative to the stack pointer, but now have to be expressed
10215 relative to the frame pointer.
10216 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10218 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10219 IP0 and IP1 values for SVE frames.
10220 (aarch64_expand_vec_series): New function.
10221 (aarch64_expand_sve_widened_duplicate): Likewise.
10222 (aarch64_expand_sve_const_vector): Likewise.
10223 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10224 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10225 into the register, rather than emitting a SET directly.
10226 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10227 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10228 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10229 (offset_9bit_signed_scaled_p): New functions.
10230 (aarch64_replicate_bitmask_imm): New function.
10231 (aarch64_bitmask_imm): Use it.
10232 (aarch64_cannot_force_const_mem): Reject expressions involving
10233 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10234 (aarch64_classify_index): Handle SVE indices, by requiring
10235 a plain register index with a scale that matches the element size.
10236 (aarch64_classify_address): Handle SVE addresses. Assert that
10237 the mode of the address is VOIDmode or an integer mode.
10238 Update call to aarch64_classify_symbol.
10239 (aarch64_classify_symbolic_expression): Update call to
10240 aarch64_classify_symbol.
10241 (aarch64_const_vec_all_in_range_p): New function.
10242 (aarch64_print_vector_float_operand): Likewise.
10243 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10244 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10245 and the FP immediates 1.0 and 0.5.
10246 (aarch64_print_address_internal): Handle SVE addresses.
10247 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10248 (aarch64_regno_regclass): Handle predicate registers.
10249 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10251 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10252 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10253 (aarch64_convert_sve_vector_bits): New function.
10254 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10255 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10256 rather than an rtx.
10257 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10258 Handle SVE vector and predicate modes. Accept VL-based constants
10259 that need only one temporary register, and VL offsets that require
10260 no temporary registers.
10261 (aarch64_conditional_register_usage): Mark the predicate registers
10262 as fixed if SVE isn't available.
10263 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10264 Return true for SVE vector and predicate modes.
10265 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10266 rather than an unsigned int. Handle SVE modes.
10267 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10269 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10271 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10272 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10273 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10274 (aarch64_sve_float_mul_immediate_p): New functions.
10275 (aarch64_sve_valid_immediate): New function.
10276 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10277 Explicitly reject structure modes. Check for INDEX constants.
10278 Handle PTRUE and PFALSE constants.
10279 (aarch64_check_zero_based_sve_index_immediate): New function.
10280 (aarch64_simd_imm_zero_p): Delete.
10281 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10282 vector modes. Accept constants in the range of CNT[BHWD].
10283 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10284 ask for an Advanced SIMD mode.
10285 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10286 (aarch64_simd_vector_alignment): Handle SVE predicates.
10287 (aarch64_vectorize_preferred_vector_alignment): New function.
10288 (aarch64_simd_vector_alignment_reachable): Use it instead of
10290 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10291 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10293 (MAX_VECT_LEN): Delete.
10294 (expand_vec_perm_d): Add a vec_flags field.
10295 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10296 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10297 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10299 (aarch64_evpc_rev): Rename to...
10300 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10301 (aarch64_evpc_rev_global): New function.
10302 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10303 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10305 (aarch64_evpc_sve_tbl): New function.
10306 (aarch64_expand_vec_perm_const_1): Update after rename of
10307 aarch64_evpc_rev. Handle SVE permutes too, trying
10308 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10309 than aarch64_evpc_tbl.
10310 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10311 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10312 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10313 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10314 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10315 (aarch64_expand_sve_vcond): New functions.
10316 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10317 of aarch64_vector_mode_p.
10318 (aarch64_dwarf_poly_indeterminate_value): New function.
10319 (aarch64_compute_pressure_classes): Likewise.
10320 (aarch64_can_change_mode_class): Likewise.
10321 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10322 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10323 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10324 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10325 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10326 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10327 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10328 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10330 (Dn, Dl, Dr): Accept const as well as const_vector.
10331 (Dz): Likewise. Compare against CONST0_RTX.
10332 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10333 of "vector" where appropriate.
10334 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10335 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10336 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10337 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10338 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10339 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10340 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10341 (v_int_equiv): Extend to SVE modes.
10342 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10344 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10345 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10346 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10347 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10348 (SVE_COND_FP_CMP): New int iterators.
10349 (perm_hilo): Handle the new unpack unspecs.
10350 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10352 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10353 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10354 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10355 (aarch64_equality_operator, aarch64_constant_vector_operand)
10356 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10357 (aarch64_sve_nonimmediate_operand): Likewise.
10358 (aarch64_sve_general_operand): Likewise.
10359 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10360 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10361 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10362 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10363 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10364 (aarch64_sve_float_arith_immediate): Likewise.
10365 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10366 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10367 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10368 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10369 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10370 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10371 (aarch64_sve_float_arith_operand): Likewise.
10372 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10373 (aarch64_sve_float_mul_operand): Likewise.
10374 (aarch64_sve_vec_perm_operand): Likewise.
10375 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10376 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10377 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10378 as well as const_vector.
10379 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10380 in file. Use CONST0_RTX and CONSTM1_RTX.
10381 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10382 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10383 Use aarch64_simd_imm_zero.
10384 * config/aarch64/aarch64-sve.md: New file.
10385 * config/aarch64/aarch64.md: Include it.
10386 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10387 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10388 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10389 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10390 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10391 (sve): New attribute.
10392 (enabled): Disable instructions with the sve attribute unless
10394 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10395 aarch64_expand_mov_immediate.
10396 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10397 CNT[BHSD] immediates.
10398 (movti): Split CONST_POLY_INT moves into two halves.
10399 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10400 Split additions that need a temporary here if the destination
10401 is the stack pointer.
10402 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10403 (*add<mode>3_poly_1): New instruction.
10404 (set_clobber_cc): New expander.
10406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10408 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10409 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10410 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10411 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10412 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10413 Change innermode from fixed_mode_size to machine_mode.
10414 (simplify_subreg): Update call accordingly. Handle a constant-sized
10415 subreg of a variable-length CONST_VECTOR.
10417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10418 Alan Hayward <alan.hayward@arm.com>
10419 David Sherwood <david.sherwood@arm.com>
10421 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10422 (add_offset_to_base): New function, split out from...
10423 (create_mem_ref): ...here. When handling a scale other than 1,
10424 check first whether the address is valid without the offset.
10425 Add it into the base if so, leaving the index and scale as-is.
10427 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10430 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10431 fold_for_warn before checking if arg2 is INTEGER_CST.
10433 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10435 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10436 (store_multiple_operation): Delete.
10437 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10438 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10439 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10440 guarded by TARGET_STRING.
10441 (rs6000_output_load_multiple): Delete.
10442 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10443 OPTION_MASK_STRING / TARGET_STRING handling.
10444 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10445 (const rs6000_opt_masks) <"string">: Change mask to 0.
10446 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10447 (MASK_STRING): Delete.
10448 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10450 (load_multiple): Delete.
10457 (store_multiple): Delete.
10464 (movmemsi_8reg): Delete.
10465 (corresponding unnamed define_insn): Delete.
10466 (movmemsi_6reg): Delete.
10467 (corresponding unnamed define_insn): Delete.
10468 (movmemsi_4reg): Delete.
10469 (corresponding unnamed define_insn): Delete.
10470 (movmemsi_2reg): Delete.
10471 (corresponding unnamed define_insn): Delete.
10472 (movmemsi_1reg): Delete.
10473 (corresponding unnamed define_insn): Delete.
10474 * config/rs6000/rs6000.opt (mno-string): New.
10475 (mstring): Replace by deprecation warning stub.
10476 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10478 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10480 * regrename.c (regrename_do_replace): If replacing the same
10481 reg multiple times, try to reuse last created gen_raw_REG.
10484 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10485 main to workaround a bug in GDB.
10487 2018-01-12 Tom de Vries <tom@codesourcery.com>
10490 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10492 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10494 PR rtl-optimization/80481
10495 * ira-color.c (get_cap_member): New function.
10496 (allocnos_conflict_by_live_ranges_p): Use it.
10497 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10498 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10500 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10503 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10504 (*saddl_se_1): Ditto.
10505 (*ssubsi_1): Ditto.
10506 (*ssubl_se_1): Ditto.
10508 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10510 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10511 rather than wi::to_widest for DR_INITs.
10512 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10513 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10514 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10516 (vect_analyze_group_access_1): Note that here.
10518 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10520 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10521 polynomial type sizes.
10523 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10525 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10526 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10527 (gimple_add_tmp_var): Likewise.
10529 2018-01-12 Martin Liska <mliska@suse.cz>
10531 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10532 (gimple_alloc_sizes): Likewise.
10533 (dump_gimple_statistics): Use PRIu64 in printf format.
10534 * gimple.h: Change uint64_t to int.
10536 2018-01-12 Martin Liska <mliska@suse.cz>
10538 * tree-core.h: Use uint64_t instead of int.
10539 * tree.c (tree_node_counts): Likewise.
10540 (tree_node_sizes): Likewise.
10541 (dump_tree_statistics): Use PRIu64 in printf format.
10543 2018-01-12 Martin Liska <mliska@suse.cz>
10545 * Makefile.in: As qsort_chk is implemented in vec.c, add
10546 vec.o to linkage of gencfn-macros.
10547 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10548 passing the info to record_node_allocation_statistics.
10549 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10551 * ggc-common.c (struct ggc_usage): Add operator== and use
10552 it in operator< and compare function.
10553 * mem-stats.h (struct mem_usage): Likewise.
10554 * vec.c (struct vec_usage): Remove operator< and compare
10555 function. Can be simply inherited.
10557 2018-01-12 Martin Jambor <mjambor@suse.cz>
10560 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10561 * tree-ssa-math-opts.c: Include domwalk.h.
10562 (convert_mult_to_fma_1): New function.
10563 (fma_transformation_info): New type.
10564 (fma_deferring_state): Likewise.
10565 (cancel_fma_deferring): New function.
10566 (result_of_phi): Likewise.
10567 (last_fma_candidate_feeds_initial_phi): Likewise.
10568 (convert_mult_to_fma): Added deferring logic, split actual
10569 transformation to convert_mult_to_fma_1.
10570 (math_opts_dom_walker): New type.
10571 (math_opts_dom_walker::after_dom_children): New method, body moved
10572 here from pass_optimize_widening_mul::execute, added deferring logic
10574 (pass_optimize_widening_mul::execute): Moved most of code to
10575 math_opts_dom_walker::after_dom_children.
10576 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10577 * config/i386/i386.c (ix86_option_override_internal): Added
10578 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10580 2018-01-12 Richard Biener <rguenther@suse.de>
10583 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10584 inline instance vars.
10586 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10589 * config/rx/rx.c (rx_is_restricted_memory_address):
10590 Handle SUBREG case.
10592 2018-01-12 Richard Biener <rguenther@suse.de>
10594 PR tree-optimization/80846
10595 * target.def (split_reduction): New target hook.
10596 * targhooks.c (default_split_reduction): New function.
10597 * targhooks.h (default_split_reduction): Declare.
10598 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10599 target requests first reduce vectors by combining low and high
10601 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10602 (get_vectype_for_scalar_type_and_size): Export.
10603 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10604 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10605 * doc/tm.texi: Regenerate.
10606 * config/i386/i386.c (ix86_split_reduction): Implement
10607 TARGET_VECTORIZE_SPLIT_REDUCTION.
10609 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10612 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10613 in PIC mode except for TARGET_VXWORKS_RTP.
10614 * config/sparc/sparc.c: Include cfgrtl.h.
10615 (TARGET_INIT_PIC_REG): Define.
10616 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10617 (sparc_pic_register_p): New predicate.
10618 (sparc_legitimate_address_p): Use it.
10619 (sparc_legitimize_pic_address): Likewise.
10620 (sparc_delegitimize_address): Likewise.
10621 (sparc_mode_dependent_address_p): Likewise.
10622 (gen_load_pcrel_sym): Remove 4th parameter.
10623 (load_got_register): Adjust call to above. Remove obsolete stuff.
10624 (sparc_expand_prologue): Do not call load_got_register here.
