1 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2 Jakub Jelinek <jakub@redhat.com>
4 PR tree-optimization/85692
5 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6 source permute as well.
8 2018-05-11 Martin Liska <mliska@suse.cz>
11 * doc/extend.texi: Document LLVM style format for no_sanitize
14 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
17 mode_supports_vsx_dform_quad to mode_supports_dq_form.
18 (mode_supports_vsx_dform_quad): Likewise.
19 (mode_supports_vmx_dform): Move these functions to be next to the
20 other mode_supports functions.
21 (mode_supports_dq_form): Likewise.
22 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
23 mode_supports_dq_form.
24 (reg_offset_addressing_ok_p): Likewise.
25 (offsettable_ok_by_alignment): Likewise.
26 (rs6000_legitimate_offset_address_p): Likewise.
27 (legitimate_lo_sum_address_p): Likewise.
28 (rs6000_legitimize_address): Likewise.
29 (rs6000_legitimize_reload_address): Likewise.
30 (rs6000_secondary_reload_inner): Likewise.
31 (rs6000_preferred_reload_class): Likewise.
32 (rs6000_output_move_128bit): Likewise.
34 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
36 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
37 Generate SImode target register for null target.
38 <case IX86_BUILTIN_XGETBV>: Ditto.
39 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
40 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
42 2018-05-10 Carl Love <cel@us.ibm.com>
44 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
45 dcbtt and dcbtstt if operands[2] is 0.
47 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
50 * config/i386/sse.md (usadv64qi): New expander.
52 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
54 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
55 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
57 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
58 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
59 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
60 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
61 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
62 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
63 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
64 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
65 altivec_vsumsws): Adjust.
66 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
67 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
68 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
70 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
71 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
72 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
73 (altivec_lve<VI_char>x): Delete expand.
74 (*altivec_lve<VI_char>x_internal): Rename to...
75 (altivec_lve<VI_char>x): ... this.
76 (altivec_lvxl_<mode>): Delete expand.
77 (*altivec_lvxl_<mode>_internal): Rename to ...
78 (altivec_lvxl_<mode>): ... this.
79 (altivec_stvxl_<mode>): Delete expand.
80 (*altivec_stvxl_<mode>_internal): Rename to ...
81 (altivec_stvxl_<mode>): ... this.
82 (altivec_stve<VI_char>x): Delete expand.
83 (*altivec_stve<VI_char>x_internal): Rename to ...
84 (altivec_stve<VI_char>x): ... this.
85 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
86 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
87 reduc_plus_scal_<mode>): Adjust.
88 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
90 (rs6000_cpu_cpp_builtins): Adjust.
91 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
92 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
93 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
94 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
96 (rs6000_split_vec_extract_var): Adjust.
97 (rs6000_split_v4si_init): Adjust.
98 (swap_selector_for_mode): Delete.
99 (altivec_expand_lvx_be, altivec_expand_stvx_be,
100 altivec_expand_stvex_be): Delete.
101 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
102 -maltivec=be support.
103 (rs6000_gimple_fold_builtin): Ditto.
104 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
106 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
107 (TARGET_DIRECT_MOVE_64BIT): Adjust.
108 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
109 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
110 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
111 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
112 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
113 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
114 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
115 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
116 anonymous split): Adjust.
117 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
118 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
120 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
122 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
123 when --with-gxx-include-dir is also specified.
124 * configure: Regenerate.
126 2018-05-09 Jim Wilson <jimw@sifive.com>
129 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
130 * config/riscv/t-withmultilib: New.
131 * config/riscv/withmultilib.h: New.
132 * doc/install.texi: Document RISC-V --with-multilib-list support.
134 2018-05-09 Richard Biener <rguenther@suse.de>
136 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
138 (vect_bb_vectorization_profitable_p): Adjust. Compute
139 actual scalar cost using the cost vector and the add_stmt_cost
142 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
144 PR rtl-optimization/85645
145 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
146 in the REG_CFA_REGISTER note for LR, don't leave it empty.
148 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
150 PR rtl-optimization/85645
151 * shrink-wrap.c (spread_components): Return a boolean saying if
152 anything was changed.
153 (try_shrink_wrapping_separate): Iterate spread_components until
154 nothing changes anymore.
156 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
158 PR rtl-optimization/85645
159 * regrename.c (build_def_use): Also kill the chains that include the
160 destination of a REG_CFA_REGISTER note.
162 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
164 PR rtl-optimization/85645
165 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
166 insn that has a REG_CFA_REGISTER note.
168 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
170 * cfgexpand.c (expand_clobber): New function.
171 (expand_gimple_stmt_1): Use it.
172 * tree-vect-stmts.c (vect_clobber_variable): New function,
174 (vectorizable_simd_clone_call): ...here.
175 (vectorizable_store): Emit a clobber either side of an
176 IFN_STORE_LANES sequence.
177 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
179 2018-05-09 Tom de Vries <tom@codesourcery.com>
182 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
183 (define_insn "trap_if_false"): Add exit after trap.
185 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
187 PR rtl-optimization/85638
188 * bb-reorder.c: Include common/common-target.h.
189 (create_forwarder_block): New function extracted from...
190 (fix_up_crossing_landing_pad): ...here. Rename into...
191 (dw2_fix_up_crossing_landing_pad): ...this.
192 (sjlj_fix_up_crossing_landing_pad): New function.
193 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
194 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
195 from both partitions and exit the loop after one iteration.
197 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
200 * doc/extend.texi (PowerPC Built-in Functions): Rename this
202 (Basic PowerPC Built-in Functions): The new name of the
203 subsection previously known as "PowerPC Built-in Functions".
204 (Basic PowerPC Built-in Functions Available on all Configurations):
206 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
208 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
210 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
212 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
215 2018-05-08 Jim Wilson <jimw@sifive.com>
217 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
218 (LD_EMUL_SUFFIX): New.
221 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
223 * doc/extend.texi (PowerPC Built-in Functions): Rename this
225 (Basic PowerPC Built-in Functions): The new name of the
226 subsection previously known as "PowerPC Built-in Functions".
227 (Basic PowerPC Built-in Functions Available on all Configurations):
229 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
231 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
233 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
235 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
238 2018-05-08 Jakub Jelinek <jakub@redhat.com>
241 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
242 after cmpelim optimization.
244 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
246 * config.gcc: Support "goldmont".
247 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
248 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
250 * config/i386/i386.c (m_GOLDMONT): Define.
251 (processor_target_table): Add "goldmont".
252 (PTA_GOLDMONT): Define.
253 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
254 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
255 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
256 (fold_builtin_cpu): Add "goldmont".
257 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
258 (ix86_option_override_internal): Add "goldmont".
259 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
260 (processor_type): Add PROCESSOR_GOLDMONT.
261 * config/i386/i386.md: Add CPU "glm".
262 * config/i386/glm.md: New file.
263 * config/i386/x86-tune.def: Add m_GOLDMONT.
264 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
266 2018-05-08 Jakub Jelinek <jakub@redhat.com>
269 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
271 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
272 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
273 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
276 * config/i386/i386.c (ix86_fold_builtin): Handle
277 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
280 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
281 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
283 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
286 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
287 (check_arch): Likewise.
288 (check_fpu): Return the result rather than printing it.
289 (end arch): Fix operator precedence.
291 (END): Print the result from check_fpu.
293 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
294 Alan Hayward <alan.hayward@arm.com>
295 David Sherwood <david.sherwood@arm.com>
297 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
298 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
299 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
300 (*fcmuo<mode>_and): New patterns.
302 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
304 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
305 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
306 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
307 (cmp_op, sve_imm_con): New code attributes.
308 (SVE_COND_INT_CMP, imm_con): Delete.
309 (cmp_op): Remove above unspecs from int attribute.
310 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
312 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
313 comparison-specific unspecs.
314 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
315 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
316 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
317 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
318 (*vec_fcm<cmp_op><mode>): Rename to...
319 (*fcm<cmp_op><mode>): ...this and adjust likewise.
320 (*vec_fcmuo<mode>): Rename to...
321 (*fcmuo<mode>): ...this and adjust likewise.
322 (*pred_fcm<cmp_op><mode>): New pattern.
323 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
324 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
326 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
328 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
329 (aarch64_emit_sve_predicated_cond): New function.
330 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
331 (aarch64_emit_unspec_cond_or): Replace with...
332 (aarch64_emit_sve_or_conds): ...this new function. Use
333 aarch64_emit_sve_ptrue_op for the individual comparisons and
334 aarch64_emit_binop to OR them together.
335 (aarch64_emit_inverted_unspec_cond): Replace with...
336 (aarch64_emit_sve_inverted_cond): ...this new function. Use
337 aarch64_emit_sve_ptrue_op for the comparison and
338 aarch64_emit_unop to invert the result.
339 (aarch64_expand_sve_vec_cmp_float): Update after the above
340 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
342 2018-05-07 Nathan Sidwell <nathan@acm.org>
344 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
345 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
346 (Backwards Compatibility): Likewise.
348 2018-05-07 Luis Machado <luis.machado@linaro.org>
352 2018-05-07 Luis Machado <luis.machado@linaro.org>
354 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
355 <prefetch_dynamic_strides>: New const bool field.
356 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
357 prefetch_dynamic_strides.
358 (exynosm1_prefetch_tune): Likewise.
359 (thunderxt88_prefetch_tune): Likewise.
360 (thunderx_prefetch_tune): Likewise.
361 (thunderx2t99_prefetch_tune): Likewise.
362 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
363 (aarch64_override_options_internal): Update to set
364 PARAM_PREFETCH_DYNAMIC_STRIDES.
365 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
366 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
367 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
368 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
369 prefetch-dynamic-strides setting.
371 2018-05-07 Luis Machado <luis.machado@linaro.org>
373 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
374 <minimum_stride>: New const int field.
375 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
376 minimum_stride field.
377 (exynosm1_prefetch_tune): Likewise.
378 (thunderxt88_prefetch_tune): Likewise.
379 (thunderx_prefetch_tune): Likewise.
380 (thunderx2t99_prefetch_tune): Likewise.
381 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
382 (aarch64_override_options_internal): Update to set
383 PARAM_PREFETCH_MINIMUM_STRIDE.
384 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
385 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
386 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
387 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
388 stride is constant and is below the minimum stride threshold.
390 2018-05-07 Luis Machado <luis.machado@linaro.org>
392 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
395 2018-05-07 Luis Machado <luis.machado@linaro.org>
397 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
398 <prefetch_dynamic_strides>: New const bool field.
399 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
400 prefetch_dynamic_strides.
401 (exynosm1_prefetch_tune): Likewise.
402 (thunderxt88_prefetch_tune): Likewise.
403 (thunderx_prefetch_tune): Likewise.
404 (thunderx2t99_prefetch_tune): Likewise.
405 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
406 (aarch64_override_options_internal): Update to set
407 PARAM_PREFETCH_DYNAMIC_STRIDES.
408 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
409 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
410 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
411 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
412 prefetch-dynamic-strides setting.
414 2018-05-07 Luis Machado <luis.machado@linaro.org>
416 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
417 <minimum_stride>: New const int field.