10625 (sparc_flat_expand_prologue): Likewise.
10626 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10627 (sparc_use_pseudo_pic_reg): New function.
10628 (sparc_init_pic_reg): Likewise.
10629 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10630 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10632 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10634 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10635 Add item for branch_cost.
10637 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10639 PR rtl-optimization/83565
10640 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10641 not extend the result to a larger mode for rotate operations.
10642 (num_sign_bit_copies1): Likewise.
10644 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10647 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10649 Use values-Xc.o for -pedantic.
10650 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10652 2018-01-12 Martin Liska <mliska@suse.cz>
10655 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10657 (possible_polymorphic_call_targets): Use it.
10658 (ipa_devirt): Likewise.
10660 2018-01-12 Martin Liska <mliska@suse.cz>
10662 * profile-count.h (enum profile_quality): Use 0 as invalid
10663 enum value of profile_quality.
10665 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10667 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10668 -mext-string options.
10670 2018-01-12 Richard Biener <rguenther@suse.de>
10672 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10673 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10674 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10676 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10678 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10680 * configure.ac (--with-long-double-format): Add support for the
10681 configuration option to change the default long double format on
10683 * config.gcc (powerpc*-linux*-*): Likewise.
10684 * configure: Regenerate.
10685 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10686 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10687 used without modification.
10689 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10691 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10692 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10693 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10694 MISC_BUILTIN_SPEC_BARRIER.
10695 (rs6000_init_builtins): Likewise.
10696 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10698 (speculation_barrier): New define_insn.
10699 * doc/extend.texi: Document __builtin_speculation_barrier.
10701 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10704 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10705 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10706 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10708 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10709 integral modes instead of "ss" and "sd".
10710 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10711 vectors with 32-bit and 64-bit elements.
10712 (vecdupssescalarmodesuffix): New mode attribute.
10713 (vec_dup<mode>): Use it.
10715 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10718 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10719 frame if argument is passed on stack.
10721 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10724 * ree.c (combine_reaching_defs): Optimize also
10725 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10726 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10728 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10730 PR middle-end/83189
10731 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10733 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10735 PR middle-end/83718
10736 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10737 after they are computed.
10739 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10741 PR tree-optimization/83695
10742 * gimple-loop-linterchange.cc
10743 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10744 reset cached scev information after interchange.
10745 (pass_linterchange::execute): Remove call to scev_reset_htab.
10747 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10749 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10750 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10751 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10752 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10753 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10754 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10755 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10756 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10757 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10758 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10759 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10760 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10761 (V_lane_reg): Likewise.
10762 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10764 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10765 (vfmal_lane_low<mode>_intrinsic,
10766 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10767 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10768 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10769 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10770 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10771 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10773 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10775 * config/arm/arm-cpus.in (fp16fml): New feature.
10776 (ALL_SIMD): Add fp16fml.
10777 (armv8.2-a): Add fp16fml as an option.
10778 (armv8.3-a): Likewise.
10779 (armv8.4-a): Add fp16fml as part of fp16.
10780 * config/arm/arm.h (TARGET_FP16FML): Define.
10781 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10783 * config/arm/arm-modes.def (V2HF): Define.
10784 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10785 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10786 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10787 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10788 vfmsl_low, vfmsl_high): New set of builtins.
10789 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10790 (vfml_op): New code attribute.
10791 (VFMLHALVES): New int iterator.
10792 (VFML, VFMLSEL): New mode attributes.
10793 (V_reg): Define mapping for V2HF.
10794 (V_hi, V_lo): New mode attributes.
10795 (VF_constraint): Likewise.
10796 (vfml_half, vfml_half_selector): New int attributes.
10797 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10799 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10800 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10802 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10803 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10804 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10805 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10807 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10808 Document new effective target and option set.
10810 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10812 * config/arm/arm-cpus.in (armv8_4): New feature.
10813 (ARMv8_4a): New fgroup.
10814 (armv8.4-a): New arch.
10815 * config/arm/arm-tables.opt: Regenerate.
10816 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10817 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10818 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10819 Add matching rules for -march=armv8.4-a and extensions.
10820 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10822 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10825 * config/rx/rx.md (BW): New mode attribute.
10826 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10828 2018-01-11 Richard Biener <rguenther@suse.de>
10830 PR tree-optimization/83435
10831 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10832 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10833 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10835 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10836 Alan Hayward <alan.hayward@arm.com>
10837 David Sherwood <david.sherwood@arm.com>
10839 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10841 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10842 (aarch64_print_address_internal): Use it to check for a zero offset.
10844 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10845 Alan Hayward <alan.hayward@arm.com>
10846 David Sherwood <david.sherwood@arm.com>
10848 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10849 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10850 Return a poly_int64 rather than a HOST_WIDE_INT.
10851 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10852 rather than a HOST_WIDE_INT.
10853 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10854 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10855 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10856 final_offset from HOST_WIDE_INT to poly_int64.
10857 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10858 to_constant when getting the number of units in an Advanced SIMD
10860 (aarch64_builtin_vectorized_function): Check for a constant number
10862 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10864 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10865 attribute instead of GET_MODE_NUNITS.
10866 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10867 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10868 GET_MODE_SIZE for fixed-size registers.
10869 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10870 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10871 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10872 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10873 (aarch64_print_operand, aarch64_print_address_internal)
10874 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10875 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10876 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10877 Handle polynomial GET_MODE_SIZE.
10878 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10879 wider than SImode without modification.
10880 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10881 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10882 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10883 passing and returning SVE modes.
10884 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10885 rather than GEN_INT.
10886 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10887 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10888 (aarch64_allocate_and_probe_stack_space): Likewise.
10889 (aarch64_layout_frame): Cope with polynomial offsets.
10890 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10891 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10892 polynomial offsets.
10893 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10894 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10895 poly_int64 rather than a HOST_WIDE_INT.
10896 (aarch64_get_separate_components, aarch64_process_components)
10897 (aarch64_expand_prologue, aarch64_expand_epilogue)
10898 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10899 (aarch64_anchor_offset): New function, split out from...
10900 (aarch64_legitimize_address): ...here.
10901 (aarch64_builtin_vectorization_cost): Handle polynomial
10902 TYPE_VECTOR_SUBPARTS.
10903 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10905 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10906 number of elements from the PARALLEL rather than the mode.
10907 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10908 rather than GET_MODE_BITSIZE.
10909 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10910 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10911 (aarch64_expand_vec_perm_const_1): Handle polynomial
10912 d->perm.length () and d->perm elements.
10913 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10914 Apply to_constant to d->perm elements.
10915 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10916 polynomial CONST_VECTOR_NUNITS.
10917 (aarch64_move_pointer): Take amount as a poly_int64 rather
10919 (aarch64_progress_pointer): Avoid temporary variable.
10920 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10921 the mode attribute instead of GET_MODE.
10923 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10924 Alan Hayward <alan.hayward@arm.com>
10925 David Sherwood <david.sherwood@arm.com>
10927 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10928 x exists before using it.
10929 (aarch64_add_constant_internal): Rename to...
10930 (aarch64_add_offset_1): ...this. Replace regnum with separate
10931 src and dest rtxes. Handle the case in which they're different,
10932 including when the offset is zero. Replace scratchreg with an rtx.
10933 Use 2 additions if there is no spare register into which we can
10934 move a 16-bit constant.
10935 (aarch64_add_constant): Delete.
10936 (aarch64_add_offset): Replace reg with separate src and dest
10937 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10938 Use aarch64_add_offset_1.
10939 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10940 an rtx rather than an int. Take the delta as a poly_int64
10941 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10942 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10943 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10944 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10945 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10946 and aarch64_add_sp.
10947 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10948 aarch64_add_constant.
10950 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10952 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10953 Use scalar_float_mode.
10955 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10957 * config/aarch64/aarch64-simd.md
10958 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10959 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10960 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10961 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10962 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10963 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10964 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10965 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10966 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10967 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10969 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10972 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10973 targ_options->x_arm_arch_string is non NULL.
10975 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10977 * config/aarch64/aarch64.h
10978 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10980 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10983 * expmed.c (emit_store_flag_force): Swap if const op0
10984 and change VOIDmode to mode of op0.
10986 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10988 PR rtl-optimization/83761
10989 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10990 than bytes to mode_for_size.
10992 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10994 PR middle-end/83189
10995 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10996 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10999 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11001 PR middle-end/83575
11002 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11003 when in layout mode.
11004 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11005 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11008 2018-01-10 Michael Collison <michael.collison@arm.com>
11010 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11011 * config/aarch64/aarch64-option-extension.def: Add
11012 AARCH64_OPT_EXTENSION of 'fp16fml'.
11013 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11014 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11015 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11016 * config/aarch64/constraints.md (Ui7): New constraint.
11017 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11018 (VFMLA_SEL_W): Ditto.
11021 (VFMLA16_LOW): New int iterator.
11022 (VFMLA16_HIGH): Ditto.
11023 (UNSPEC_FMLAL): New unspec.
11024 (UNSPEC_FMLSL): Ditto.
11025 (UNSPEC_FMLAL2): Ditto.
11026 (UNSPEC_FMLSL2): Ditto.
11027 (f16mac): New code attribute.
11028 * config/aarch64/aarch64-simd-builtins.def
11029 (aarch64_fmlal_lowv2sf): Ditto.
11030 (aarch64_fmlsl_lowv2sf): Ditto.
11031 (aarch64_fmlalq_lowv4sf): Ditto.
11032 (aarch64_fmlslq_lowv4sf): Ditto.
11033 (aarch64_fmlal_highv2sf): Ditto.
11034 (aarch64_fmlsl_highv2sf): Ditto.
11035 (aarch64_fmlalq_highv4sf): Ditto.
11036 (aarch64_fmlslq_highv4sf): Ditto.
11037 (aarch64_fmlal_lane_lowv2sf): Ditto.
11038 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11039 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11040 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11041 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11042 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11043 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11044 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11045 (aarch64_fmlal_lane_highv2sf): Ditto.
11046 (aarch64_fmlsl_lane_highv2sf): Ditto.
11047 (aarch64_fmlal_laneq_highv2sf): Ditto.
11048 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11049 (aarch64_fmlalq_lane_highv4sf): Ditto.
11050 (aarch64_fmlsl_lane_highv4sf): Ditto.
11051 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11052 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11053 * config/aarch64/aarch64-simd.md:
11054 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11055 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11056 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11057 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11058 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11059 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11060 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11061 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11062 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11063 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11064 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11065 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11066 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11067 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11068 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11069 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11070 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11071 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11072 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11073 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11074 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11075 (vfmlsl_low_u32): Ditto.
11076 (vfmlalq_low_u32): Ditto.
11077 (vfmlslq_low_u32): Ditto.
11078 (vfmlal_high_u32): Ditto.
11079 (vfmlsl_high_u32): Ditto.
11080 (vfmlalq_high_u32): Ditto.
11081 (vfmlslq_high_u32): Ditto.
11082 (vfmlal_lane_low_u32): Ditto.
11083 (vfmlsl_lane_low_u32): Ditto.
11084 (vfmlal_laneq_low_u32): Ditto.
11085 (vfmlsl_laneq_low_u32): Ditto.
11086 (vfmlalq_lane_low_u32): Ditto.
11087 (vfmlslq_lane_low_u32): Ditto.
11088 (vfmlalq_laneq_low_u32): Ditto.
11089 (vfmlslq_laneq_low_u32): Ditto.
11090 (vfmlal_lane_high_u32): Ditto.
11091 (vfmlsl_lane_high_u32): Ditto.
11092 (vfmlal_laneq_high_u32): Ditto.
11093 (vfmlsl_laneq_high_u32): Ditto.
11094 (vfmlalq_lane_high_u32): Ditto.
11095 (vfmlslq_lane_high_u32): Ditto.
11096 (vfmlalq_laneq_high_u32): Ditto.
11097 (vfmlslq_laneq_high_u32): Ditto.
11098 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11099 (AARCH64_FL_FOR_ARCH8_4): New.
11100 (AARCH64_ISA_F16FML): New ISA flag.
11101 (TARGET_F16FML): New feature flag for fp16fml.
11102 (doc/invoke.texi): Document new fp16fml option.
11104 2018-01-10 Michael Collison <michael.collison@arm.com>
11106 * config/aarch64/aarch64-builtins.c:
11107 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11108 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11109 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11110 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11111 (AARCH64_ISA_SHA3): New ISA flag.
11112 (TARGET_SHA3): New feature flag for sha3.
11113 * config/aarch64/iterators.md (sha512_op): New int attribute.
11114 (CRYPTO_SHA512): New int iterator.
11115 (UNSPEC_SHA512H): New unspec.
11116 (UNSPEC_SHA512H2): Ditto.
11117 (UNSPEC_SHA512SU0): Ditto.
11118 (UNSPEC_SHA512SU1): Ditto.
11119 * config/aarch64/aarch64-simd-builtins.def
11120 (aarch64_crypto_sha512hqv2di): New builtin.
11121 (aarch64_crypto_sha512h2qv2di): Ditto.
11122 (aarch64_crypto_sha512su0qv2di): Ditto.
11123 (aarch64_crypto_sha512su1qv2di): Ditto.
11124 (aarch64_eor3qv8hi): Ditto.
11125 (aarch64_rax1qv2di): Ditto.
11126 (aarch64_xarqv2di): Ditto.
11127 (aarch64_bcaxqv8hi): Ditto.
11128 * config/aarch64/aarch64-simd.md:
11129 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11130 (aarch64_crypto_sha512su0qv2di): Ditto.
11131 (aarch64_crypto_sha512su1qv2di): Ditto.
11132 (aarch64_eor3qv8hi): Ditto.
11133 (aarch64_rax1qv2di): Ditto.
11134 (aarch64_xarqv2di): Ditto.
11135 (aarch64_bcaxqv8hi): Ditto.
11136 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11137 (vsha512h2q_u64): Ditto.
11138 (vsha512su0q_u64): Ditto.
11139 (vsha512su1q_u64): Ditto.
11140 (veor3q_u16): Ditto.
11141 (vrax1q_u64): Ditto.
11142 (vxarq_u64): Ditto.
11143 (vbcaxq_u16): Ditto.
11144 * config/arm/types.md (crypto_sha512): New type attribute.
11145 (crypto_sha3): Ditto.
11146 (doc/invoke.texi): Document new sha3 option.
11148 2018-01-10 Michael Collison <michael.collison@arm.com>
11150 * config/aarch64/aarch64-builtins.c:
11151 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11152 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11153 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11154 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11155 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11156 (AARCH64_ISA_SM4): New ISA flag.
11157 (TARGET_SM4): New feature flag for sm4.
11158 * config/aarch64/aarch64-simd-builtins.def
11159 (aarch64_sm3ss1qv4si): Ditto.
11160 (aarch64_sm3tt1aq4si): Ditto.
11161 (aarch64_sm3tt1bq4si): Ditto.
11162 (aarch64_sm3tt2aq4si): Ditto.
11163 (aarch64_sm3tt2bq4si): Ditto.
11164 (aarch64_sm3partw1qv4si): Ditto.
11165 (aarch64_sm3partw2qv4si): Ditto.
11166 (aarch64_sm4eqv4si): Ditto.
11167 (aarch64_sm4ekeyqv4si): Ditto.
11168 * config/aarch64/aarch64-simd.md:
11169 (aarch64_sm3ss1qv4si): Ditto.
11170 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11171 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11172 (aarch64_sm4eqv4si): Ditto.
11173 (aarch64_sm4ekeyqv4si): Ditto.
11174 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11175 (sm3part_op): Ditto.
11176 (CRYPTO_SM3TT): Ditto.
11177 (CRYPTO_SM3PART): Ditto.
11178 (UNSPEC_SM3SS1): New unspec.
11179 (UNSPEC_SM3TT1A): Ditto.
11180 (UNSPEC_SM3TT1B): Ditto.
11181 (UNSPEC_SM3TT2A): Ditto.
11182 (UNSPEC_SM3TT2B): Ditto.
11183 (UNSPEC_SM3PARTW1): Ditto.
11184 (UNSPEC_SM3PARTW2): Ditto.
11185 (UNSPEC_SM4E): Ditto.
11186 (UNSPEC_SM4EKEY): Ditto.
11187 * config/aarch64/constraints.md (Ui2): New constraint.
11188 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11189 * config/arm/types.md (crypto_sm3): New type attribute.
11190 (crypto_sm4): Ditto.
11191 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11192 (vsm3tt1aq_u32): Ditto.
11193 (vsm3tt1bq_u32): Ditto.
11194 (vsm3tt2aq_u32): Ditto.
11195 (vsm3tt2bq_u32): Ditto.
11196 (vsm3partw1q_u32): Ditto.
11197 (vsm3partw2q_u32): Ditto.
11198 (vsm4eq_u32): Ditto.
11199 (vsm4ekeyq_u32): Ditto.
11200 (doc/invoke.texi): Document new sm4 option.
11202 2018-01-10 Michael Collison <michael.collison@arm.com>
11204 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11205 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11206 (AARCH64_FL_FOR_ARCH8_4): New.
11207 (AARCH64_FL_V8_4): New flag.
11208 (doc/invoke.texi): Document new armv8.4-a option.
11210 2018-01-10 Michael Collison <michael.collison@arm.com>
11212 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11213 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11214 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11215 * config/aarch64/aarch64-option-extension.def: Add
11216 AARCH64_OPT_EXTENSION of 'sha2'.
11217 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11218 (crypto): Disable sha2 and aes if crypto disabled.
11219 (crypto): Enable aes and sha2 if enabled.
11220 (simd): Disable sha2 and aes if simd disabled.
11221 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11223 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11224 (TARGET_SHA2): New feature flag for sha2.
11225 (TARGET_AES): New feature flag for aes.
11226 * config/aarch64/aarch64-simd.md:
11227 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11228 conditional on TARGET_AES.
11229 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11230 (aarch64_crypto_sha1hsi): Make pattern conditional
11232 (aarch64_crypto_sha1hv4si): Ditto.
11233 (aarch64_be_crypto_sha1hv4si): Ditto.
11234 (aarch64_crypto_sha1su1v4si): Ditto.
11235 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11236 (aarch64_crypto_sha1su0v4si): Ditto.
11237 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11238 (aarch64_crypto_sha256su0v4si): Ditto.
11239 (aarch64_crypto_sha256su1v4si): Ditto.
11240 (doc/invoke.texi): Document new aes and sha2 options.
11242 2018-01-10 Martin Sebor <msebor@redhat.com>
11244 PR tree-optimization/83781
11245 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11248 2018-01-11 Martin Sebor <msebor@gmail.com>
11249 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11251 PR tree-optimization/83501
11252 PR tree-optimization/81703
11254 * tree-ssa-strlen.c (get_string_cst): Rename...
11255 (get_string_len): ...to this. Handle global constants.
11256 (handle_char_store): Adjust.
11258 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11259 Jim Wilson <jimw@sifive.com>
11261 * config/riscv/riscv-protos.h (riscv_output_return): New.
11262 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11263 (riscv_attribute_table, riscv_output_return),
11264 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11265 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11266 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11268 (riscv_expand_prologue): Add early return for naked function.
11269 (riscv_expand_epilogue): Likewise.
11270 (riscv_function_ok_for_sibcall): Return false for naked function.
11271 (riscv_set_current_function): New.
11272 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11273 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11274 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11275 * doc/extend.texi (RISC-V Function Attributes): New.
11277 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11279 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11280 check for 128-bit long double before checking TCmode.
11281 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11282 128-bit long doubles before checking TFmode or TCmode.
11283 (FLOAT128_IBM_P): Likewise.
11285 2018-01-10 Martin Sebor <msebor@redhat.com>
11287 PR tree-optimization/83671
11288 * builtins.c (c_strlen): Unconditionally return zero for the empty
11290 Use -Warray-bounds for warnings.
11291 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11292 for non-constant array indices with COMPONENT_REF, arrays of
11293 arrays, and pointers to arrays.
11294 (gimple_fold_builtin_strlen): Determine and set length range for
11295 non-constant character arrays.
11297 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11299 PR middle-end/81897
11300 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11303 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11305 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11307 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11310 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11311 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11312 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11313 indexed_or_indirect_operand predicate.
11314 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11315 (*vsx_le_perm_load_v8hi): Likewise.
11316 (*vsx_le_perm_load_v16qi): Likewise.
11317 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11318 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11319 (*vsx_le_perm_store_v8hi): Likewise.
11320 (*vsx_le_perm_store_v16qi): Likewise.
11321 (eight unnamed splitters): Likewise.
11323 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11325 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11326 * config/rs6000/emmintrin.h: Likewise.
11327 * config/rs6000/mmintrin.h: Likewise.
11328 * config/rs6000/xmmintrin.h: Likewise.
11330 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11333 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11335 * tree.c (tree_nop_conversion): Return true for location wrapper
11337 (maybe_wrap_with_location): New function.
11338 (selftest::check_strip_nops): New function.
11339 (selftest::test_location_wrappers): New function.
11340 (selftest::tree_c_tests): Call it.
11341 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11342 (maybe_wrap_with_location): New decl.
11343 (EXPR_LOCATION_WRAPPER_P): New macro.
11344 (location_wrapper_p): New inline function.
11345 (tree_strip_any_location_wrapper): New inline function.
11347 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11350 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11351 stack_realign_offset for the largest alignment of stack slot
11353 (ix86_find_max_used_stack_alignment): New function.
11354 (ix86_finalize_stack_frame_flags): Use it. Set
11355 max_used_stack_alignment if we don't realign stack.
11356 * config/i386/i386.h (machine_function): Add
11357 max_used_stack_alignment.
11359 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11361 * config/arm/arm.opt (-mbranch-cost): New option.
11362 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11365 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11368 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11369 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11371 2018-01-10 Richard Biener <rguenther@suse.de>
11374 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11375 early out so it also covers the case where we have a non-NULL
11378 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11380 PR tree-optimization/83753
11381 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11382 for non-strided grouped accesses if the number of elements is 1.
11384 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11387 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11388 * i386.h (TARGET_USE_GATHER): Define.
11389 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11391 2018-01-10 Martin Liska <mliska@suse.cz>
11394 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11395 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11397 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11398 CLEANUP_NO_PARTITIONING is not set.
11400 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11402 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11403 for vectors, as a partial revert of r254296.
11404 * rtl.h (const_vec_p): Delete.
11405 (const_vec_duplicate_p): Don't test for vector CONSTs.
11406 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11407 * expmed.c (make_tree): Likewise.
11410 * common.md (E, F): Use CONSTANT_P instead of checking for
11412 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11413 checking for CONST_VECTOR.
11415 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11417 PR middle-end/83575
11418 * predict.c (force_edge_cold): Handle in more sane way edges
11419 with no prediction.
11421 2018-01-09 Carl Love <cel@us.ibm.com>
11423 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11425 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11426 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11427 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11428 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11429 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11430 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11431 * config/rs6000/rs6000-protos.h: Add extern defition for
11432 rs6000_generate_float2_double_code.
11433 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11435 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11436 (float2_v2df): Add define_expand.
11438 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11441 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11442 op_mode in the force_to_mode call.
11444 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11446 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11447 instead of checking each element individually.
11448 (aarch64_evpc_uzp): Likewise.
11449 (aarch64_evpc_zip): Likewise.
11450 (aarch64_evpc_ext): Likewise.
11451 (aarch64_evpc_rev): Likewise.
11452 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11453 instead of checking each element individually. Return true without
11455 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11456 whether all selected elements come from the same input, instead of
11457 checking each element individually. Remove calls to gen_rtx_REG,
11458 start_sequence and end_sequence and instead assert that no rtl is
11461 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11463 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11464 order of HIGH and CONST checks.
11466 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11468 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11469 if the destination isn't an SSA_NAME.
11471 2018-01-09 Richard Biener <rguenther@suse.de>
11473 PR tree-optimization/83668
11474 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11476 (canonicalize_loop_form): ... here, renamed from ...