418 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
419 minimum_stride field.
420 (exynosm1_prefetch_tune): Likewise.
421 (thunderxt88_prefetch_tune): Likewise.
422 (thunderx_prefetch_tune): Likewise.
423 (thunderx2t99_prefetch_tune): Likewise.
424 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
425 (aarch64_override_options_internal): Update to set
426 PARAM_PREFETCH_MINIMUM_STRIDE.
427 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
428 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
429 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
430 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
431 stride is constant and is below the minimum stride threshold.
433 2018-05-06 Jakub Jelinek <jakub@redhat.com>
436 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
437 the type is addressable. Don't force op into register if it has
440 2018-05-05 Roland McGrath <mcgrathr@google.com>
443 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
444 any section for which we don't know a specific type it should have,
445 regardless of name. Previously this was done only for the exact
446 names ".init_array", ".fini_array", and ".preinit_array".
447 (default_elf_asm_named_section): Add comment about
448 relationship with default_section_type_flags and SECTION_NOTYPE.
449 (get_section): Don't consider it a type conflict if one side has
450 SECTION_NOTYPE and the other doesn't, as long as neither has the
451 SECTION_BSS et al used in the default_section_type_flags logic.
453 2018-05-05 Tom de Vries <tom@codesourcery.com>
456 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
457 (workaround_barsyncs): New function.
458 (nvptx_reorg): Use workaround_barsyncs.
459 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
460 (define_expand "nvptx_membar_cta"): New define_expand.
461 (define_insn "*nvptx_membar_cta"): New insn.
463 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
465 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
466 To improve optimization opportunities.
467 * builtin-types.def: The new needed builtin types for the above.
469 2018-05-04 Richard Biener <rguenther@suse.de>
471 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
472 * gimple-ssa-store-merging.c
473 (imm_store_chain_info::output_merged_store): Remove redundant create,
474 release split_store vector contents on failure.
475 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
476 scalar stmt vector on cache hit.
478 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
480 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
482 * config.gcc (powerpc-xilinx-eabi*): Remove.
483 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
485 (fusion_addis_mem_combo_load): Ditto.
486 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
488 (rs6000_cpu_cpp_builtins): Ditto.
489 * config/rs6000/rs6000-linux.c
490 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
491 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
492 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
494 (rs6000_setup_reg_addr_masks): Ditto.
495 (rs6000_init_hard_regno_mode_ok): Ditto.
496 (rs6000_option_override_internal): Ditto.
497 (legitimate_lo_sum_address_p): Ditto.
498 (rs6000_legitimize_address): Ditto.
499 (rs6000_legitimize_reload_address): Ditto.
500 (rs6000_legitimate_address_p): Ditto.
501 (abi_v4_pass_in_fpr): Ditto.
502 (setup_incoming_varargs): Ditto.
503 (rs6000_gimplify_va_arg): Ditto.
504 (rs6000_split_multireg_move): Ditto.
505 (rs6000_savres_strategy): Ditto.
506 (rs6000_emit_prologue_components): Ditto.
507 (rs6000_emit_epilogue_components): Ditto.
508 (rs6000_emit_prologue): Ditto.
509 (rs6000_emit_epilogue): Ditto.
510 (rs6000_elf_file_end): Ditto.
511 (rs6000_function_value): Ditto.
512 (rs6000_libcall_value): Ditto.
513 * config/rs6000/rs6000.h: Ditto.
514 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
515 (TARGET_MINMAX): ... this. New.
516 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
517 * config/rs6000/rs6000.md: Remove Xilinx FP support.
518 (*movsi_internal1_single): Delete.
519 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
520 mfpu=, mxilinx-fpu): Delete.
521 * config/rs6000/singlefp.h: Delete.
522 * config/rs6000/sysv4.h: Remove Xilinx FP support.
523 * config/rs6000/t-rs6000: Ditto.
524 * config/rs6000/t-xilinx: Delete.
525 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
526 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
527 (VStype_simple): Delete.
528 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
529 * config/rs6000/xfpu.h: Delete.
530 * config/rs6000/xfpu.md: Delete.
531 * config/rs6000/xilinx.h: Delete.
532 * config/rs6000/xilinx.opt: Delete.
533 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
534 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
536 2018-05-04 Tom de Vries <tom@codesourcery.com>
539 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
542 2018-05-04 Richard Biener <rguenther@suse.de>
545 * tree-complex.c (update_complex_assignment): We are always in SSA form.
546 (expand_complex_div_wide): Likewise.
547 (expand_complex_operations_1): Likewise.
548 (expand_complex_libcall): Preserve EH info of the original stmt.
549 (tree_lower_complex): Handle removed blocks.
550 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
551 on complex multiplication and division libcall builtins.
553 2018-05-04 Richard Biener <rguenther@suse.de>
556 * fold-const.c (negate_expr_p): Restrict negation of operand
557 zero of a division to when we know that can happen without
559 (fold_negate_expr_1): Likewise.
561 2018-05-04 Jakub Jelinek <jakub@redhat.com>
564 * real.h (real_nextafter): Declare.
565 * real.c (real_nextafter): New function.
566 * fold-const-call.c (fold_const_nextafter): New function.
567 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
569 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
570 even when arg1_mode is different from arg0_mode.
572 2018-05-03 Nathan Sidwell <nathan@acm.org>
574 * doc/extend.texi (Deprecated Features): Remove
576 (Backwards Compatibility): Likewise.
577 * doc/invoke.texi (C++ Language Options): Likewise.
578 (C++ Dialect Options): Likewise.
580 2018-05-03 Jakub Jelinek <jakub@redhat.com>
583 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
584 _mm512_mask_mullox_epi64): New intrinsics.
586 2018-05-03 Tom de Vries <tom@codesourcery.com>
589 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
590 dump files): Add offload-tree.
592 2018-05-03 Richard Biener <rguenther@suse.de>
594 PR tree-optimization/85615
595 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
596 to loops not nested in BBs loop father to avoid creating multi-entry
599 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
601 PR tree-optimization/70291
602 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
603 arguments. Change return type to tree. Emit libcall as a new
604 statement rather than replacing existing one when inplace_p is true.
605 (expand_complex_multiplication_components): New function.
606 (expand_complex_multiplication): Expand floating-point complex
607 multiplication using the above.
608 (expand_complex_division): Rename inner_type parameter to type.
609 Update expand_complex_libcall call-site.
610 (expand_complex_operations_1): Update expand_complex_multiplication
611 and expand_complex_division call-sites.
613 2018-05-02 Jakub Jelinek <jakub@redhat.com>
616 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
617 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
618 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
619 the highest significant bit of the shift count mask is clear. In
620 check whether and[sq]i3 is needed verify that all significant bits
621 of the shift count other than the highest are set.
623 2018-05-02 Tom de Vries <tom@codesourcery.com>
626 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
627 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
628 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
629 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
630 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
631 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
632 __builtin_goacc_parlevel_size.
634 2018-05-02 Richard Biener <rguenther@suse.de>
636 PR tree-optimization/85597
637 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
638 do not use split vect_get_vec_defs call but call vect_get_slp_defs
641 2018-05-02 Tom de Vries <tom@codesourcery.com>
644 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
645 dump files): Add ltrans-tree.
647 2018-05-02 Tom de Vries <tom@codesourcery.com>
650 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
651 dump files): Add wpa-ipa.
653 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
655 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
656 powerpc*-*-linux*paired* target.
657 * config/rs6000/750cl.h: Delete.
658 * config/rs6000/paired.h: Delete.
659 * config/rs6000/paired.md: Delete.
660 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
662 * config/rs6000/rs6000-builtin.def: Remove paired float support.
663 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
664 comment. Remove paired float support.
665 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
666 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
668 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
669 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
671 * config/rs6000/rs6000.c: Remove paired float support.
672 (paired_expand_vector_init, paired_expand_vector_move,
673 paired_emit_vector_compare, paired_emit_vector_cond_expr,
674 (paired_expand_lv_builtin, paired_expand_stv_builtin,
675 paired_expand_builtin, paired_expand_predicate_builtin,
676 paired_init_builtins): Delete.
677 * config/rs6000/rs6000.h: Remove paired float support.
678 * config/rs6000/rs6000.md: Remove paired float support.
679 (move_from_CR_ov_bit): Delete.
680 * config/rs6000/rs6000.opt (mpaired): Delete.
681 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
682 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
684 2018-05-02 Richard Biener <rguenther@suse.de>
687 * gimplify.c (gimplify_save_expr): When in SSA form allow
688 SAVE_EXPRs to compute to SSA vars.
690 2018-05-02 Jakub Jelinek <jakub@redhat.com>
693 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
694 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
695 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
696 clobber operands[2], instead use a new pseudo. Formatting fixes.
698 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
700 PR tree-optimization/85586
701 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
702 exit early for statements in the same group if the accesses are
705 2018-05-02 Tom de Vries <tom@codesourcery.com>
708 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
711 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
713 PR tree-optimization/85143
714 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
716 2018-05-01 Tom de Vries <tom@codesourcery.com>
719 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
720 not found" error message.
722 2018-05-01 Tom de Vries <tom@codesourcery.com>
725 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
726 * vec.c (test_ordered_remove_if): New function.
727 (vec_c_tests): Call test_ordered_remove_if.
728 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
729 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
730 * tree-vect-patterns.c (vect_pattern_recog_1): Use
731 VEC_ORDERED_REMOVE_IF.
733 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
735 PR tree-optimization/82665
736 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
737 pointer subtraction where arguments come from a memchr call.
739 2018-05-01 Jakub Jelinek <jakub@redhat.com>
741 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
742 --push-state --as-needed and --pop-state instead of --as-needed and
743 --no-as-needed if ld supports it.
744 * configure: Regenerated.
747 * doc/install.texi2html: Replace _002d with - and _002a with * in
748 generated html files using sed.
750 2018-04-30 David Malcolm <dmalcolm@redhat.com>
753 * gcc-rich-location.c (blank_line_before_p): New function.
754 (use_new_line): New function.
755 (gcc_rich_location::add_fixit_insert_formatted): New function.
756 * gcc-rich-location.h
757 (gcc_rich_location::add_fixit_insert_formatted): New function.
759 2018-04-30 David Malcolm <dmalcolm@redhat.com>
761 * selftest.c (assert_streq): Rename "expected" and "actual" to
762 "val1" and "val2". Extend NULL-handling to cover both inputs
763 symmetrically, while still requiring both to be non-NULL for a pass.
764 * selftest.h (assert_streq): Rename "expected" and "actual" to
766 (ASSERT_EQ): Likewise.
767 (ASSERT_EQ_AT): Likewise.
768 (ASSERT_KNOWN_EQ): Likewise.
769 (ASSERT_KNOWN_EQ_AT): Likewise.
770 (ASSERT_NE): Likewise.
771 (ASSERT_MAYBE_NE): Likewise.
772 (ASSERT_MAYBE_NE_AT): Likewise.
773 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
774 the assertion to pass.
775 (ASSERT_STREQ_AT): Likewise.
777 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
779 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
780 interaction with -pie.