11477 (canonicalize_loop_closed_ssa_form): ... this and amended to
11478 swap successor edges for loop exit blocks to make us use
11479 the RPO order we need for initial schedule generation.
11481 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11483 PR tree-optimization/64811
11484 * match.pd: When optimizing comparisons with Inf, avoid
11485 introducing or losing exceptions from comparisons with NaN.
11487 2018-01-09 Martin Liska <mliska@suse.cz>
11490 * asan.c (shadow_mem_size): Add gcc_assert.
11492 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11494 Don't save registers in main().
11497 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11498 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11499 * config/avr/avr.c (avr_set_current_function): Don't error if
11500 naked, OS_task or OS_main are specified at the same time.
11501 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11503 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11505 * common/config/avr/avr-common.c (avr_option_optimization_table):
11506 Switch on -mmain-is-OS_task for optimizing compilations.
11508 2018-01-09 Richard Biener <rguenther@suse.de>
11510 PR tree-optimization/83572
11511 * graphite.c: Include cfganal.h.
11512 (graphite_transform_loops): Connect infinite loops to exit
11513 and remove fake edges at the end.
11515 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11517 * ipa-inline.c (edge_badness): Revert accidental checkin.
11519 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11522 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11523 symbols; not inline clones.
11525 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11528 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11529 hard registers. Formatting fixes.
11531 PR preprocessor/83722
11532 * gcc.c (try_generate_repro): Pass
11533 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11534 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11537 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11538 Kito Cheng <kito.cheng@gmail.com>
11540 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11541 (riscv_leaf_function_p): Delete.
11542 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11544 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11546 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11548 (do_ifelse): New function.
11549 (do_isel): New function.
11550 (do_sub3): New function.
11551 (do_add3): New function.
11552 (do_load_mask_compare): New function.
11553 (do_overlap_load_compare): New function.
11554 (expand_compare_loop): New function.
11555 (expand_block_compare): Call expand_compare_loop() when appropriate.
11556 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11557 option description.
11558 (-mblock-compare-inline-loop-limit): New option.
11560 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11563 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11564 Reverse order of second and third operands in first alternative.
11565 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11566 of first and second elements in UNSPEC_VPERMR vector.
11567 (altivec_expand_vec_perm_le): Likewise.
11569 2018-01-08 Jeff Law <law@redhat.com>
11571 PR rtl-optimizatin/81308
11572 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11573 (process_switch): If group_case_labels makes a change, then set
11575 (pass_convert_switch::execute): If a switch is converted, then
11576 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11578 PR rtl-optimization/81308
11579 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11582 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11584 PR target/83663 - Revert r255946
11585 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11586 generation for cases where splatting a value is not useful.
11587 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11588 across a vec_duplicate and a paradoxical subreg forming a vector
11589 mode to a vec_concat.
11591 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11593 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11594 -march=armv8.3-a variants.
11595 * config/arm/t-multilib: Likewise.
11596 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11598 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11600 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11602 (cceq_ior_compare_complement): Give it a name so I can use it, and
11603 change boolean_or_operator predicate to boolean_operator so it can
11604 be used to generate a crand.
11605 (eqne): New code iterator.
11606 (bd/bd_neg): New code_attrs.
11607 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11608 a single define_insn.
11609 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11610 decrement (bdnzt/bdnzf/bdzt/bdzf).
11611 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11612 with the new names of the branch decrement patterns, and added the
11613 names of the branch decrement conditional patterns.
11615 2018-01-08 Richard Biener <rguenther@suse.de>
11617 PR tree-optimization/83563
11618 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11621 2018-01-08 Richard Biener <rguenther@suse.de>
11623 PR middle-end/83713
11624 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11626 2018-01-08 Richard Biener <rguenther@suse.de>
11628 PR tree-optimization/83685
11629 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11630 references to abnormals.
11632 2018-01-08 Richard Biener <rguenther@suse.de>
11635 * dwarf2out.c (output_indirect_strings): Handle empty
11636 skeleton_debug_str_hash.
11637 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11639 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11641 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11642 (emit_store_direct): Likewise.
11643 (arc_trampoline_adjust_address): Likewise.
11644 (arc_asm_trampoline_template): New function.
11645 (arc_initialize_trampoline): Use asm_trampoline_template.
11646 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11647 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11648 * config/arc/arc.md (flush_icache): Delete pattern.
11650 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11652 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11653 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11656 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11659 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11660 by not USED_FOR_TARGET.
11661 (make_pass_resolve_sw_modes): Likewise.
11663 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11665 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11668 2018-01-08 Richard Biener <rguenther@suse.de>
11670 PR middle-end/83580
11671 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11673 2018-01-08 Richard Biener <rguenther@suse.de>
11675 PR middle-end/83517
11676 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11678 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11680 PR middle-end/81897
11681 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11682 basic blocks with a small number of successors.
11683 (convert_control_dep_chain_into_preds): Improve handling of
11685 (dump_predicates): Split apart into...
11686 (dump_pred_chain): ...here...
11687 (dump_pred_info): ...and here.
11688 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11689 (can_chain_union_be_invalidated_p): Improve check for invalidation
11691 (uninit_uses_cannot_happen): Avoid unnecessary if
11692 convert_control_dep_chain_into_preds yielded nothing.
11694 2018-01-06 Martin Sebor <msebor@redhat.com>
11696 PR tree-optimization/83640
11697 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11698 subtracting negative offset from size.
11699 (builtin_access::overlap): Adjust offset bounds of the access to fall
11700 within the size of the object if possible.
11702 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11704 PR rtl-optimization/83699
11705 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11706 extract_bit_field_as_subreg to cases in which the extracted
11707 value is also a vector.
11709 * lra-constraints.c (process_alt_operands): Test for the equivalence
11710 substitutions when detecting a possible reload cycle.
11712 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11715 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11716 by default if flag_selective_schedling{,2}. Formatting fixes.
11718 PR rtl-optimization/83682
11719 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11720 if it has non-VECTOR_MODE element mode.
11721 (vec_duplicate_p): Likewise.
11723 PR middle-end/83694
11724 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11725 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11727 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11730 * config/i386/i386-builtin.def
11731 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11732 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11733 Require also OPTION_MASK_ISA_AVX512F in addition to
11734 OPTION_MASK_ISA_GFNI.
11735 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11736 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11737 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11738 to OPTION_MASK_ISA_GFNI.
11739 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11740 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11741 OPTION_MASK_ISA_AVX512BW.
11742 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11743 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11744 addition to OPTION_MASK_ISA_GFNI.
11745 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11746 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11747 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11748 to OPTION_MASK_ISA_GFNI.
11749 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11750 a requirement for all ISAs rather than any of them with a few
11752 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11754 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11755 bitmasks to be enabled with 3 exceptions, instead of requiring any
11756 enabled ISA with lots of exceptions.
11757 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11758 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11759 Change avx512bw in isa attribute to avx512f.
11760 * config/i386/sgxintrin.h: Add license boilerplate.
11761 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11762 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11763 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11764 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11766 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11767 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11768 temporarily sse2 rather than sse if not enabled already.
11771 * config/i386/sse.md (VI248_VLBW): Rename to ...
11772 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11773 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11774 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11775 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11776 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11777 mode iterator instead of VI248_VLBW.
11779 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11781 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11782 (record_modified): Skip clobbers; add debug output.
11783 (param_change_prob): Use sreal frequencies.
11785 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11787 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11788 punt for user-aligned variables.
11790 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11792 * tree-chrec.c (chrec_contains_symbols): Return true for
11795 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11798 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11799 of (x|y) == x for BICS pattern.
11801 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11803 PR tree-optimization/83605
11804 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11805 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11808 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11810 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11811 * config/epiphany/rtems.h: New file.
11813 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11814 Uros Bizjak <ubizjak@gmail.com>
11817 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11818 QIreg_operand instead of register_operand predicate.
11819 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11820 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11821 comments instead of -fmitigate[-_]rop.
11823 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11826 * cgraphunit.c (symbol_table::compile): Switch to text_section
11827 before calling assembly_start debug hook.
11828 * run-rtl-passes.c (run_rtl_passes): Likewise.
11831 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11833 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11834 range_int_cst_p rather than !symbolic_range_p before calling
11835 extract_range_from_multiplicative_op_1.
11837 2018-01-04 Jeff Law <law@redhat.com>
11839 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11840 redundant test in assertion.
11842 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11844 * doc/rtl.texi: Document machine_mode wrapper classes.
11846 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11848 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11849 using tree_to_uhwi.
11851 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11853 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11854 the VEC_PERM_EXPR fold to fail.
11856 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11859 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11860 to switched_sections.
11862 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11865 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11866 test for d.testing.
11868 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11871 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11872 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11874 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11877 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11878 is BLKmode and bitpos not zero or mode change is needed.
11880 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11883 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11886 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11889 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11890 instead of MULT rtx. Update all corresponding splitters.
11891 (*saddl_se): Ditto.
11892 (*ssub<modesuffix>): Ditto.
11893 (*ssubl_se): Ditto.
11894 (*cmp_sadd_di): Update split patterns.
11895 (*cmp_sadd_si): Ditto.
11896 (*cmp_sadd_sidi): Ditto.
11897 (*cmp_ssub_di): Ditto.
11898 (*cmp_ssub_si): Ditto.
11899 (*cmp_ssub_sidi): Ditto.
11900 * config/alpha/predicates.md (const23_operand): New predicate.
11901 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11902 Look for ASHIFT, not MULT inner operand.
11903 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11905 2018-01-04 Martin Liska <mliska@suse.cz>
11907 PR gcov-profile/83669
11908 * gcov.c (output_intermediate_file): Add version to intermediate
11910 * doc/gcov.texi: Document new field 'version' in intermediate
11911 file format. Fix location of '-k' option of gcov command.
11913 2018-01-04 Martin Liska <mliska@suse.cz>
11916 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11918 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11920 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11922 2018-01-03 Martin Sebor <msebor@redhat.com>
11924 PR tree-optimization/83655
11925 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11926 checking calls with invalid arguments.
11928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11930 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11931 (vectorizable_mask_load_store): Delete.
11932 (vectorizable_call): Return false for masked loads and stores.
11933 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11934 instead of gimple_assign_rhs1.
11935 (vectorizable_load): Handle IFN_MASK_LOAD.
11936 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11940 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11942 (vectorizable_mask_load_store): ...here.
11943 (vectorizable_load): ...and here.
11945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11947 * tree-vect-stmts.c (vect_build_all_ones_mask)
11948 (vect_build_zero_merge_argument): New functions, split out from...
11949 (vectorizable_load): ...here.
11951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11953 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11955 (vectorizable_mask_load_store): ...here.
11956 (vectorizable_store): ...and here.
11958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11960 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11962 (vectorizable_mask_load_store): ...here.
11964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11966 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11967 (vect_model_store_cost): Take a vec_load_store_type instead of a
11969 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11970 (vect_model_store_cost): Take a vec_load_store_type instead of a
11972 (vectorizable_mask_load_store): Update accordingly.
11973 (vectorizable_store): Likewise.
11974 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11978 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11979 IFN_MASK_LOAD calls here rather than...
11980 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11983 Alan Hayward <alan.hayward@arm.com>
11984 David Sherwood <david.sherwood@arm.com>
11986 * expmed.c (extract_bit_field_1): For vector extracts,
11987 fall back to extract_bit_field_as_subreg if vec_extract
11990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11991 Alan Hayward <alan.hayward@arm.com>
11992 David Sherwood <david.sherwood@arm.com>
11994 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11995 they are variable or constant sized.
11996 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11997 slots for constant-sized data.
11999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12000 Alan Hayward <alan.hayward@arm.com>
12001 David Sherwood <david.sherwood@arm.com>
12003 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12004 handling COND_EXPRs with boolean comparisons, try to find a better
12005 basis for the mask type than the boolean itself.
12007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12009 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12010 is calculated and how it can be overridden.
12011 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12012 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12014 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12018 Alan Hayward <alan.hayward@arm.com>
12019 David Sherwood <david.sherwood@arm.com>
12021 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12022 Remove the mode argument.
12023 (aarch64_simd_valid_immediate): Remove the mode and inverse
12025 * config/aarch64/iterators.md (bitsize): New iterator.