782 2018-04-30 David Malcolm <dmalcolm@redhat.com>
784 * selftest.h: Fix alphabetization of per-source-file selftest
787 2018-04-30 Jason Merrill <jason@redhat.com>
789 PR c++/61982 - dead stores to destroyed objects.
790 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
793 2018-04-30 Jason Merrill <jason@redhat.com>
795 * tree.c (build_clobber): New.
796 * tree.h: Declare it.
797 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
799 2018-04-30 David Malcolm <dmalcolm@redhat.com>
801 * diagnostic-show-locus.c (layout::layout): Update for
802 location_get_source_line returning a char_span.
803 (struct char_span): Move to input.h.
804 (struct correction): Update for fields in char_span becoming
806 (struct source_line): Update for location_get_source_line
807 returning a char_span.
808 (layout::print_line): Likewise.
809 * edit-context.c (edited_file::print_content): Likewise.
810 (edited_file::print_diff_hunk): Likewise.
811 (edited_file::print_run_of_changed_lines): Likewise.
812 (edited_file::get_num_lines): Likewise.
813 (edited_line::edited_line): Likewise.
814 * final.c (asm_show_source): Likewise.
815 * input.c (location_get_source_line): Convert return type
816 from const char * to char_span, losing the final "line_len"
818 (dump_location_info): Update for the above.
819 (get_substring_ranges_for_loc): Likewise. Use a char_span
820 when handling the literal within the line.
821 (test_reading_source_line): Update for location_get_source_line
822 returning a char_span.
823 * input.h (class char_span): Move here from
824 diagnostic-show-locus.c, converting from a struct to a class.
825 Make data members private.
826 (char_span::operator bool): New.
827 (char_span::length): New.
828 (char_span::get_buffer): New.
829 (char_span::operator[]): New.
830 (char_span::subspan): Make const.
831 (char_span::xstrdup): New.
832 (location_get_source_line): Convert return type from const char *
833 to char_span, losing the final "line_size" param.
835 2018-04-30 Jan Hubicka <jh@suse.cz>
837 * lto-wrapper.c (ltrans_priorities): New static var.
839 (run_gcc): Read priorities and if doing parallel build order
840 the Makefile by them.
842 2018-04-30 David Malcolm <dmalcolm@redhat.com>
844 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
846 2018-04-30 Richard Biener <rguenther@suse.de>
848 * tree-cfg.c (verify_address): Remove base argument, add
849 flag whether to check TREE_ADDRESSABLE and do that.
850 (verify_expr): Remove.
851 (verify_types_in_gimple_reference): Add pieces from verify_expr.
852 (verify_gimple_assign_single): Likewise.
853 (verify_gimple_switch): Likewise.
854 (verify_expr_location_1): Dereference tp once. Add (disabled)
855 piece from verify_expr.
856 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
858 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
860 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
862 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
864 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
865 (small_data_pattern): Likewise.
866 (arc_rewrite_small_data): Likewise.
867 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
868 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
869 (get_symbol_alignment): New function.
870 (legitimate_small_data_address_p): Likewise.
871 (legitimate_scaled_address): Update, call
872 legitimate_small_data_address_p.
873 (output_sdata): New static variable.
874 (arc_print_operand): Update how we handle small data operands.
875 (arc_print_operand_address): Likewise.
876 (arc_legitimate_address_p): Update, use
877 legitimate_small_data_address_p.
878 (arc_rewrite_small_data_p): Remove.
879 (arc_rewrite_small_data_1): Likewise.
880 (arc_rewrite_small_data): Likewise.
881 (small_data_pattern): Likewise.
882 (compact_sda_memory_operand): Update to use
883 legitimate_small_data_address_p and get_symbol_alignment.
884 (prepare_move_operands): Don't rewite sdata pattern.
885 (prepare_extend_operands): Remove.
886 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
888 (zero_extendqisi2): Likewise.
889 (zero_extendhisi2): Likewise.
890 (extendqihi2): Likewise.
891 (extendqisi2): Likewise.
892 (extendhisi2): Likewise.
896 * config/arc/constraints.md (Usd): Change it to memory constraint.
898 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
900 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
901 as source of std instructions.
902 * config/arc/arc.md (movsi_insn): Update pattern predicate to
903 allow 6-bit constants as source for store instructions.
904 (movdi_insn): Update instruction pattern to allow 6-bit constants
905 as source for store instructions.
907 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
909 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
911 2018-04-30 Nathan Sidwell <nathan@acm.org>
912 Sandra Loosemore <sandra@codesourcery.com>
914 * dumpfile.c (dump_open): Allow '-' for stdout.
915 * doc/invoke.texi (Developer Options): Document dump filename
916 determination early. Document stdin/stdout selection.
918 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
920 Microblaze Target: PIC data text relative
922 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
923 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
925 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
926 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
927 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
928 New addressing mode for data-text relative position indepenedent code.
929 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
930 'ADDRESS_SYMBOLIC_TXT_REL'.
931 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
932 (microblaze_legitimate_pic_operand): Exclude function calls from
933 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
934 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
936 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
937 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
938 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
939 for 'address + offset'.
940 (microblaze_expand_prologue): Add new function prologue call for
942 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
943 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
944 table in case of TARGET_PIC_DATA_TEXT_REL.
945 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
946 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
947 Add new macros 'UNSPEC_TEXT',
948 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
949 + exclude function calls from 'UNSPEC_PLT' in case of data text
951 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
952 new target hook for generating address diff vector tables in case of
954 * doc/tm.texi : Regenerate.
955 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
956 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
957 of addr diff vector generation.
958 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
959 target hook definition.
960 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
961 Add default function for generate_pic_addr_diff_vec -> flag_pic.
962 * doc/invoke.texi (Add new pic option): Add new microblaze pic
963 option for data text relative.
965 2018-04-30 Richard Biener <rguenther@suse.de>
967 * tree-chrec.h (evolution_function_is_constant_p): Remove
969 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
971 2018-04-30 Richard Biener <rguenther@suse.de>
974 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
976 2018-04-30 Richard Biener <rguenther@suse.de>
978 PR tree-optimization/28364
979 PR tree-optimization/85275
980 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
981 copying first exit test.
983 2018-04-28 Mark Wielaard <mark@klomp.org>
985 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
987 (dwarf_AT): Handle DW_AT_addr_base.
988 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
990 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
993 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
994 (*ashl<dwi>3_doubleword_mask_1): Ditto.
995 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
996 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
998 2018-04-28 Richard Biener <rguenther@suse.de>
1000 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1001 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1002 to reflect use. Only add interesting stmts.
1004 2018-04-27 Martin Jambor <mjambor@suse.cz>
1007 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1008 the jump function allows for passing through aggregate values.
1010 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1012 * input.h (in_system_header_at): Convert from macro to inline
1014 (from_macro_expansion_at): Likewise.
1015 (from_macro_definition_at): Likewise.
1017 2018-04-27 Jeff Law <law@redhat.com>
1019 * config.gcc: Mark tile* targets as deprecated/obsolete.
1021 2018-04-27 Richard Biener <rguenther@suse.de>
1023 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1026 2018-04-27 Richard Biener <rguenther@suse.de>
1028 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1030 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1032 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1033 with Yd constraint. Set "preferred_for_speed" attribute from
1034 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1036 (*movdi_internal): Ditto.
1037 (movti_interunit splitters): Remove
1038 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1039 (movdi_interunit splitters): Ditto.
1040 * config/i386/constraints.md (Ye): Remove.
1041 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1043 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1046 * config/aarch64/constraints.md (Usg): Limit to 31.
1049 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1051 PR tree-optimization/85529
1052 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1053 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1054 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1055 zero extension or masking of the MSB bit.
1056 (optimize_range_tests): Add FIRST_BB argument, pass it through
1057 to optimize_range_tests_var_bound.
1058 (maybe_optimize_range_tests, reassociate_bb): Adjust
1059 optimize_range_tests callers.
1061 2018-04-26 Richard Biener <rguenther@suse.de>
1062 Jakub Jelinek <jakub@redhat.com>
1064 * cgraph.h (symbol_table): Just declare debug method here.
1065 * symtab.c (symbol_table::debug): Define.
1067 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1069 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1071 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1073 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1074 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1075 (*movdi_internal): Substitute Yi and Yj constraint with x
1076 and Ym and Yn constraint with y constraint. Update "isa"
1077 attribute and set "preferred_for_speed" attribute from
1078 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1079 (*movsi_internal): Ditto.
1080 (*movdf_internal): Ditto.
1081 (*movsf_internal): Ditto.
1082 (*zero_extendsidi2): Ditto.
1083 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1084 (sse2_loadld): Ditto.
1085 (*vec_extract<ssevecmodelower>_0): Ditto.
1086 (*vec_extractv4si_0_zext_sse4): Ditto.
1087 (vec_concatv2di): Ditto.
1088 (*vec_dup<mode>): Ditto.
1089 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1090 * config/i386/constraints.md (Yi): Remove.
1095 2018-04-26 Nathan Sidwell <nathan@acm.org>
1097 * dumpfile.c (dump_open): New.
1098 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1099 (dump_finish): Detect stdio/stderr by value not name.
1101 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1103 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1105 2018-04-26 Tom de Vries <tom@codesourcery.com>
1108 * config/nvptx/nvptx.c (verify_neutering_jumps)
1109 (verify_neutering_labels): New function
1110 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1112 2018-04-26 Tom de Vries <tom@codesourcery.com>
1115 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1116 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1119 2018-04-26 Richard Biener <rguenther@suse.de>
1120 Tom de Vries <tom@codesourcery.com>
1123 * lto-streamer-out.c (output_function): Fixup loops if required to match
1124 discovery done in the reader.
1126 2018-04-26 Richard Biener <rguenther@suse.de>
1128 PR tree-optimization/85116
1129 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1130 have a loop exit from the single latch predecessor. Remove
1131 case of header with just condition.
1132 (ch_base::copy_headers): Exclude infinite loops from any
1134 (pass_ch::execute): Record exits.
1136 2018-04-26 Richard Biener <rguenther@suse.de>
1138 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1139 prologue cost vector and pass it to vect_get_load_cost.
1140 (vect_get_peeling_costs_all_drs): Likewise.
1141 (vect_peeling_hash_get_lowest_cost): Likewise.
1142 (vect_enhance_data_refs_alignment): Likewise.
1144 2018-04-26 Richard Biener <rguenther@suse.de>
1147 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1148 checking of integer<->pointer conversions.
1149 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1150 sign-/zero-extending pointer types.
1151 (expand_omp_for_static_chunk): Likewise.
1153 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1154 Jean Lee <xiaoyur347@gmail.com>
1156 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1157 (TARGET_ASAN_SHADOW_OFFSET): Define.
1158 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1159 true for -fsanitize=address.
1161 2018-04-25 Mark Wielaard <mark@klomp.org>
1163 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1166 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1168 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1169 than "alu", remove explicit "memory" and "imm_disp" attributes.
1170 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1173 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1174 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1175 gen_lowpart_no_emit.
1177 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1180 * config/i386/i386.c (ix86_expand_builtin): Change memory
1181 operand to XI, extend p0 to Pmode.
1182 * config/i386/i386.md: Change unspec volatile and operand
1183 1 mode to XI, change operand 0 mode to P.
1185 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1187 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1188 GET_MODE_MASK before any checking.
1189 (nds32_can_use_bset_p): Likewise.
1190 (nds32_can_use_btgl_p): Likewise.
1192 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1194 * config/nds32/nds32-doubleword.md: New define_split pattern for
1195 illegal register number.
1197 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1199 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1201 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1203 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1205 2018-04-25 Richard Biener <rguenther@suse.de>
1207 * lto-streamer.h (LTO_major_version): Bump to 8.
1209 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1211 * BASE-VER: Set to 9.0.0.
1213 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1215 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1216 in __abskf2 and __powikf2.
1218 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1221 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1222 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1223 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1224 Use the above on operand 2. Reindent.
1225 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1227 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1230 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1232 * config/i386/i386.opt (mcet): Removed.
1233 * doc/install.texi: Remove -mcet documentation.
1234 * doc/invoke.texi: Likewise.
1236 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1239 * doc/install.texi: Remove -mcet from bootstrap-cet.
1241 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1244 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1245 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1249 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1250 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1251 containing a CONST_VECTOR.
1253 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1255 * doc/install.texi: Update newlib dependency for nvptx.
1257 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1260 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1261 instead of INTVAL when shifting x left.
1263 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1265 PR tree-optimization/85478
1266 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1267 vect_grouped_store_supported for single element vectors.
1269 2018-04-24 Richard Biener <rguenther@suse.de>
1272 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1273 load cost increase to the case of non-constant step.
1275 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1278 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1279 destination if any_malformed_asm.
1281 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1284 * expr.c (store_field): In the bitfield case, if the value comes from
1285 a function call and is returned in registers by means of a PARALLEL,
1286 do not change the mode of the temporary unless BLKmode and VOIDmode.
1288 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1290 PR rtl-optimization/85423
1291 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1292 dependencies to debug insns when the previous insn is non-debug.
1294 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1296 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1297 enums into a single definition.
1298 (fls): Fix predicates and printing.
1301 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1303 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1304 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1305 and short u6 immediate.
1306 (check_if_valid_sleep_operand): Remove.
1307 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1309 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1311 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1312 flag_always_save_lp condition.
1313 * config/nds32/nds32.opt (malways-save-lp): New option.
1315 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1317 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1318 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1319 * config/nds32/nds32.h
1320 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1321 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1323 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1325 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1326 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1328 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1329 Chung-Ju Wu <jasonwucj@gmail.com>
1331 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1333 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1334 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1336 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1338 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1340 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1342 * config/nds32/nds32-protos.h (nds32_data_alignment,
1343 nds32_local_alignment): Declare.
1344 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1345 nds32_local_alignment): New functions.
1346 (TARGET_CONSTANT_ALIGNMENT): Define.
1347 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1349 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1351 * config/nds32/nds32.c
1352 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1353 (TARGET_MODES_TIEABLE_P): Likewise.
1355 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1357 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1360 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1361 Chung-Ju Wu <jasonwucj@gmail.com>
1363 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1364 for unaligned access.
1365 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1367 * config/nds32/nds32-intrinsic.md: Likewise.
1368 * config/nds32/nds32_intrinsic.h: Likewise.
1369 * config/nds32/nds32.h (nds32_builtins): Likewise.
1370 * config/nds32/nds32.opt (munaligned-access): New option.
1371 * config/nds32/nds32.c (nds32_asm_file_start): Display
1372 flag_unaligned_access status.
1374 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1376 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1377 -mno-relax is present.
1378 * config/riscv/linux.h (LINK_SPEC): Ditto.
1380 2018-04-20 Martin Sebor <msebor@redhat.com>
1383 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1385 (gimple_fold_builtin_stxcpy_chk): Same.
1386 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1388 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1391 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1392 __powikf2 when long double is IEEE 128-bit.
1394 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1396 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1397 step to make sure stack always aligned.
1399 2018-04-20 Carl Love <cel@us.ibm.com>
1402 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1403 size check for arg0.
1405 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1406 Tom de Vries <tom@codesourcery.com>
1409 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1410 Emit insns for calls too.
1411 (nvptx_find_par): Always look for worker-level predecessor insn.
1412 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1414 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1415 (nvptx_process_pars): Propagate frames for calls.
1417 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1420 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1422 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1423 (ix86_handle_option): Don't handle OPT_mibt.
1424 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1426 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1428 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1429 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1430 (ix86_target_macros): Define __CET__ with flag_cf_protection
1431 for -fcf-protection.
1432 * config/i386/i386.c (isa2_opts): Remove -mibt.
1433 * config/i386/i386.h (TARGET_IBT): Removed.
1434 (TARGET_IBT_P): Likewise.
1435 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1436 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1437 * config/i386/i386.opt (mcet): Update help message.
1440 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1441 -mcet as an alias for -mshstk.
1443 2018-04-20 Richard Biener <rguenther@suse.de>
1446 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1447 complexity by forcing a single use of the multiply operand.
1449 2018-04-20 Martin Jambor <mjambor@suse.cz>
1452 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1453 recursion dependency to only apply to non-clones.
1455 2018-04-20 Martin Jambor <mjambor@suse.cz>
1458 * ipa-cp.c (create_specialized_node): Check that clones of
1459 self-recursive edges exist during IPA-CP.
1461 2018-04-19 Toon Moene <toon@moene.org>
1463 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1466 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1468 PR tree-optimization/85467
1469 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1470 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1471 VECTOR_CST element to type.
1473 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1476 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1477 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1478 (builtin_longjmp): Likewise.
1479 (save_stack_nonlocal): New pattern.
1480 (restore_stack_nonlocal): Likewise.
1482 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1485 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1486 Replace ASM_OUTPUT_LABEL with fprintf.
1488 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1491 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1492 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1493 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1494 define __IBT__ and __SHSTK__ for -fcf-protection.
1495 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1497 (ix86_trampoline_init): Likewise.
1498 (x86_output_mi_thunk): Likewise.
1499 (ix86_notrack_prefixed_insn_p): Likewise.
1500 (ix86_option_override_internal): Don't disallow -fcf-protection.
1501 * config/i386/i386.md (rdssp<mode>): Also enable for
1503 (incssp<mode>): Likewise.
1504 (nop_endbr): Likewise.
1505 * config/i386/i386.opt (mcet): Change help message to built-in
1509 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1510 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1511 enable CET built-in functions.
1513 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1515 * common/config/i386/i386-common.c
1516 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1517 OPTION_MASK_ISA_MOVDIRI_UNSET,
1518 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1519 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1520 * config.gcc (movdirintrin.h): New header.
1521 * config/i386/cpuid.h (bit_MOVDIRI,
1522 bit_MOVDIR64B): New bits.
1523 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1525 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1526 (VOID, PVOID, PCVOID)): New function types.
1527 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1528 __builtin_ia32_directstoreu_u64,
1529 __builtin_ia32_movdir64b): New builtins.
1530 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1531 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1533 (ix86_valid_target_attribute_inner_p): Ditto.
1534 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1535 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1536 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1537 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1538 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1539 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1540 (movdiri<mode>, movdir64b_<mode>): New.
1541 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1542 * config/i386/immintrin.h: Include movdirintrin.h.
1543 * config/i386/movdirintrin.h: New file.
1544 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1546 2018-04-19 Richard Biener <rguenther@suse.de>
1549 * cfg.c (clear_bb_flags): When loop state says we have
1550 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1552 2018-04-19 Richard Biener <rguenther@suse.de>
1554 PR tree-optimization/84737
1555 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1556 copying restrict info.
1557 (vect_setup_realignment): Use it.
1558 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1559 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1560 the first DR to all generated stores.
1561 (vectorizable_load): Likewise for loads.
1563 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1565 PR tree-optimization/85446
1566 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1567 the integral and pointer types to have the same precision.
1569 * doc/install.texi: Document --disable-cet being the default and
1572 2018-04-18 Martin Liska <mliska@suse.cz>
1574 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1577 2018-04-18 Martin Liska <mliska@suse.cz>
1580 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1583 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1584 arguments if they are comparable.
1586 2018-04-18 Martin Liska <mliska@suse.cz>
1589 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1592 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1595 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1598 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1599 ENDBR after calling __morestack.
1601 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1604 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1605 by using gcc_base_ver to generate a gcc_driver_version, and use
1606 it when generating GCC_DRIVER_NAME.
1607 * configure: Regenerate.
1609 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1612 * config.gcc: Obsolete powerpc*-*-*spe*.
1614 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1617 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1618 (stabstr_D): Change type of unum from unsigned int to
1619 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1622 2018-04-17 Jim Wilson <jimw@sifive.com>
1625 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1626 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1627 Set arg_pointer_offset after using pretend_args_size.
1629 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1631 PR rtl-optimization/85431
1632 * dse.c (record_store): Ignore zero width stores.
1635 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1636 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1637 __builtin_stack_restore rather than after it.
1638 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1639 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1640 argument instead of virtual_dynamic_stack_rtx.
1642 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1644 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1646 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1647 Add note to error message to explain internal mapping of overloaded
1648 built-in function name to non-overloaded built-in function name.
1649 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1652 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1655 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1656 where the inputs overlap with the output.
1658 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1660 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1661 (=v, v) alternative and explicit "memory" attribute.
1662 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1663 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1665 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1666 "sselog1" type instead of "sselog".
1667 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1668 "sselog". Remove explicit "memory" attribute.
1669 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1670 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1672 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1673 "sselog1" type instead of "sselog". Remove explicit "memory"
1675 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1676 use "sselog1" type instead of "sselog". Remove explicit "memory"
1678 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1679 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1681 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1682 "sselog1" type instead of "sselog". Remove explicit "memory"
1684 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1685 use "sselog1" type instead of "sselog". Remove explicit "memory"
1689 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1692 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1695 2018-04-17 Martin Jambor <mjambor@suse.cz>
1698 * ipa-cp.c (create_specialized_node): Call
1699 expand_all_artificial_thunks if necessary.
1701 2018-04-17 Martin Liska <mliska@suse.cz>
1704 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1705 in message, remote space in between '_G' and '('.
1707 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1710 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1711 avx512f_vmcmp<mode>3<round_saeonly_name>,
1712 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1713 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1714 avx512f_rndscale<mode><round_saeonly_name>,
1715 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1716 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1717 Use %<iptr>2 instead of %2 for -masm=intel.
1718 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1719 avx512f_vcvttss2usi<round_saeonly_name>,
1720 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1722 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1723 avx512f_vcvttsd2usi<round_saeonly_name>,
1724 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1725 Use %q1 instead of %1 for -masm=intel.
1726 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1727 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1728 of %3 for -masm=intel.
1729 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1731 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1733 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1735 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1736 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1738 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1740 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1741 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1742 %g1 and one with %0 and %1.
1743 (avx512er_vmrcp28<mode><round_saeonly_name>,
1744 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1746 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1747 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1748 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1749 of %0 and %{%4%} for -masm=intel.
1750 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1751 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1752 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1753 order of %0 and %{%5%}%{z%} for -masm=intel.