12026 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12027 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12028 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12029 aarch64_simd_valid_immediate.
12030 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12031 (aarch64_reg_or_bic_imm): Likewise.
12032 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12033 with an insn_type enum and msl with a modifier_type enum.
12034 Replace element_width with a scalar_mode. Change the shift
12035 to unsigned int. Add constructors for scalar_float_mode and
12036 scalar_int_mode elements.
12037 (aarch64_vect_float_const_representable_p): Delete.
12038 (aarch64_can_const_movi_rtx_p)
12039 (aarch64_simd_scalar_immediate_valid_for_move)
12040 (aarch64_simd_make_constant): Update call to
12041 aarch64_simd_valid_immediate.
12042 (aarch64_advsimd_valid_immediate_hs): New function.
12043 (aarch64_advsimd_valid_immediate): Likewise.
12044 (aarch64_simd_valid_immediate): Remove mode and inverse
12045 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12046 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12047 and aarch64_float_const_representable_p on the result.
12048 (aarch64_output_simd_mov_immediate): Remove mode argument.
12049 Update call to aarch64_simd_valid_immediate and use of
12050 simd_immediate_info.
12051 (aarch64_output_scalar_simd_mov_immediate): Update call
12054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12055 Alan Hayward <alan.hayward@arm.com>
12056 David Sherwood <david.sherwood@arm.com>
12058 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12059 (mode_nunits): Likewise CONST_MODE_NUNITS.
12060 * machmode.def (ADJUST_NUNITS): Document.
12061 * genmodes.c (mode_data::need_nunits_adj): New field.
12062 (blank_mode): Update accordingly.
12063 (adj_nunits): New variable.
12064 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12066 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12067 listed in adj_nunits.
12068 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12069 listed in adj_nunits. Don't emit case statements for such modes.
12070 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12071 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12072 nothing if adj_nunits is nonnull.
12073 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12074 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12075 (emit_mode_fbit): Update use of print_maybe_const_decl.
12076 (emit_move_size): Likewise. Treat the array as non-const
12078 (emit_mode_adjustments): Handle adj_nunits.
12080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12082 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12083 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12084 (VECTOR_MODES): Use it.
12085 (make_vector_modes): Take the prefix as an argument.
12087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12088 Alan Hayward <alan.hayward@arm.com>
12089 David Sherwood <david.sherwood@arm.com>
12091 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12092 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12093 for MODE_VECTOR_BOOL.
12094 * machmode.def (VECTOR_BOOL_MODE): Document.
12095 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12096 (make_vector_bool_mode): New function.
12097 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12099 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12100 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12102 * stor-layout.c (int_mode_for_mode): Likewise.
12103 * tree.c (build_vector_type_for_mode): Likewise.
12104 * varasm.c (output_constant_pool_2): Likewise.
12105 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12106 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12107 for MODE_VECTOR_BOOL.
12108 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12109 of mode class checks.
12110 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12111 instead of a list of mode class checks.
12112 (expand_vector_scalar_condition): Likewise.
12113 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12116 Alan Hayward <alan.hayward@arm.com>
12117 David Sherwood <david.sherwood@arm.com>
12119 * machmode.h (mode_size): Change from unsigned short to
12121 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12122 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12123 or if measurement_type is not polynomial.
12124 (fixed_size_mode::includes_p): Check for constant-sized modes.
12125 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12126 return a poly_uint16 rather than an unsigned short.
12127 (emit_mode_size): Change the type of mode_size from unsigned short
12128 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12129 (emit_mode_adjustments): Cope with polynomial vector sizes.
12130 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12132 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12134 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12135 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12136 * caller-save.c (setup_save_areas): Likewise.
12137 (replace_reg_with_saved_mem): Likewise.
12138 * calls.c (emit_library_call_value_1): Likewise.
12139 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12140 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12141 (gen_lowpart_for_combine): Likewise.
12142 * convert.c (convert_to_integer_1): Likewise.
12143 * cse.c (equiv_constant, cse_insn): Likewise.
12144 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12145 (cselib_subst_to_values): Likewise.
12146 * dce.c (word_dce_process_block): Likewise.
12147 * df-problems.c (df_word_lr_mark_ref): Likewise.
12148 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12149 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12150 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12151 (rtl_for_decl_location): Likewise.
12152 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12153 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12154 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12155 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12156 (expand_expr_real_1): Likewise.
12157 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12158 (pad_below): Likewise.
12159 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12160 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12161 * ira.c (get_subreg_tracking_sizes): Likewise.
12162 * ira-build.c (ira_create_allocno_objects): Likewise.
12163 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12164 (ira_sort_regnos_for_alter_reg): Likewise.
12165 * ira-costs.c (record_operand_costs): Likewise.
12166 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12167 (resolve_simple_move): Likewise.
12168 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12169 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12170 (lra_constraints): Likewise.
12171 (CONST_POOL_OK_P): Reject variable-sized modes.
12172 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12173 (add_pseudo_to_slot, lra_spill): Likewise.
12174 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12175 * optabs-query.c (get_best_extraction_insn): Likewise.
12176 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12177 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12178 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12179 * recog.c (offsettable_address_addr_space_p): Likewise.
12180 * regcprop.c (maybe_mode_change): Likewise.
12181 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12182 * regrename.c (build_def_use): Likewise.
12183 * regstat.c (dump_reg_info): Likewise.
12184 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12185 (find_reloads, find_reloads_subreg_address): Likewise.
12186 * reload1.c (eliminate_regs_1): Likewise.
12187 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12188 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12189 (simplify_binary_operation_1, simplify_subreg): Likewise.
12190 * targhooks.c (default_function_arg_padding): Likewise.
12191 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12192 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12193 (verify_gimple_assign_ternary): Likewise.
12194 * tree-inline.c (estimate_move_cost): Likewise.
12195 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12196 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12197 (get_address_cost_ainc): Likewise.
12198 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12199 (vect_supportable_dr_alignment): Likewise.
12200 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12201 (vectorizable_reduction): Likewise.
12202 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12203 (vectorizable_operation, vectorizable_load): Likewise.
12204 * tree.c (build_same_sized_truth_vector_type): Likewise.
12205 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12206 * var-tracking.c (emit_note_insn_var_location): Likewise.
12207 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12208 (ADDR_VEC_ALIGN): Likewise.
12210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12211 Alan Hayward <alan.hayward@arm.com>
12212 David Sherwood <david.sherwood@arm.com>
12214 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12216 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12217 or if measurement_type is polynomial.
12218 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12219 * combine.c (make_extraction): Likewise.
12220 * dse.c (find_shift_sequence): Likewise.
12221 * dwarf2out.c (mem_loc_descriptor): Likewise.
12222 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12223 (extract_bit_field, extract_low_bits): Likewise.
12224 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12225 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12226 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12227 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12228 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12229 * reload.c (find_reloads): Likewise.
12230 * reload1.c (alter_reg): Likewise.
12231 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12232 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12233 * tree-if-conv.c (predicate_mem_writes): Likewise.
12234 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12235 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12236 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12237 * valtrack.c (dead_debug_insert_temp): Likewise.
12238 * varasm.c (mergeable_constant_section): Likewise.
12239 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12242 Alan Hayward <alan.hayward@arm.com>
12243 David Sherwood <david.sherwood@arm.com>
12245 * expr.c (expand_assignment): Cope with polynomial mode sizes
12246 when assigning to a CONCAT.
12248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12249 Alan Hayward <alan.hayward@arm.com>
12250 David Sherwood <david.sherwood@arm.com>
12252 * machmode.h (mode_precision): Change from unsigned short to
12254 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12256 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12257 or if measurement_type is not polynomial.
12258 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12259 in which the mode is already known to be a scalar_int_mode.
12260 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12261 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12263 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12264 for GET_MODE_PRECISION.
12265 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12266 for GET_MODE_PRECISION.
12267 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12269 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12270 (expand_field_assignment, make_extraction): Likewise.
12271 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12272 (get_last_value): Likewise.
12273 * convert.c (convert_to_integer_1): Likewise.
12274 * cse.c (cse_insn): Likewise.
12275 * expr.c (expand_expr_real_1): Likewise.
12276 * lra-constraints.c (simplify_operand_subreg): Likewise.
12277 * optabs-query.c (can_atomic_load_p): Likewise.
12278 * optabs.c (expand_atomic_load): Likewise.
12279 (expand_atomic_store): Likewise.
12280 * ree.c (combine_reaching_defs): Likewise.
12281 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12282 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12283 * tree.h (type_has_mode_precision_p): Likewise.
12284 * ubsan.c (instrument_si_overflow): Likewise.
12286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12287 Alan Hayward <alan.hayward@arm.com>
12288 David Sherwood <david.sherwood@arm.com>
12290 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12291 polynomial numbers of units.
12292 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12293 (valid_vector_subparts_p): New function.
12294 (build_vector_type): Remove temporary shim and take the number
12295 of units as a poly_uint64 rather than an int.
12296 (build_opaque_vector_type): Take the number of units as a
12297 poly_uint64 rather than an int.
12298 * tree.c (build_vector_from_ctor): Handle polynomial
12299 TYPE_VECTOR_SUBPARTS.
12300 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12301 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12302 (build_vector_from_val): If the number of units is variable,
12303 use build_vec_duplicate_cst for constant operands and
12304 VEC_DUPLICATE_EXPR otherwise.
12305 (make_vector_type): Remove temporary is_constant ().
12306 (build_vector_type, build_opaque_vector_type): Take the number of
12307 units as a poly_uint64 rather than an int.
12308 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12310 * cfgexpand.c (expand_debug_expr): Likewise.
12311 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12312 (store_constructor, expand_expr_real_1): Likewise.
12313 (const_scalar_mask_from_tree): Likewise.
12314 * fold-const-call.c (fold_const_reduction): Likewise.
12315 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12316 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12317 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12318 (fold_relational_const): Likewise.
12319 (native_interpret_vector): Likewise. Change the size from an
12320 int to an unsigned int.
12321 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12322 TYPE_VECTOR_SUBPARTS.
12323 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12324 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12325 duplicating a non-constant operand into a variable-length vector.
12326 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12327 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12328 * ipa-icf.c (sem_variable::equals): Likewise.
12329 * match.pd: Likewise.
12330 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12331 * print-tree.c (print_node): Likewise.
12332 * stor-layout.c (layout_type): Likewise.
12333 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12334 * tree-cfg.c (verify_gimple_comparison): Likewise.
12335 (verify_gimple_assign_binary): Likewise.
12336 (verify_gimple_assign_ternary): Likewise.
12337 (verify_gimple_assign_single): Likewise.
12338 * tree-pretty-print.c (dump_generic_node): Likewise.
12339 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12340 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12341 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12342 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12343 (vect_shift_permute_load_chain): Likewise.
12344 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12345 (expand_vector_condition, optimize_vector_constructor): Likewise.
12346 (lower_vec_perm, get_compute_type): Likewise.
12347 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12348 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12349 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12350 (vect_recog_mask_conversion_pattern): Likewise.
12351 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12352 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12353 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12354 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12355 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12356 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12357 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12358 (supportable_widening_operation): Likewise.
12359 (supportable_narrowing_operation): Likewise.
12360 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12362 * varasm.c (output_constant): Likewise.
12364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12365 Alan Hayward <alan.hayward@arm.com>
12366 David Sherwood <david.sherwood@arm.com>
12368 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12369 so that both the length == 3 and length != 3 cases set up their
12370 own permute vectors. Add comments explaining why we know the
12371 number of elements is constant.
12372 (vect_permute_load_chain): Likewise.
12374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12375 Alan Hayward <alan.hayward@arm.com>
12376 David Sherwood <david.sherwood@arm.com>
12378 * machmode.h (mode_nunits): Change from unsigned char to
12380 (ONLY_FIXED_SIZE_MODES): New macro.
12381 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12382 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12383 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12385 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12386 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12387 or if measurement_type is not polynomial.
12388 * genmodes.c (ZERO_COEFFS): New macro.
12389 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12391 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12392 Use ZERO_COEFFS when emitting initializers.