1755 2018-04-17 Jan Hubicka <jh@suse.cz>
1758 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1760 2018-04-17 Martin Liska <mliska@suse.cz>
1763 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1764 for target_clone error message. Make default implementation
1765 clone to be a local declaration.
1766 (separate_attrs): Add new argument and check for an empty
1768 (expand_target_clones): Handle it.
1769 (ipa_target_clone): Make redirection just for target_clones
1772 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1773 Tom de Vries <tom@codesourcery.com>
1776 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1777 tiled basic blocks without omp continue statements.
1779 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1782 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1783 vec_extract expression as having side effects to make sure it gets
1786 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1789 * config/i386/i386.c (get_builtin_code_for_version): Check
1792 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1795 * gcc/config.gcc: Support "skylake".
1796 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1798 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1799 (processor_target_table): Add "skylake".
1800 (ix86_option_override_internal): Add "skylake".
1801 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1802 PROCESSOR_CANNONLAKE.
1803 (get_builtin_code_for_version): Fix priority for
1804 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1805 PROCESSOR_SKYLAKE-AVX512.
1806 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1807 (processor_type): Add PROCESSOR_SKYLAKE.
1809 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1810 Jason Merrill <jason@redhat.com>
1813 * convert.c (convert_to_integer_1): Use direct recursion for
1814 enumeral types and types with a precision less than the number
1815 of bits in their mode.
1817 2018-04-16 Julia Koval <julia.koval@intel.com>
1820 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1821 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1823 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1826 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1827 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1828 and -mno-direct-move.
1830 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1833 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1834 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1835 Ensure negative shifts result in {0}.
1837 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1839 PR rtl-optimization/79916
1840 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1841 regs (if any) to define how to gnerate SD moves when LRA is in
1844 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1846 PR rtl-optimization/85393
1847 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1848 * except.c (expand_dw2_landing_pad_for_region): Make static.
1849 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1850 a label and unconditional jump to old_bb, rather than
1851 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1854 PR rtl-optimization/85376
1855 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1856 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1857 instead of a specific value.
1859 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1860 Bin Cheng <bin.cheng@arm.com>
1862 PR tree-optimization/82965
1863 PR tree-optimization/83991
1864 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1865 by_profile_only parameter.
1866 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1867 information if the loop was predicted to iterate too many times.
1868 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1870 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1873 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1876 2018-04-13 Martin Liska <mliska@suse.cz>
1877 Jakub Jelinek <jakub@redhat.com>
1880 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1881 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1882 * builtins.c (expand_builtin_memory_copy_args): Use
1883 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1884 handle dest_addr == pc_rtx.
1886 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1889 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1890 asked to not generate direct moves.
1891 (fix_trunc<mode>si2_stfiwx): Similar.
1892 (fix_trunc<mode>si2_internal): Similar.
1894 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1897 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1898 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1899 lookup if dest in some wider mode is known to be const0_rtx and
1900 if so, record permanent equivalence for it to be ZERO_EXTEND of
1901 the narrower mode destination.
1903 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1905 * lto-streamer-out.c (output_function): Revert 259346.
1906 * omp-expand.c (expand_oacc_for): Likewise.
1908 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1910 PR rtl-optimization/85354
1911 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1912 * sel-sched.c (sel_global_init): ... here.
1914 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1917 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1918 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1919 mode for PE-COFF targets.
1920 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1921 (i386_pe_asm_lto_end): Likewise.
1922 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1923 (TARGET_ASM_LTO_END): Likewise.
1924 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1925 (i386_pe_asm_lto_start): New function.
1926 (i386_pe_asm_lto_end): Likewise.
1928 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1929 Richard Biener <rguenther@suse.de>
1932 * lto-streamer-out.c (output_function): Fix CFG loop state before
1934 * omp-expand.c (expand_oacc_for): Handle calls to internal
1935 functions like regular functions.
1937 2018-04-12 Richard Biener <rguenther@suse.de>
1940 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1941 for the early LTO debug to properly generate references to it
1942 during DIE emission. Do not re-use that for the skeleton for
1944 (dwarf2out_early_finish): Likewise.
1946 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1949 * config/i386/sse.md
1950 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1951 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1952 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1953 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1954 and output is a reg, avoid creating invalid lowpart subreg, but
1955 instead split into a 512-bit move. Don't split if not AVX512VL,
1956 input is xmm16+ reg and output is a mem.
1957 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1958 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1959 xmm16+ reg and output is a mem.
1961 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1963 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1964 also for flag_dwarf2_cfi_asm.
1966 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1968 PR rtl-optimization/85342
1969 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1970 a bool scalar var inside of the loop instead. Don't try to update
1971 recog_data.operand after failed apply_change_group.
1973 2018-04-12 Tom de Vries <tom@codesourcery.com>
1976 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1977 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1978 array with flexible array member as array without given dimension.
1979 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1980 argument for undefined param to true.
1982 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1985 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1986 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1987 from PowerPC section.
1988 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1989 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1990 help text that is too long.
1991 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1992 help text that is too long.
1993 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1994 help text that is too long.
1996 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1998 * config/alpha/alpha.md (stack_probe_internal): Rename
1999 from "probe_stack". Update all callers.
2001 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2003 PR rtl-optimization/84566
2004 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2005 sched_macro_fuse_insns.
2007 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2010 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2011 (compute_block_dependences): ... from here.
2013 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2015 PR tree-optimization/85331
2016 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2017 from int to HOST_WIDE_INT.
2019 2018-04-11 Martin Jambor <mjambor@suse.cz>
2022 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2023 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2024 not the same as the source val.
2025 (cgraph_edge_brings_value_p): New parameter.
2026 (gather_edges_for_value): Pass destination value to
2027 cgraph_edge_brings_value_p.
2028 (perhaps_add_new_callers): Likewise.
2029 (get_info_about_necessary_edges): Likewise and exclude values brought
2030 only by self-recursive edges.
2031 (create_specialized_node): Redirect only clones of self-calling edges.
2032 (+self_recursive_pass_through_p): New function.
2033 (find_more_scalar_values_for_callers_subset): Use it.
2034 (find_aggregate_values_for_callers_subset): Likewise.
2035 (known_aggs_to_agg_replacement_list): Removed.
2036 (decide_whether_version_node): Re-calculate known constants for all
2037 remaining context clones.
2039 2018-04-11 Richard Biener <rguenther@suse.de>
2042 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2043 from early DWARF output.
2044 (dwarf2out_early_finish): Output line info unconditionally into
2045 early DWARF and add reference to it.
2047 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2050 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2051 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2052 other than V2DFmode using iptr mode attribute.
2053 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2055 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2057 PR rtl-optimization/84659
2058 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2060 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2063 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2065 (output_loc_list): Pass address of a dummy size variable even in the
2066 locview handling loop.
2067 (index_location_lists): Add comment on why skip_loc_list_entry can't
2070 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2073 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2076 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2079 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2080 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2081 and -mstring-compare-inline-limit.
2083 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2086 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2087 for stack clash protection in a register whenever we need it to be in
2090 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2092 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2093 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2095 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2098 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2100 (mlong-double-): Ditto.
2101 * config/rs6000/sysv4.opt (msdata=): Ditto.
2102 (mtls-size=): Ditto.
2104 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2106 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2107 erroneous entries for
2108 "vector int vec_ldl (int, long int *)", and
2109 "vector unsigned int vec_ldl (int, unsigned long int *)".
2110 Add comments and entries for
2111 "vector bool char vec_ldl (int, bool char *)",
2112 "vector bool short vec_ldl (int, bool short *)",
2113 "vector bool int vec_ldl (int, bool int *)",
2114 "vector bool long long vec_ldl (int, bool long long *)",
2115 "vector pixel vec_ldl (int, pixel *)",
2116 "vector long long vec_ldl (int, long long *)",
2117 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2118 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2119 type tree bool_long_long_type_node and correct definition of
2120 bool_V2DI_type_node to make reference to this new type tree.
2121 (rs6000_mangle_type): Replace erroneous reference to
2122 bool_long_type_node with bool_long_long_type_node.
2123 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2124 comments to emphasize sign distinctions for char and int types and
2125 replace RS6000_BTI_bool_long constant with
2126 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2127 use of RS6000_BTI_pixel.
2128 (bool_long_type_node): Remove this macro definition.
2129 (bool_long_long_type_node): New macro definition
2131 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2133 PR rtl-optimization/85300
2134 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2135 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2136 simplify_unary_operation fails.
2138 2018-04-10 Martin Liska <mliska@suse.cz>
2140 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2141 cgraph_edge and ipa_ref.
2143 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2147 * config/i386/sse.md
2148 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2149 computation of the VEC_MERGE selector from mask.
2150 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2151 Fix decoding of the VEC_MERGE selector into mask.
2153 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2155 PR tree-optimization/85286
2156 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2158 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2160 * final.c (final_1): Set insn_last_address as well as
2161 insn_current_address.
2163 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2166 * explow.c (emit_stack_probe): Call validize_mem on memory location
2167 before passing it to gen_probe_stack. Create address operand and
2168 legitimize it for the probe_stack_address case.
2170 2018-04-09 Jan Hubicka <jh@suse.cz>
2173 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2174 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2175 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2177 (free_lang_data): Rebuild type inheritance graph.
2179 2018-04-09 Martin Sebor <msebor@redhat.com>
2181 * invoke.texi (-finline-small-functions): Mention other optimization
2183 (-findirect-inlining, -fpartial-inlining): Same.
2184 (-finline-functions-called-once): Same.
2185 (-freorder-blocks-and-partition): Same.
2187 2018-04-09 Jan Hubicka <jh@suse.cz>
2190 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2191 jumps; choose last target that matches the criteria (i.e.
2192 no partition changes for non-crossing jumps).
2193 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2194 support for redirecting crossing jumps to non-crossing.
2196 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2198 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2199 also for naked functions.
2201 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2203 * config/arc/arc.md (add_shift): New pattern.
2204 (add_shift2): Likewise.
2205 (sub_shift): Likewise.
2206 (sub_shift_cmp0_noout): Likewise.
2207 (compare_si_ashiftsi): Likewise.
2208 (xbfu_cmp0_noout): New combine pattern.
2209 (xbfu_cmp0"): Likewise.
2210 (movsi_set_cc_insn): Place the predicable variant first.
2211 (commutative_binary_cmp0_noout): Remove clobber.
2212 (commutative_binary_cmp0): New pattern.
2213 (noncommutative_binary_cmp0): Likewise.
2214 (noncommutative_binary_cmp0_noout): Likewise.
2215 (noncommutative_binary_comparison_result_used): Removed.
2216 (rsub_cmp0): New pattern.
2217 (rsub_cmp0_noout): Likewise.
2218 (extzvsi): Changed, keep only meaningful variants.
2219 (SQH, SEZ): New iterators.
2220 (SQH_postfix): New mode attribute.
2221 (SEZ_prefix): New code attribute.