12393 * data-streamer.h (bp_pack_poly_value): New function.
12394 (bp_unpack_poly_value): Likewise.
12395 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12396 for GET_MODE_NUNITS.
12397 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12398 for GET_MODE_NUNITS.
12399 * tree.c (make_vector_type): Remove temporary shim and make
12400 the real function take the number of units as a poly_uint64
12401 rather than an int.
12402 (build_vector_type_for_mode): Handle polynomial nunits.
12403 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12404 * emit-rtl.c (const_vec_series_p_1): Likewise.
12405 (gen_rtx_CONST_VECTOR): Likewise.
12406 * fold-const.c (test_vec_duplicate_folding): Likewise.
12407 * genrecog.c (validate_pattern): Likewise.
12408 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12409 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12410 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12411 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12412 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12413 * rtlanal.c (subreg_get_info): Likewise.
12414 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12415 (vect_grouped_load_supported): Likewise.
12416 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12417 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12418 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12419 (simplify_const_unary_operation, simplify_binary_operation_1)
12420 (simplify_const_binary_operation, simplify_ternary_operation)
12421 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12422 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12423 instead of CONST_VECTOR_NUNITS.
12424 * varasm.c (output_constant_pool_2): Likewise.
12425 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12426 explicit-encoded elements in the XVEC for variable-length vectors.
12428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12430 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12433 Alan Hayward <alan.hayward@arm.com>
12434 David Sherwood <david.sherwood@arm.com>
12436 * coretypes.h (fixed_size_mode): Declare.
12437 (fixed_size_mode_pod): New typedef.
12438 * builtins.h (target_builtins::x_apply_args_mode)
12439 (target_builtins::x_apply_result_mode): Change type to
12440 fixed_size_mode_pod.
12441 * builtins.c (apply_args_size, apply_result_size, result_vector)
12442 (expand_builtin_apply_args_1, expand_builtin_apply)
12443 (expand_builtin_return): Update accordingly.
12445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12447 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12448 * cselib.c (cselib_hash_rtx): Likewise.
12449 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12450 CONST_VECTOR encoding.
12452 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12453 Jeff Law <law@redhat.com>
12456 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12457 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12458 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12459 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12462 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12463 explicitly probe *sp in a noreturn function if there were any callee
12464 register saves or frame pointer is needed.
12466 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12469 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12470 BLKmode for ternary, binary or unary expressions.
12473 * var-tracking.c (delete_vta_debug_insn): New inline function.
12474 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12475 insns from get_insns () to NULL instead of each bb separately.
12476 Use delete_vta_debug_insn. No longer static.
12477 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12478 delete_vta_debug_insns callers.
12479 * rtl.h (delete_vta_debug_insns): Declare.
12480 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12481 instead of variable_tracking_main.
12483 2018-01-03 Martin Sebor <msebor@redhat.com>
12485 PR tree-optimization/83603
12486 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12487 arguments past the endof the argument list in functions declared
12488 without a prototype.
12489 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12490 Avoid checking when arguments are null.
12492 2018-01-03 Martin Sebor <msebor@redhat.com>
12495 * doc/extend.texi (attribute const): Fix a typo.
12496 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12497 issuing -Wsuggest-attribute for void functions.
12499 2018-01-03 Martin Sebor <msebor@redhat.com>
12501 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12502 offset_int::from instead of wide_int::to_shwi.
12503 (maybe_diag_overlap): Remove assertion.
12504 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12505 * gimple-ssa-sprintf.c (format_directive): Same.
12506 (parse_directive): Same.
12507 (sprintf_dom_walker::compute_format_length): Same.
12508 (try_substitute_return_value): Same.
12510 2018-01-03 Jeff Law <law@redhat.com>
12512 PR middle-end/83654
12513 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12514 non-constant residual for zero at runtime and avoid probing in
12515 that case. Reorganize code for trailing problem to mirror handling
12518 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12520 PR tree-optimization/83501
12521 * tree-ssa-strlen.c (get_string_cst): New.
12522 (handle_char_store): Call get_string_cst.
12524 2018-01-03 Martin Liska <mliska@suse.cz>
12526 PR tree-optimization/83593
12527 * tree-ssa-strlen.c: Include tree-cfg.h.
12528 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12529 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12530 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12532 (strlen_dom_walker::before_dom_children): Call
12533 gimple_purge_dead_eh_edges. Dump tranformation with details
12535 (strlen_dom_walker::before_dom_children): Update call by adding
12536 new argument cleanup_eh.
12537 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12539 2018-01-03 Martin Liska <mliska@suse.cz>
12542 * cif-code.def (VARIADIC_THUNK): New enum value.
12543 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12546 2018-01-03 Jan Beulich <jbeulich@suse.com>
12548 * sse.md (mov<mode>_internal): Tighten condition for when to use
12549 vmovdqu<ssescalarsize> for TI and OI modes.
12551 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12553 Update copyright years.
12555 2018-01-03 Martin Liska <mliska@suse.cz>
12558 * ipa-visibility.c (function_and_variable_visibility): Skip
12559 functions with noipa attribure.
12561 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12563 * gcc.c (process_command): Update copyright notice dates.
12564 * gcov-dump.c (print_version): Ditto.
12565 * gcov.c (print_version): Ditto.
12566 * gcov-tool.c (print_version): Ditto.
12567 * gengtype.c (create_file): Ditto.
12568 * doc/cpp.texi: Bump @copying's copyright year.
12569 * doc/cppinternals.texi: Ditto.
12570 * doc/gcc.texi: Ditto.
12571 * doc/gccint.texi: Ditto.
12572 * doc/gcov.texi: Ditto.
12573 * doc/install.texi: Ditto.
12574 * doc/invoke.texi: Ditto.
12576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12578 * vector-builder.h (vector_builder::m_full_nelts): Change from
12579 unsigned int to poly_uint64.
12580 (vector_builder::full_nelts): Update prototype accordingly.
12581 (vector_builder::new_vector): Likewise.
12582 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12583 (vector_builder::operator ==): Likewise.
12584 (vector_builder::finalize): Likewise.
12585 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12586 Take the number of elements as a poly_uint64 rather than an
12588 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12589 from unsigned int to poly_uint64.
12590 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12591 (vec_perm_indices::new_vector): Likewise.
12592 (vec_perm_indices::length): Likewise.
12593 (vec_perm_indices::nelts_per_input): Likewise.
12594 (vec_perm_indices::input_nelts): Likewise.
12595 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12596 number of elements per input as a poly_uint64 rather than an
12597 unsigned int. Use the original encoding for variable-length
12598 vectors, rather than clamping each individual element.
12599 For the second and subsequent elements in each pattern,
12600 clamp the step and base before clamping their sum.
12601 (vec_perm_indices::series_p): Handle polynomial element counts.
12602 (vec_perm_indices::all_in_range_p): Likewise.
12603 (vec_perm_indices_to_tree): Likewise.
12604 (vec_perm_indices_to_rtx): Likewise.
12605 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12606 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12607 (tree_vector_builder::new_binary_operation): Handle polynomial
12608 element counts. Return false if we need to know the number
12609 of elements at compile time.
12610 * fold-const.c (fold_vec_perm): Punt if the number of elements
12611 isn't known at compile time.
12613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12615 * vec-perm-indices.h (vec_perm_builder): Change element type
12616 from HOST_WIDE_INT to poly_int64.
12617 (vec_perm_indices::element_type): Update accordingly.
12618 (vec_perm_indices::clamp): Handle polynomial element_types.
12619 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12620 (vec_perm_indices::all_in_range_p): Likewise.
12621 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12623 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12624 polynomial vec_perm_indices element types.
12625 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12626 * fold-const.c (fold_vec_perm): Likewise.
12627 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12628 * tree-vect-generic.c (lower_vec_perm): Likewise.
12629 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12630 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12631 element type to HOST_WIDE_INT.
12633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12634 Alan Hayward <alan.hayward@arm.com>
12635 David Sherwood <david.sherwood@arm.com>
12637 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12638 rather than an int. Use plus_constant.
12639 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12640 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12643 Alan Hayward <alan.hayward@arm.com>
12644 David Sherwood <david.sherwood@arm.com>
12646 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12647 a HOST_WIDE_INT to a poly_int64.
12649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12650 Alan Hayward <alan.hayward@arm.com>
12651 David Sherwood <david.sherwood@arm.com>
12653 * calls.c (load_register_parameters): Cope with polynomial
12654 mode sizes. Require a constant size for BLKmode parameters
12655 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12656 forces a parameter to be padded at the lsb end in order to
12657 fill a complete number of words, require the parameter size
12658 to be ordered wrt UNITS_PER_WORD.
12660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12661 Alan Hayward <alan.hayward@arm.com>
12662 David Sherwood <david.sherwood@arm.com>
12664 * reload1.c (spill_stack_slot_width): Change element type
12665 from unsigned int to poly_uint64_pod.
12666 (alter_reg): Treat mode sizes as polynomial.
12668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12669 Alan Hayward <alan.hayward@arm.com>
12670 David Sherwood <david.sherwood@arm.com>
12672 * reload.c (complex_word_subreg_p): New function.
12673 (reload_inner_reg_of_subreg, push_reload): Use it.
12675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12676 Alan Hayward <alan.hayward@arm.com>
12677 David Sherwood <david.sherwood@arm.com>
12679 * lra-constraints.c (process_alt_operands): Reject matched
12680 operands whose sizes aren't ordered.
12681 (match_reload): Refer to this check here.
12683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12684 Alan Hayward <alan.hayward@arm.com>
12685 David Sherwood <david.sherwood@arm.com>
12687 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12688 that the mode size is in the set {1, 2, 4, 8, 16}.
12690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12691 Alan Hayward <alan.hayward@arm.com>
12692 David Sherwood <david.sherwood@arm.com>
12694 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12695 Use plus_constant instead of gen_rtx_PLUS.
12697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12698 Alan Hayward <alan.hayward@arm.com>
12699 David Sherwood <david.sherwood@arm.com>
12701 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12702 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12703 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12704 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12705 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12706 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12707 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12708 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12709 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12710 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12712 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12713 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12714 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12715 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12716 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12717 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12718 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12719 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12720 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12721 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12723 * expr.c (emit_move_resolve_push): Treat the input and result
12724 of PUSH_ROUNDING as a poly_int64.
12725 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12726 (emit_push_insn): Likewise.
12727 * lra-eliminations.c (mark_not_eliminable): Likewise.
12728 * recog.c (push_operand): Likewise.
12729 * reload1.c (elimination_effects): Likewise.
12730 * rtlanal.c (nonzero_bits1): Likewise.
12731 * calls.c (store_one_arg): Likewise. Require the padding to be
12732 known at compile time.
12734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12735 Alan Hayward <alan.hayward@arm.com>
12736 David Sherwood <david.sherwood@arm.com>
12738 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12739 Use plus_constant instead of gen_rtx_PLUS.
12741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12742 Alan Hayward <alan.hayward@arm.com>
12743 David Sherwood <david.sherwood@arm.com>
12745 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12746 rather than an int.
12748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12749 Alan Hayward <alan.hayward@arm.com>
12750 David Sherwood <david.sherwood@arm.com>
12752 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12753 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12754 via stack temporaries. Treat the mode size as polynomial too.
12756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12757 Alan Hayward <alan.hayward@arm.com>
12758 David Sherwood <david.sherwood@arm.com>
12760 * expr.c (expand_expr_real_2): When handling conversions involving
12761 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12762 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12763 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12772 Alan Hayward <alan.hayward@arm.com>
12773 David Sherwood <david.sherwood@arm.com>
12775 * combine.c (can_change_dest_mode): Handle polynomial
12776 REGMODE_NATURAL_SIZE.
12777 * expmed.c (store_bit_field_1): Likewise.
12778 * expr.c (store_constructor): Likewise.
12779 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12780 and polynomial REGMODE_NATURAL_SIZE.
12781 (gen_lowpart_common): Likewise.
12782 * reginfo.c (record_subregs_of_mode): Likewise.
12783 * rtlanal.c (read_modify_subreg_p): Likewise.