2222 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2223 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2224 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2226 (noncommutative_operator): Check the availability of barrel
2229 2018-04-09 Richard Biener <rguenther@suse.de>
2231 PR tree-optimization/85284
2232 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2233 Only use the niter constraining form of simple_iv when the exit
2236 2018-04-09 Tom de Vries <tom@codesourcery.com>
2239 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2240 (define_expand "*memory_barrier"): New define_expand.
2241 (define_insn "memory_barrier"): New insn.
2243 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2245 PR rtl-optimization/80463
2246 PR rtl-optimization/83972
2247 PR rtl-optimization/83480
2249 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2250 correct producer for the insn.
2251 (tidy_control_flow): Fixup seqnos in case of debug insns.
2253 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2255 PR rtl-optimization/83913
2257 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2258 different sched-times when merging exprs.
2260 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2262 PR rtl-optimization/83962
2264 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2265 tidy_fallthru_edge and tidy_control_flow.
2267 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2269 PR rtl-optimization/83530
2271 * sel-sched.c (force_next_insn): New global variable.
2272 (remove_insn_for_debug): When force_next_insn is true, also leave only
2273 next insn in the ready list.
2274 (sel_sched_region): When the region wasn't scheduled, make another pass
2275 over it with force_next_insn set to 1.
2277 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2279 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2281 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2282 for interrupt control.
2283 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2284 functions for interrupt control.
2285 * config/nds32/nds32-intrinsic.md: Likewise.
2286 * config/nds32/nds32_intrinsic.h: Likewise.
2287 * config/nds32/nds32.h (nds32_builtins): Likewise.
2289 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2291 * config/nds32/nds32.c (nds32_init_machine_status,
2292 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2293 strict_aligned_p field.
2294 (nds32_expand_to_rtl_hook): New function.
2295 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2296 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2298 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2299 Chung-Ju Wu <jasonwucj@gmail.com>
2301 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2302 * config/nds32/nds32-n7.md: New file.
2303 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2304 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2306 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2307 * config/nds32/nds32.md (pipeline_model): Add n7.
2308 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2309 * config/nds32/pipelines.md: Include n7 settings.
2311 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2312 Chung-Ju Wu <jasonwucj@gmail.com>
2314 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2315 * config/nds32/nds32-e8.md: New file.
2316 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2317 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2319 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2320 * config/nds32/nds32.md (pipeline_model): Add e8.
2321 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2322 * config/nds32/pipelines.md: Include e8 settings.
2324 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2325 Chung-Ju Wu <jasonwucj@gmail.com>
2327 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2328 * config/nds32/nds32-n8.md: New file.
2329 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2330 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2332 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2333 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2334 * config/nds32/nds32.md (pipeline_model): Add n8.
2335 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2336 * config/nds32/pipelines.md: Include n8 settings.
2338 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2339 Chung-Ju Wu <jasonwucj@gmail.com>
2341 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2342 * config/nds32/nds32-n9-2r1w.md: New file.
2343 * config/nds32/nds32-n9-3r2w.md: New file.
2344 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2345 nds32_register_ports): New or modify for cpu n9.
2346 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2348 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2349 * config/nds32/nds32-utils.c: New file.
2350 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2351 TARGET_MUL_SLOW): Define.
2352 * config/nds32/nds32.md (pipeline_model): New attribute.
2353 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2354 New options that support cpu n9.
2355 * config/nds32/pipelines.md: Include n9 settings.
2356 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2358 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2360 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2361 information if necessary.
2362 (output_cond_branch_compare_zero): Likewise.
2363 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2364 (nds32_target_alignment): Refine for alignment.
2365 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2366 (FUNCTION_BOUNDARY): Modify.
2367 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2369 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2371 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2373 * config/nds32/constants.md (unspec_volatile_element): Add values for
2374 TLB operation and data prefetch.
2375 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2376 functions for TLB operation and data prefetch.
2377 * config/nds32/nds32-intrinsic.md: Likewise.
2378 * config/nds32/nds32_intrinsic.h: Likewise.
2379 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2380 (nds32_print_operand): Likewise.
2381 * config/nds32/nds32.h (nds32_builtins): Likewise.
2383 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2384 Andrew Pinski <pinsika@gcc.gnu.org>
2387 * match.pd: Use constant_boolean_node of correct type instead of
2388 boolean_true_node or boolean_false_node for simplifying
2389 pointer comparisons to zero.
2391 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2393 PR tree-optimization/80021
2394 * tree.c (verify_type_variant): Make error call in verify_variant_match
2395 translatable and remove final full stop.
2397 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2399 * config/nds32/constants.md (unspec_volatile_element): Add
2400 UNSPEC_VOLATILE_EH_RETURN.
2401 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2402 nds32_output_stack_pop): Support dwarf exception handling process.
2403 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2404 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2405 exception handling process.
2406 (nds32_compute_stack_frame): Likewise.
2407 (nds32_return_addr_rtx): Likewise.
2408 (nds32_initial_elimination_offset): Likewise.
2409 (nds32_expand_prologue): Likewise.
2410 (nds32_expand_epilogue): Likewise.
2411 (nds32_dynamic_chain_address): New function.
2412 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2414 (DYNAMIC_CHAIN_ADDRESS): Define.
2415 (EH_RETURN_DATA_REGNO): Define.
2416 (EH_RETURN_STACKADJ_RTX): Define.
2417 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2418 patterns for dwarf exception handling.
2420 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2422 * config/nds32/nds32.h: Clean up obsolete macros.
2424 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2426 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2427 Add enum values for particular instructions.
2428 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2429 particular intrinsic functions.
2430 * config/nds32/nds32-intrinsic.md: Likewise.
2431 * config/nds32/nds32_intrinsic.h: Likewise.
2432 * config/nds32/nds32.h (nds32_builtins): Likewise.
2433 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2434 (btst, ave): New patterns for particular instructions.
2436 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2438 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2439 Add enum values for atomic load/store and memory sync.
2440 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2442 * config/nds32/nds32-intrinsic.md: Likewise.
2443 * config/nds32/nds32_intrinsic.h: Likewise.
2444 * config/nds32/nds32.h (nds32_builtins): Likewise.
2446 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2448 PR tree-optimization/85257
2449 * fold-const.c (native_encode_vector): If not all elts could fit
2450 and off is -1, return 0 rather than offset.
2451 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2452 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2453 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2454 adjust buffer in native_interpret_expr call.
2456 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2458 * config/nds32/constants.md (unspec_volatile_element): Add cache
2459 control enum values.
2460 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2461 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2462 * config/nds32/nds32.c (nds32_cctl_names): New.
2463 (nds32_print_operand): Handle cache control register names.
2464 * config/nds32/nds32.h (nds32_builtins): New enum values.
2465 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2467 * config/nds32/nds32.md (type): Add mmu.
2468 * config/nds32/pipelines.md (simple_insn): Add mmu.
2470 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2472 * config/nds32/nds32.md (type): Remove call.
2473 * config/nds32/pipelines.md (simple_insn): Likewise.
2475 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2477 * config/nds32/constants.md (unspec_volatile_element): Add
2478 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2479 UNSPEC_VOLATILE_FMFCFG.
2480 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2481 description for fmfcfg and fmfcsr.
2482 (bdesc_1arg): Add fmtcsr.
2483 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2484 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2485 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2486 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2487 unspec_fmfcfg): New patterns.
2488 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2489 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2490 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2491 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2492 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2493 __nds32__fmfcfg): Define.
2495 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2497 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2498 intrinsic register names.
2499 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2500 intrinsic register enum values and macros.
2502 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2504 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2505 for load/store addressing form.
2506 (nds32_print_operand_address): Likewise.
2508 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2511 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2512 based on LABEL_REF. Remove useless assertion.
2513 (pic_address_needs_scratch): Fix formatting.
2514 (sparc_legitimize_pic_address): Minor tweaks.
2515 (sparc_delegitimize_address): Adjust assertion accordingly.
2516 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2517 into symbolic_operand.
2518 (movsi_high_pic_label_ref): Likewise.
2519 (movsi_lo_sum_pic_label_ref): Likewise.
2520 (movdi_pic_label_ref): Likewise.
2521 (movdi_high_pic_label_ref): Likewise.
2522 (movdi_lo_sum_pic_label_ref): Likewise.
2524 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2526 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2527 custom LIB_SPEC setup.
2529 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2530 Kito Cheng <kito.cheng@gmail.com>
2532 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2533 * config/riscv/freebsd.h: New.
2535 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2537 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2538 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2541 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2542 Kito Cheng <kito.cheng@gmail.com>
2544 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2545 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2546 * config/nds32/nds32-protos.h (nds32_output_call,
2547 nds32_output_return): Declare.
2548 * config/nds32/nds32.md: Refine all the call and return patterns.
2550 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2553 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2554 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2556 PR rtl-optimization/84872
2557 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2558 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2561 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2563 * expr.c (copy_blkmode_to_reg): Revert 254862.
2564 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2566 2018-04-06 Richard Biener <rguenther@suse.de>
2569 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2570 after seeing a component reference with an adjacent field. Treat
2571 refs to arrays at struct end of external decls similar to
2572 refs to unconstrained commons.
2574 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2577 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2578 look through SAVE_EXPRs with non-side-effects argument. Adjust
2580 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2583 2018-04-06 Richard Biener <rguenther@suse.de>
2586 * alias.c (find_base_term): New wrapper around find_base_term
2587 unwinding CSELIB_VAL_PTR changes.
2588 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2591 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2593 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2595 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2596 constant definitions.
2597 ("nop"): lr 0,0 -> nopr r0
2598 ("nop_lr0", "nop_lr1"): New insn definitions.
2600 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2602 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2603 NDS32_V3PUSH_AVAILABLE_P macro.
2605 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2606 Chung-Ju Wu <jasonwucj@gmail.com>
2608 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2609 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2610 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2611 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2612 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2613 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2614 * config/nds32/constraints.md: New constraints and checking for hard
2615 float configuration.
2616 * config/nds32/iterators.md: New mode iterator and attribute for hard
2617 float configuration.
2618 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2620 * config/nds32/nds32-fpu.md: New file.
2621 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2622 deal with hard float code generation.
2623 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2625 (abi_type, float_reg_number): New enum type.
2626 * config/nds32/nds32-predicates.c: New predicates for hard float.
2627 * config/nds32/nds32-protos.h: Declare functions for hard float.
2628 * config/nds32/nds32.c: Implementation for hard float configuration.
2629 * config/nds32/nds32.h: Definitions for hard float configuration.
2630 * config/nds32/nds32.md: Include hard float machine description and
2631 modify patterns for hard float configuration.
2632 * config/nds32/nds32.opt: New options for hard float configuration.
2633 * config/nds32/predicates.md: New predicates for hard float
2636 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2638 * common/config/nds32/nds32-common.c
2639 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2641 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2644 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2645 CONSTRUCTOR_ELT (ctor, ...)->value.
2647 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2650 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2652 2018-04-05 Tom de Vries <tom@codesourcery.com>
2655 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2658 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2659 Kito Cheng <kito.cheng@gmail.com>
2661 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2662 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2663 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2664 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2666 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2667 Kito Cheng <kito.cheng@gmail.com>
2669 * config/nds32/constraints.md (Ufe): New memory constraint.