12785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12786 Alan Hayward <alan.hayward@arm.com>
12787 David Sherwood <david.sherwood@arm.com>
12789 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12790 numbers of elements.
12792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12793 Alan Hayward <alan.hayward@arm.com>
12794 David Sherwood <david.sherwood@arm.com>
12796 * match.pd: Cope with polynomial numbers of vector elements.
12798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12799 Alan Hayward <alan.hayward@arm.com>
12800 David Sherwood <david.sherwood@arm.com>
12802 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12803 in a POINTER_PLUS_EXPR.
12805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12806 Alan Hayward <alan.hayward@arm.com>
12807 David Sherwood <david.sherwood@arm.com>
12809 * omp-simd-clone.c (simd_clone_subparts): New function.
12810 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12811 (ipa_simd_modify_function_body): Likewise.
12813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12814 Alan Hayward <alan.hayward@arm.com>
12815 David Sherwood <david.sherwood@arm.com>
12817 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12818 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12819 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12820 (expand_vector_condition, vector_element): Likewise.
12821 (subparts_gt): New function.
12822 (get_compute_type): Use subparts_gt.
12823 (count_type_subparts): Delete.
12824 (expand_vector_operations_1): Use subparts_gt instead of
12825 count_type_subparts.
12827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12828 Alan Hayward <alan.hayward@arm.com>
12829 David Sherwood <david.sherwood@arm.com>
12831 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12832 (vect_compile_time_alias): ...this new function. Do the calculation
12833 on poly_ints rather than trees.
12834 (vect_prune_runtime_alias_test_list): Update call accordingly.
12836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12837 Alan Hayward <alan.hayward@arm.com>
12838 David Sherwood <david.sherwood@arm.com>
12840 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12842 (vect_schedule_slp_instance): Likewise.
12844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12845 Alan Hayward <alan.hayward@arm.com>
12846 David Sherwood <david.sherwood@arm.com>
12848 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12849 constant and extern definitions for variable-length vectors.
12850 (vect_get_constant_vectors): Note that the number of units
12851 is known to be constant.
12853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12854 Alan Hayward <alan.hayward@arm.com>
12855 David Sherwood <david.sherwood@arm.com>
12857 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12858 of units as polynomial. Choose between WIDE and NARROW based
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 * tree-vect-stmts.c (simd_clone_subparts): New function.
12866 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12869 Alan Hayward <alan.hayward@arm.com>
12870 David Sherwood <david.sherwood@arm.com>
12872 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12873 vectors as polynomial. Use build_index_vector for
12874 IFN_GOMP_SIMD_LANE.
12876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12877 Alan Hayward <alan.hayward@arm.com>
12878 David Sherwood <david.sherwood@arm.com>
12880 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12881 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12882 for variable-length vectors.
12883 (vectorizable_mask_load_store): Treat the number of units as
12884 polynomial, asserting that it is constant if the condition has
12885 already been enforced.
12886 (vectorizable_store, vectorizable_load): Likewise.
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 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12893 of units as polynomial. Punt if we can't tell at compile time
12894 which vector contains the final result.
12896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12897 Alan Hayward <alan.hayward@arm.com>
12898 David Sherwood <david.sherwood@arm.com>
12900 * tree-vect-loop.c (vectorizable_induction): Treat the number
12901 of units as polynomial. Punt on SLP inductions. Use an integer
12902 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12903 cast of such a series for variable-length floating-point
12906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12907 Alan Hayward <alan.hayward@arm.com>
12908 David Sherwood <david.sherwood@arm.com>
12910 * tree.h (build_index_vector): Declare.
12911 * tree.c (build_index_vector): New function.
12912 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12913 of units as polynomial, forcibly converting it to a constant if
12914 vectorizable_reduction has already enforced the condition.
12915 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12916 to create a {1,2,3,...} vector.
12917 (vectorizable_reduction): Treat the number of units as polynomial.
12918 Choose vectype_in based on the largest scalar element size rather
12919 than the smallest number of units. Enforce the restrictions
12922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12923 Alan Hayward <alan.hayward@arm.com>
12924 David Sherwood <david.sherwood@arm.com>
12926 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12927 number of units as polynomial.
12929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12930 Alan Hayward <alan.hayward@arm.com>
12931 David Sherwood <david.sherwood@arm.com>
12933 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12934 * target.def (autovectorize_vector_sizes): Return the vector sizes
12935 by pointer, using vector_sizes rather than a bitmask.
12936 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12937 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12938 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12940 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12941 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12942 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12943 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12944 * omp-general.c (omp_max_vf): Likewise.
12945 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12946 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12947 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12948 * tree-vect-slp.c (vect_slp_bb): Likewise.
12949 * doc/tm.texi: Regenerate.
12950 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12952 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12953 the vector size as a poly_uint64 rather than an unsigned int.
12954 (current_vector_size): Change from an unsigned int to a poly_uint64.
12955 (get_vectype_for_scalar_type): Update accordingly.
12956 * tree.h (build_truth_vector_type): Take the size and number of
12957 units as a poly_uint64 rather than an unsigned int.
12958 (build_vector_type): Add a temporary overload that takes
12959 the number of units as a poly_uint64 rather than an unsigned int.
12960 * tree.c (make_vector_type): Likewise.
12961 (build_truth_vector_type): Take the number of units as a poly_uint64
12962 rather than an unsigned int.
12964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12965 Alan Hayward <alan.hayward@arm.com>
12966 David Sherwood <david.sherwood@arm.com>
12968 * target.def (get_mask_mode): Take the number of units and length
12969 as poly_uint64s rather than unsigned ints.
12970 * targhooks.h (default_get_mask_mode): Update accordingly.
12971 * targhooks.c (default_get_mask_mode): Likewise.
12972 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12973 * doc/tm.texi: Regenerate.
12975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12976 Alan Hayward <alan.hayward@arm.com>
12977 David Sherwood <david.sherwood@arm.com>
12979 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12980 * omp-general.c (omp_max_vf): Likewise.
12981 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12982 (expand_omp_simd): Handle polynomial safelen.
12983 * omp-low.c (omplow_simd_context): Add a default constructor.
12984 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12985 (lower_rec_simd_input_clauses): Update accordingly.
12986 (lower_rec_input_clauses): Likewise.
12988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12989 Alan Hayward <alan.hayward@arm.com>
12990 David Sherwood <david.sherwood@arm.com>
12992 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12993 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12994 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12995 (vect_analyze_slp_cost): Likewise.
12996 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12997 (vect_model_load_cost): Likewise.
12999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13000 Alan Hayward <alan.hayward@arm.com>
13001 David Sherwood <david.sherwood@arm.com>
13003 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13004 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13005 from an unsigned int * to a poly_uint64_pod *.
13006 (calculate_unrolling_factor): New function.
13007 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13010 Alan Hayward <alan.hayward@arm.com>
13011 David Sherwood <david.sherwood@arm.com>
13013 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13014 from an unsigned int to a poly_uint64.
13015 (_loop_vec_info::slp_unrolling_factor): Likewise.
13016 (_loop_vec_info::vectorization_factor): Change from an int
13018 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13019 (vect_get_num_vectors): New function.
13020 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13021 (vect_get_num_copies): Use vect_get_num_vectors.
13022 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13023 to an unsigned int *.
13024 (vect_analyze_data_refs): Change min_vf from an int * to a
13026 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13027 than an unsigned HOST_WIDE_INT.
13028 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13029 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13030 to an unsigned int *.
13031 (vect_analyze_data_ref_dependences): Likewise.
13032 (vect_compute_data_ref_alignment): Handle polynomial vf.
13033 (vect_enhance_data_refs_alignment): Likewise.
13034 (vect_prune_runtime_alias_test_list): Likewise.
13035 (vect_shift_permute_load_chain): Likewise.
13036 (vect_supportable_dr_alignment): Likewise.
13037 (dependence_distance_ge_vf): Take the vectorization factor as a
13038 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13039 (vect_analyze_data_refs): Change min_vf from an int * to a
13041 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13042 vfm1 as a poly_uint64 rather than an int. Make the same change
13043 for the returned bound_scalar.
13044 (vect_gen_vector_loop_niters): Handle polynomial vf.
13045 (vect_do_peeling): Likewise. Update call to
13046 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13047 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13049 * tree-vect-loop.c (vect_determine_vectorization_factor)
13050 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13051 (vect_get_known_peeling_cost): Likewise.
13052 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13053 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13054 (vect_transform_loop): Likewise. Use the lowest possible VF when
13055 updating the upper bounds of the loop.
13056 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13057 rather than an int.
13058 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13059 polynomial unroll factors.
13060 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13061 (vect_make_slp_decision): Likewise.
13062 (vect_supported_load_permutation_p): Likewise, and polynomial
13064 (vect_analyze_slp_cost): Handle polynomial vf.
13065 (vect_slp_analyze_node_operations): Likewise.
13066 (vect_slp_analyze_bb_1): Likewise.
13067 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13068 than an unsigned HOST_WIDE_INT.
13069 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13070 (vectorizable_load): Handle polynomial vf.
13071 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13073 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13076 Alan Hayward <alan.hayward@arm.com>
13077 David Sherwood <david.sherwood@arm.com>
13079 * match.pd: Handle bit operations involving three constants
13080 and try to fold one pair.
13082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13084 * tree-vect-loop-manip.c: Include gimple-fold.h.
13085 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13086 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13087 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13088 Add a path that uses a step of VF instead of 1, but disable it
13090 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13091 and niters_no_overflow parameters. Update calls to
13092 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13093 Create a new SSA name if the latter choses to use a ste other
13094 than zero, and return it via niters_vector_mult_vf_var.
13095 * tree-vect-loop.c (vect_transform_loop): Update calls to
13096 vect_do_peeling, vect_gen_vector_loop_niters and
13097 slpeel_make_loop_iterate_ntimes.
13098 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13099 (vect_gen_vector_loop_niters): Update declarations after above changes.
13101 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13103 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13104 128-bit round to integer instructions.
13105 (ceil<mode>2): Likewise.
13106 (btrunc<mode>2): Likewise.
13107 (round<mode>2): Likewise.
13109 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13111 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13112 unaligned VSX load/store on P8/P9.
13113 (expand_block_clear): Allow the use of unaligned VSX
13114 load/store on P8/P9.
13116 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13118 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13120 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13121 swap associated with both a load and a store.
13123 2018-01-02 Andrew Waterman <andrew@sifive.com>
13125 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13126 * config/riscv/riscv.md (clear_cache): Use it.
13128 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13130 * web.c: Remove out-of-date comment.
13132 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13134 * expr.c (fixup_args_size_notes): Check that any existing
13135 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13136 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13137 (emit_single_push_insn): ...here.
13139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13141 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13142 (const_vector_encoded_nelts): New function.
13143 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13144 (const_vector_int_elt, const_vector_elt): Declare.
13145 * emit-rtl.c (const_vector_int_elt_1): New function.
13146 (const_vector_elt): Likewise.
13147 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13148 of CONST_VECTOR_ELT.
13150 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13152 * expr.c: Include rtx-vector-builder.h.
13153 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13154 directly on the tree encoding.
13155 (const_vector_from_tree): Likewise.
13156 * optabs.c: Include rtx-vector-builder.h.
13157 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13158 sequence of "u" values.
13159 * vec-perm-indices.c: Include rtx-vector-builder.h.
13160 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13161 directly on the vec_perm_indices encoding.
13163 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13165 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13166 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13167 * rtx-vector-builder.h: New file.
13168 * rtx-vector-builder.c: Likewise.
13169 * rtl.h (rtx_def::u2): Add a const_vector field.
13170 (CONST_VECTOR_NPATTERNS): New macro.
13171 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13172 (CONST_VECTOR_DUPLICATE_P): Likewise.
13173 (CONST_VECTOR_STEPPED_P): Likewise.
13174 (CONST_VECTOR_ENCODED_ELT): Likewise.
13175 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13176 (unwrap_const_vec_duplicate): Likewise.
13177 (const_vec_series_p): Check for a non-duplicated vector encoding.
13178 Say that the function only returns true for integer vectors.
13179 * emit-rtl.c: Include rtx-vector-builder.h.
13180 (gen_const_vec_duplicate_1): Delete.
13181 (gen_const_vector): Call gen_const_vec_duplicate instead of
13182 gen_const_vec_duplicate_1.
13183 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13184 (gen_const_vec_duplicate): Use rtx_vector_builder.
13185 (gen_const_vec_series): Likewise.
13186 (gen_rtx_CONST_VECTOR): Likewise.
13187 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13188 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13189 Build a new vector rather than modifying a CONST_VECTOR in-place.
13190 (handle_special_swappables): Update call accordingly.
13191 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13192 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13193 Build a new vector rather than modifying a CONST_VECTOR in-place.
13194 (handle_special_swappables): Update call accordingly.
13196 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13198 * simplify-rtx.c (simplify_const_binary_operation): Use
13199 CONST_VECTOR_ELT instead of XVECEXP.
13201 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13203 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13204 the selector elements to be different from the data elements
13205 if the selector is a VECTOR_CST.
13206 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13207 ssizetype for the selector.
13209 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13211 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13212 before testing each element individually.
13213 * tree-vect-generic.c (lower_vec_perm): Likewise.
13215 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13217 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13218 * selftest-run-tests.c (selftest::run_tests): Call it.
13219 * vector-builder.h (vector_builder::operator ==): New function.
13220 (vector_builder::operator !=): Likewise.
13221 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13222 (vec_perm_indices::all_from_input_p): New function.
13223 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13224 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13225 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13226 instead of reading the VECTOR_CST directly. Detect whether both
13227 vector inputs are the same before constructing the vec_perm_indices,
13228 and update the number of inputs argument accordingly. Use the
13229 utility functions added above. Only construct sel2 if we need to.
13231 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13233 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13234 the broadcast of the low byte.
13235 (expand_mult_highpart): Use an explicit encoding for the permutes.
13236 * optabs-query.c (can_mult_highpart_p): Likewise.
13237 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13238 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13239 (vectorizable_bswap): Likewise.
13240 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13241 explicit encoding for the power-of-2 permutes.
13242 (vect_permute_store_chain): Likewise.
13243 (vect_grouped_load_supported): Likewise.
13244 (vect_permute_load_chain): Likewise.
13246 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13248 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13249 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13250 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13251 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13252 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13253 (vect_gen_perm_mask_any): Likewise.
13255 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13257 * int-vector-builder.h: New file.
13258 * vec-perm-indices.h: Include int-vector-builder.h.
13259 (vec_perm_indices): Redefine as an int_vector_builder.
13260 (auto_vec_perm_indices): Delete.
13261 (vec_perm_builder): Redefine as a stand-alone class.
13262 (vec_perm_indices::vec_perm_indices): New function.
13263 (vec_perm_indices::clamp): Likewise.
13264 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13265 (vec_perm_indices::new_vector): New function.
13266 (vec_perm_indices::new_expanded_vector): Update for new
13267 vec_perm_indices class.
13268 (vec_perm_indices::rotate_inputs): New function.
13269 (vec_perm_indices::all_in_range_p): Operate directly on the
13270 encoded form, without computing elided elements.
13271 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13272 encoding. Update for new vec_perm_indices class.
13273 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13274 the given vec_perm_builder.
13275 (expand_vec_perm_var): Update vec_perm_builder constructor.
13276 (expand_mult_highpart): Use vec_perm_builder instead of
13277 auto_vec_perm_indices.
13278 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13279 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13280 or double series encoding as appropriate.
13281 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13282 vec_perm_indices instead of auto_vec_perm_indices.
13283 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13284 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13285 (vect_permute_store_chain): Likewise.
13286 (vect_grouped_load_supported): Likewise.
13287 (vect_permute_load_chain): Likewise.
13288 (vect_shift_permute_load_chain): Likewise.
13289 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13290 (vect_transform_slp_perm_load): Likewise.
13291 (vect_schedule_slp_instance): Likewise.
13292 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13293 (vectorizable_mask_load_store): Likewise.
13294 (vectorizable_bswap): Likewise.
13295 (vectorizable_store): Likewise.
13296 (vectorizable_load): Likewise.
13297 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13298 vec_perm_indices instead of auto_vec_perm_indices. Use
13299 tree_to_vec_perm_builder to read the vector from a tree.
13300 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13301 vec_perm_builder instead of a vec_perm_indices.
13302 (have_whole_vector_shift): Use vec_perm_builder and
13303 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13304 truncation to calc_vec_perm_mask_for_shift.
13305 (vect_create_epilog_for_reduction): Likewise.
13306 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13307 from auto_vec_perm_indices to vec_perm_indices.
13308 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13309 instead of changing individual elements.
13310 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13311 the vector in d.perm.
13312 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13313 from auto_vec_perm_indices to vec_perm_indices.
13314 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13315 instead of changing individual elements.
13316 (arm_vectorize_vec_perm_const): Use new_vector to install
13317 the vector in d.perm.
13318 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13319 Update vec_perm_builder constructor.
13320 (rs6000_expand_interleave): Likewise.
13321 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13322 (rs6000_expand_interleave): Likewise.
13324 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13326 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13327 to qimode could truncate the indices.
13328 * optabs.c (expand_vec_perm_var): Likewise.
13330 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13332 * Makefile.in (OBJS): Add vec-perm-indices.o.
13333 * vec-perm-indices.h: New file.
13334 * vec-perm-indices.c: Likewise.
13335 * target.h (vec_perm_indices): Replace with a forward class
13337 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13338 * optabs.h: Include vec-perm-indices.h.
13339 (expand_vec_perm): Delete.
13340 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13341 (expand_vec_perm_const): Declare.
13342 * target.def (vec_perm_const_ok): Replace with...
13343 (vec_perm_const): ...this new hook.
13344 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13345 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13346 * doc/tm.texi: Regenerate.
13347 * optabs.def (vec_perm_const): Delete.
13348 * doc/md.texi (vec_perm_const): Likewise.
13349 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13350 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13351 expand_vec_perm for constant permutation vectors. Assert that
13352 the mode of variable permutation vectors is the integer equivalent
13353 of the mode that is being permuted.
13354 * optabs-query.h (selector_fits_mode_p): Declare.
13355 * optabs-query.c: Include vec-perm-indices.h.
13356 (selector_fits_mode_p): New function.
13357 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13358 is defined, instead of checking whether the vec_perm_const_optab
13359 exists. Use targetm.vectorize.vec_perm_const instead of
13360 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13361 fit in the vector mode before using a variable permute.
13362 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13363 vec_perm_indices instead of an rtx.
13364 (expand_vec_perm): Replace with...
13365 (expand_vec_perm_const): ...this new function. Take the selector
13366 as a vec_perm_indices rather than an rtx. Also take the mode of
13367 the selector. Update call to shift_amt_for_vec_perm_mask.
13368 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13369 Use vec_perm_indices::new_expanded_vector to expand the original
13370 selector into bytes. Check whether the indices fit in the vector
13371 mode before using a variable permute.
13372 (expand_vec_perm_var): Make global.
13373 (expand_mult_highpart): Use expand_vec_perm_const.
13374 * fold-const.c: Includes vec-perm-indices.h.
13375 * tree-ssa-forwprop.c: Likewise.
13376 * tree-vect-data-refs.c: Likewise.
13377 * tree-vect-generic.c: Likewise.
13378 * tree-vect-loop.c: Likewise.
13379 * tree-vect-slp.c: Likewise.
13380 * tree-vect-stmts.c: Likewise.
13381 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13383 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13384 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13385 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13386 (aarch64_vectorize_vec_perm_const): ...this new function.
13387 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13388 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13389 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13390 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13391 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13392 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13393 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13395 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13396 check for NEON modes.
13397 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13398 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13399 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13400 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13402 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13403 the old VEC_PERM_CONST conditions.
13404 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13405 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13406 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13407 (ia64_vectorize_vec_perm_const_ok): Merge into...
13408 (ia64_vectorize_vec_perm_const): ...this new function.
13409 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13410 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13411 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13412 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13413 * config/mips/mips.c (mips_expand_vec_perm_const)
13414 (mips_vectorize_vec_perm_const_ok): Merge into...
13415 (mips_vectorize_vec_perm_const): ...this new function.
13416 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13417 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13418 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13419 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13420 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13421 (rs6000_expand_vec_perm_const): Delete.
13422 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13424 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13425 (altivec_expand_vec_perm_const_le): Take each operand individually.
13426 Operate on constant selectors rather than rtxes.
13427 (altivec_expand_vec_perm_const): Likewise. Update call to
13428 altivec_expand_vec_perm_const_le.
13429 (rs6000_expand_vec_perm_const): Delete.
13430 (rs6000_vectorize_vec_perm_const_ok): Delete.
13431 (rs6000_vectorize_vec_perm_const): New function.
13432 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13433 an element count and rtx array.
13434 (rs6000_expand_extract_even): Update call accordingly.
13435 (rs6000_expand_interleave): Likewise.
13436 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13437 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13438 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13439 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13440 (rs6000_expand_vec_perm_const): Delete.
13441 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13442 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13443 (altivec_expand_vec_perm_const_le): Take each operand individually.
13444 Operate on constant selectors rather than rtxes.
13445 (altivec_expand_vec_perm_const): Likewise. Update call to
13446 altivec_expand_vec_perm_const_le.
13447 (rs6000_expand_vec_perm_const): Delete.
13448 (rs6000_vectorize_vec_perm_const_ok): Delete.
13449 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13450 reference to the SPE evmerge intructions.
13451 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13452 an element count and rtx array.
13453 (rs6000_expand_extract_even): Update call accordingly.
13454 (rs6000_expand_interleave): Likewise.
13455 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13456 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13458 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13460 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13462 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13463 vector mode and that that mode matches the mode of the data
13465 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13466 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13467 directly using expand_vec_perm_1 when forcing selectors into
13469 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13471 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13473 * optabs-query.h (can_vec_perm_p): Delete.
13474 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13475 * optabs-query.c (can_vec_perm_p): Split into...
13476 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13477 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13478 particular selector is valid.
13479 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13480 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13481 (vect_grouped_load_supported): Likewise.
13482 (vect_shift_permute_load_chain): Likewise.
13483 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13484 (vect_transform_slp_perm_load): Likewise.
13485 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13486 (vectorizable_bswap): Likewise.
13487 (vect_gen_perm_mask_checked): Likewise.
13488 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13489 implementations of variable permutation vectors into account
13490 when deciding which selector to use.
13491 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13492 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13493 with a false third argument.
13494 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13495 to test whether the constant selector is valid and can_vec_perm_var_p
13496 to test whether a variable selector is valid.
13498 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13500 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13501 * optabs-query.c (can_vec_perm_p): Likewise.
13502 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13503 instead of vec_perm_indices.
13504 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13505 (vect_gen_perm_mask_checked): Likewise,
13506 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13507 (vect_gen_perm_mask_checked): Likewise,
13509 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13511 * optabs-query.h (qimode_for_vec_perm): Declare.
13512 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13513 (qimode_for_vec_perm): ...this new function.
13514 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13516 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13518 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13519 does not have a conditional at the top.
13521 2018-01-02 Richard Biener <rguenther@suse.de>
13523 * ipa-inline.c (big_speedup_p): Fix expression.
13525 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13528 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13531 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13535 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13536 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13537 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13538 cond_taken_branch_cost 3->4.
13540 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13542 PR tree-optimization/83581
13543 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13544 TODO_cleanup_cfg if any changes have been made.
13546 PR middle-end/83608
13547 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13548 convert_modes if target mode has the right side, but different mode
13551 PR middle-end/83609
13552 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13553 last argument when extracting from CONCAT. If either from_real or
13554 from_imag is NULL, use expansion through memory. If result is not
13555 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13556 the parts directly to inner mode, if even that fails, use expansion
13559 PR middle-end/83623
13560 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13561 check for bswap in mode rather than HImode and use that in expand_unop
13564 Copyright (C) 2018 Free Software Foundation, Inc.
13566 Copying and distribution of this file, with or without modification,
13567 are permitted in any medium without royalty provided the copyright
13568 notice and this notice are preserved.