2670 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2671 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2672 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2674 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2675 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2677 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2679 * config/nds32/nds32.md: Use optimize_size in the condition for
2680 alu-shift instructions.
2682 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2684 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2686 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2688 * config/nds32/nds32.md (negsi2): Refine pattern.
2690 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2691 Chung-Ju Wu <jasonwucj@gmail.com>
2693 * config/nds32/iterators.md (shift_rotate): New code iterator.
2694 (shift): New code attribute.
2695 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2696 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2697 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2698 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2699 bit-wise operations.
2700 (andsi3, *andsi3): Ditto.
2701 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2702 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2703 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2704 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2705 nds32_ior_operand, nds32_xor_operand): New predicates.
2707 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2709 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2710 (addsi3, subsi3): ... this.
2712 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2714 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2716 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2718 * config/nds32/nds32.md: Adjust indention.
2720 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2722 * config/nds32/nds32.md (feature): New attribute.
2724 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2726 * config/nds32/nds32.md (subtype): New attribute.
2728 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2731 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2732 expansion to perform a bitwise AND of the argument followed by a
2733 boolean negation of the result.
2735 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2737 PR rtl-optimization/84878
2738 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2739 the basic block. Assert the use reference is not artificial and that
2740 it has an associated insn.
2742 2018-04-04 Michael Matz <matz@suse.de>
2744 * builtins.c (compute_objsize): Pass correct operand
2745 to array_at_struct_end_p.
2747 2018-04-04 Richard Biener <rguenther@suse.de>
2750 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2751 from contexts for DINFO_LEVEL_TERSE and below.
2753 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2755 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2756 resiter_operand condition.
2757 * config/nds32/nds32.md (*move<mode>): Ditto.
2759 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2760 Monk Chiang <sh.chiang04@gmail.com>
2762 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2764 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2766 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2768 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2769 Kito Cheng <kito.cheng@gmail.com>
2771 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2772 nds32_cond_code_str, output_cond_branch,
2773 output_cond_branch_compare_zero, nds32_expand_cbranch,
2774 nds32_expand_cstore, nds32_expand_movcc,
2775 nds32_output_cbranchsi4_equality_zero,
2776 nds32_output_cbranchsi4_equality_reg,
2777 nds32_output_cbranchsi4_equality_reg_or_const_int,
2778 nds32_output_cbranchsi4_greater_less_zero: New functions.
2779 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2780 nds32_expand_cstore, nds32_expand_movcc,
2781 nds32_output_cbranchsi4_equality_zero,
2782 nds32_output_cbranchsi4_equality_reg,
2783 nds32_output_cbranchsi4_equality_reg_or_const_int,
2784 nds32_output_cbranchsi4_greater_less_zero): Declare.
2785 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2786 nds32_rimm11s_operand): New predicates.
2787 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2788 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2791 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2793 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2794 * config/nds32/nds32.md: Ditto.
2795 * config/nds32/pipelines.md: Ditto.
2797 2018-04-04 Richard Biener <rguenther@suse.de>
2799 PR tree-optimization/85168
2800 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2801 propagating abnormals.
2803 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2805 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2807 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2808 Kito Cheng <kito.cheng@gmail.com>
2810 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2811 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2812 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2813 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2814 * config/nds32/nds32.md (sibcall_internal): New.
2815 (sibcall_register): Remove.
2816 (sibcall_immediate): Remove.
2817 (sibcall_value_internal): New.
2818 (sibcall_value_register): Remove.
2819 (sibcall_value_immediate): Remove.
2820 * config/nds32/predicates.md (nds32_general_register_operand): New.
2821 (nds32_call_address_operand): New.
2823 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2825 PR rtl-optimization/85167
2826 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2827 bb_defs if *split_p, instead preinitialize it to NULL.
2829 PR tree-optimization/85156
2830 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2831 evaluating the argument multiple times.
2833 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2835 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2837 (_mm_cvtpd_ps): Likewise.
2838 (_mm_cvttpd_epi32): Likewise.
2839 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2840 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2841 vector, pixel, and bool following altivec.h include.
2843 2018-04-03 Martin Sebor <msebor@redhat.com>
2845 * doc/extend.texi (Common Function Attributes): Clarify.
2846 (const attribute): Likewise.
2847 (pure attribute): Likewise.
2849 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2852 * config/i386/i386.c (ix86_expand_vector_set): Use
2853 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2855 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2857 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2858 instructions when changing rounding bits to preserve precision bits
2859 in the x87 control word.
2861 2018-04-03 Martin Liska <mliska@suse.cz>
2863 PR tree-optimization/82491
2864 * rtl.h (strip_offset_and_add): Replace += suboffset with
2865 poly_uint64 () + suboffset.
2867 2018-03-29 Martin Liska <mliska@suse.cz>
2868 Martin Jambor <mjambor@suse.cz>
2871 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2872 param_type is not an integral or pointer type.
2874 2018-04-03 Richard Biener <rguenther@suse.de>
2876 * sese.h (recompute_all_dominators): Remove.
2878 2018-04-02 Martin Sebor <msebor@redhat.com>
2880 * doc/invoke.texi (-Wrestrict): Fix typos.
2882 2018-04-02 Jim Wilson <jimw@sifive.com>
2884 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2885 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2886 (<optab>di3, <optab>si3_extend): Likewise.
2887 (<optab>si3_mask, <optab>si3_mask_1): New.
2888 (<optab>di3_mask, <optab>di3_mask_1): New.
2889 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2890 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2891 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2893 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2895 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2898 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2900 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2901 (nds32_canonicalize_comparison): New function.
2903 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2904 Kito Cheng <kito.cheng@gmail.com>
2905 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2907 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2908 * config/nds32/constants.md (unspec_volatile_element): Add
2909 UNSPEC_VOLATILE_RELAX_GROUP.
2910 * config/nds32/nds32-relax-opt.c: New file.
2911 * config/nds32/nds32-predicates.c
2912 (nds32_symbol_load_store_p): New function.
2913 * config/nds32/nds32-protos.h
2914 (nds32_symbol_load_store_p): Declare function.
2915 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2916 * config/nds32/nds32.c
2917 (nds32_register_pass): New function to register pass.
2918 (nds32_register_passes): New function to register passes.
2919 * config/nds32/nds32.md (relax_group): New pattern.
2920 * config/nds32/nds32.opt (mrelax-hint): New option.
2921 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2923 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2925 * config/nds32/t-nds32: Modify files dependency.
2927 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2929 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2930 (PROFILE_HOOK): Define its implementation.
2932 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2934 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2935 type and 32-bit size.
2937 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2940 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2941 (V_128_256): New mode iterator.
2942 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2943 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2944 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2946 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2947 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2949 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2952 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2953 NaN inputs correctly.
2955 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2958 * config/rs6000/vsx.md (??r): New mode attribute.
2959 (*vsx_mov<mode>_64bit): Use it.
2960 (*vsx_mov<mode>_32bit): Likewise.
2962 2018-03-30 Martin Sebor <msebor@redhat.com>
2964 PR tree-optimization/84818
2965 * builtins.c (check_access): Use warning_n.
2967 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2970 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2972 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2975 2018-03-30 Julia Koval <julia.koval@intel.com>
2978 * x86-tune.def (movx, partial_reg_dependency): Enable for
2981 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2984 * lra-constraints.c (process_alt_operands): Move setting
2985 this_alternative_matches below.
2987 2018-03-29 Martin Liska <mliska@suse.cz>
2990 * doc/invoke.texi: Document how LTO works with debug info.
2991 Describe auto-load support of binutils. Mention 'x86-64'
2992 as valid option value of -march option.
2994 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2996 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2999 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3000 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3001 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3004 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3007 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3008 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3009 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3010 for RS6000_BTM_POWERPC64.
3011 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3012 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3013 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3018 2018-03-28 Carl Love <cel@us.ibm.com>
3021 2017-09-27 Carl Love <cel@us.ibm.com>
3023 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3024 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3025 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3028 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3030 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3031 instead of __vector bool.
3032 (_mm_max_pu8): Likewise.
3033 (_mm_min_pi16): Likewise.
3035 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3038 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3039 (DIVWEUO): Likewise.
3041 (DIVDEUO): Likewise.
3042 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3043 DIVWEUO and DIVDEUO.
3044 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3045 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3046 (div_extend): Likewise.
3047 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3049 (__builtin_divweuo): Likewise.
3050 (__builtin_divdeo): Likewise.
3051 (__builtin_divdeuo): Likewise.
3053 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3056 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3057 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3059 PR tree-optimization/82004
3060 * gimple-match-head.c (optimize_pow_to_exp): New function.
3061 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3062 Don't fold to exp if optimize_pow_to_exp is false.
3064 2018-03-28 Martin Liska <mliska@suse.cz>
3067 * calls.c (initialize_argument_information): Fix trailing space.
3068 * common.opt: Fix typo and provide better explanation for
3069 -fsanitize-coverage option.
3070 * config/i386/i386.opt: Fix typo.
3072 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3073 Martin Liska <mliska@suse.cz>
3076 * gimplify.c (asan_poison_variable): Don't do the check for
3077 gimplify_omp_ctxp here.
3078 (gimplify_decl_expr): Do it here.
3079 (gimplify_target_expr): Likewise.
3081 2018-03-28 Martin Liska <mliska@suse.cz>
3084 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3085 chkp_type_bounds_count if MPX is not enabled.
3087 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3089 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3091 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3094 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3095 function to create the function decl for complex long double
3096 multiply and divide for -mabi=ieeelongdouble.
3097 (init_float128_ieee): Call it.
3099 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3102 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3103 -fcf-protection=branch -mibt.
3104 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3106 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3109 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3111 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3114 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3115 extern array declarations.
3117 2018-03-27 Richard Biener <rguenther@suse.de>
3120 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3121 explicit single_use checks.
3123 2018-03-27 Richard Biener <rguenther@suse.de>
3125 PR tree-optimization/85082
3126 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3129 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3131 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3132 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3133 Turn on fasynchronous-unwind-tables and funwind-tables.
3135 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3138 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3139 (*bmi_blsr_<mode>_ccz): Ditto.
3141 2018-03-26 Tom de Vries <tom@codesourcery.com>
3143 PR tree-optimization/85063
3144 * omp-general.c (offloading_function_p): New function. Factor out
3146 * omp-offload.c (pass_omp_target_link::gate): ... here.
3147 * omp-general.h (offloading_function_p): Declare.
3148 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3149 with attribute omp declare target for offloading functions.
3151 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3153 PR tree-optimization/84005
3154 * tree-data-ref.h (get_base_for_alignment): Declare.
3155 * tree-data-ref.c (get_base_for_alignment_1): New function.
3156 (get_base_for_alignment): Likewise.
3157 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3158 get_base_for_alignment to find a suitable base object, instead
3159 of always using drb->base_address.
3161 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3164 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3165 known size by default.
3167 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3170 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3171 and non BLKmode operands.
3173 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3176 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3177 Clean up attributes.
3179 2018-03-23 Richard Biener <rguenther@suse.de>
3182 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3183 we are going to emit early debug for LTO.
3185 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3188 * function.c (match_asm_constraints_1): Don't optimize if input
3189 doesn't satisfy general_operand predicate for output's mode.
3192 * alias.c (write_dependence_p): Don't require for x_canonicalized
3193 non-VOIDmode if x has VOIDmode.
3196 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3197 just don't try to optimize it rather than assert it never happens.
3199 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3201 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3202 macro expansions for definition of ST_INTERNAL_<mode> and
3203 LD_INTERNAL_<mode> builtins.
3204 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3206 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3208 (altivec_expand_st_builtin): Likewise.
3209 (altivec_expand_builtin): Remove calls to deleted functions.
3210 (rs6000_address_for_altivec): Delete this function.
3211 * config/rs6000/vector.md: Remove expands for
3212 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3214 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3217 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3218 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3219 re-computing once computed.
3220 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3221 (arm_init_machine_status): Initialize
3222 machine->static_chain_stack_bytes.
3224 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3227 * doc/extend.texi: Add four new prototypes for vec_ld.
3228 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3229 definitions for more logical presentation.
3230 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3231 entries for V1TI variants of __builtin_altivec_ld builtin.
3232 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3233 handling of V1TI variant of LVX icode pattern.
3234 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3235 (rs6000_gimple_fold_builtin): Likewise.
3236 (altivec_init_builtins): Add code to define
3237 __builtin_altivec_lvx_v1ti function.
3239 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3242 * function.c (match_asm_constraints_1): Don't do the optimization
3243 if input isn't a REG, SUBREG, MEM or constant.
3245 2018-03-22 Tom de Vries <tom@codesourcery.com>
3247 PR tree-optimization/84956
3248 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3249 bb_has_abnormal_pred.
3251 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3254 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3255 DECL_INITIAL (decl) to decl at the end.
3256 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3259 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3261 * doc/extend.texi (__builtin_tgmath): Document when complex
3262 integer types are treated as _Complex _Float64.
3264 2018-03-21 Tom de Vries <tom@codesourcery.com>
3266 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3268 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3270 PR tree-optimization/84960
3271 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3272 if it is ENTRY block, move them into single succ of ENTRY in that case.
3274 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3276 PR tree-optimization/84811
3277 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3278 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3279 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3280 (known_subrange_p): Update accordingly. Cast each value involved
3281 in the size comparison, rather than casting the result of the
3284 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3286 PR tree-optimization/84982
3287 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3288 by flipping the least significant bit rather than all bits from
3289 bitpos to bitpos + bitsize - 1.
3291 2018-03-21 Nathan Sidwell <nathan@acm.org>
3293 * doc/extend.texi (Deprecated Features): Remove mention of
3294 long-deleted deprecations.
3296 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3299 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3300 * configure: Regenerate.
3302 2018-03-21 Tom de Vries <tom@codesourcery.com>
3304 PR tree-optimization/83126
3305 * tree-parloops.c (num_phis): New function.
3306 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3308 2018-03-21 Nathan Sidwell <nathan@acm.org>
3310 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3311 mention anon-struct/union members and trailing attributes.
3313 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3315 PR tree-optimization/84969
3316 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3317 builtin memset partitions if they set different rhs values.
3319 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3321 PR rtl-optimization/84989
3322 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3323 VEC_DUPLICATE with scalar result mode.
3325 2018-03-21 Martin Liska <mliska@suse.cz>
3328 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3329 not intended return statement.
3331 2018-03-21 Martin Liska <mliska@suse.cz>
3334 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3335 (chkp_find_bound_slots_1): Limit number of iterations.
3337 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3340 * Minor grammar fixes for x86 options.
3342 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3345 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3346 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3348 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3351 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3352 (altivec_lvx_<mode>_1op): Likewise.
3353 (altivec_stvx_<mode>_2op): Likewise.
3354 (altivec_stvx_<mode>_1op): Likewise.
3355 (altivec_lvx_<VM2:mode>): New define_expand.
3356 (altivec_stvx_<VM2:mode>): Likewise.
3357 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3358 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3359 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3360 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3361 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3362 (rs6000_gen_lvx): Likewise.
3363 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3364 (altivec_expand_stv_builtin): Likewise.
3365 (altivec_expand_builtin): Likewise.
3366 * config/rs6000/vector.md: Likewise.
3368 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3371 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3374 2018-03-20 Richard Biener <rguenther@suse.de>
3377 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3378 sign-conversions as zero, fall back to standard scalar_stmt
3381 2018-03-20 Martin Liska <mliska@suse.cz>
3384 * predict.c (rebuild_frequencies): Handle case when we have
3385 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3387 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3390 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3391 flag_section_anchors.
3392 * varasm.c (use_blocks_for_decl_p): Remove hack for
3393 dw2_force_const_mem.
3396 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3398 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3399 be created, use lowpart_subreg of operands[0] rather than operands[0]
3401 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3402 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3403 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3404 and n constraint instead of aarch64_shift_imm_di and Usd.
3405 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3406 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3408 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3411 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3412 to favor GPR over NEON registers.
3413 (<shift>di3_neon): Likewise.
3415 2018-03-20 Tom de Vries <tom@codesourcery.com>
3418 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3419 (nvptx_process_pars): Emit bar.sync asap and alap.
3421 2018-03-20 Tom de Vries <tom@codesourcery.com>
3424 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3425 seen_label if seen_label is already set.
3427 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3430 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3431 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3432 Use 1U instead of 1. Formatting fixes.
3435 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3436 instead of TREE_TYPE (s1) for the return value.
3438 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3440 PR tree-optimization/84946
3441 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3442 bitsize + bitsize in poly_uint64 rather than poly_int64.
3445 * dwarf2asm.c: Include fold-const.c.
3446 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3447 of decl rather than decl itself.
3449 PR rtl-optimization/84643
3450 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3452 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3455 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3456 calling assemble_variable.
3458 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3461 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3462 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3464 2018-03-19 Jim Wilson <jimw@sifive.com>
3467 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3468 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3469 (riscv_first_stack_step): Likewise.
3470 (riscv_option_override): Use STACK_BOUNDARY instead of
3472 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3474 (BIGGEST_ALIGNMENT): Set to 128.
3475 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3476 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3479 2018-03-19 Richard Biener <rguenther@suse.de>
3481 PR tree-optimization/84933
3482 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3483 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3485 2018-03-19 Richard Biener <rguenther@suse.de>
3487 PR tree-optimization/84859
3488 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3489 (cond_if_else_store_replacement): Perform sinking operation on
3490 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3491 Generalize what a BB with a single eligible store is.
3493 2018-03-19 Richard Biener <rguenther@suse.de>
3495 PR tree-optimization/84929
3496 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3497 chrec_is_positive against non-chrec arg.
3499 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3502 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3504 2018-03-18 Martin Liska <mliska@suse.cz>
3506 PR rtl-optimization/84635
3507 * regrename.c (build_def_use): Use matches_mode only when
3510 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3512 PR tree-optimization/84913
3513 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3514 vectorize chains of COND_EXPRs.
3516 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3518 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3520 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3522 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3524 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3526 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3528 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3529 Kito Cheng <kito.cheng@gmail.com>
3531 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3532 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3533 (nds32_adjust_reg_alloc_order): New function.
3534 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3536 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3538 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3539 nds32_print_operand, nds32_print_operand_address): Use
3540 HOST_WIDE_INT_PRINT_DEC instead.
3542 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3544 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3546 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3549 * config/i386/i386.c (initial_ix86_tune_features,
3550 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3552 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3553 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3554 rather than 1u << ix86_tune. Formatting fix.
3555 (ix86_option_override_internal): Change ix86_arch_mask from
3556 unsigned int to unsigned HOST_WIDE_INT, initialize to
3557 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3558 (ix86_function_specific_restore): Likewise.
3560 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3563 * postreload.c (reload_combine_recognize_pattern): Perform
3564 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3565 truncate_int_for_mode the result for the destination's mode.
3568 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3569 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3571 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3574 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3577 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3578 regno_allocno_class_array and sorted_pseudos.
3579 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3580 insns where regno is used.
3582 2018-03-16 Martin Liska <mliska@suse.cz>
3585 * multiple_target.c (create_dispatcher_calls): Redirect
3586 reference in the symbol table.
3588 2018-03-16 Martin Liska <mliska@suse.cz>
3591 * multiple_target.c (create_dispatcher_calls): Redirect also
3594 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3598 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3599 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3600 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3602 2018-03-16 Julia Koval <julia.koval@intel.com>
3604 * doc/invoke.texi (Skylake Server): Add CLWB.
3605 Cannonlake): Remove CLWB.
3607 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3609 PR tree-optimization/84841
3610 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3612 (FLOAT_ONE_CONST_TYPE): Define.
3613 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3614 (sort_by_operand_rank): Put entries with higher constant_type last
3615 rather than first to match comments.
3617 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3619 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3622 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3625 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3626 check and use address of target always.
3628 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3631 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3632 (indirect_thunk_bnd_needed): Likewise.
3633 (indirect_thunks_used): Likewise.
3634 (indirect_thunks_bnd_used): Likewise.
3635 (indirect_return_needed): New.
3636 (indirect_return_bnd_needed): Likewise.
3637 (output_indirect_thunk_function): Add a bool argument for
3639 (output_indirect_thunk_function): Don't generate alias for
3640 function return thunk.
3641 (ix86_code_end): Call output_indirect_thunk_function to generate
3642 function return thunks.
3643 (ix86_output_function_return): Set indirect_return_bnd_needed
3644 and indirect_return_needed instead of indirect_thunk_bnd_needed
3645 and indirect_thunk_needed.
3647 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3649 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3650 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3651 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3653 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3654 Paul Hua <paul.hua.gm@gmail.com>
3657 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3659 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3661 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3662 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3665 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3668 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3669 instead of GET_MODE_SIZE when comparing Units.
3671 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3674 * varasm.c (hash_section): Return an unchangeble hash value
3675 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3676 Return !aarch64_can_use_per_function_literal_pools_p ().
3678 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3681 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3682 rather than address of cmode as last argument to prepare_cmp_insn.
3684 2018-03-15 Julia Koval <julia.koval@intel.com>
3686 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3687 F_AVX512VNNI, F_AVX512BITALG): New.
3689 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3692 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3693 insn for floating-point loads and stores.
3695 2018-03-14 Carl Love <cel@us.ibm.com>
3697 * config/rs6000/rs6000-c.c: Add macro definitions for
3698 ALTIVEC_BUILTIN_VEC_PERMXOR.
3699 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3700 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3701 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3702 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3704 * config/doc/extend.texi: Add prototypes for vec_permxor.
3706 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3709 * diagnostic-show-locus.c (class layout_point): Convert m_line
3710 from int to linenum_type.
3711 (line_span::comparator): Use linenum "compare" function when
3712 comparing line numbers.
3713 (test_line_span): New function.
3714 (layout_range::contains_point): Convert param "row" from int to
3716 (layout_range::intersects_line_p): Likewise.
3717 (layout::will_show_line_p): Likewise.
3718 (layout::print_source_line): Likewise.
3719 (layout::should_print_annotation_line_p): Likewise.
3720 (layout::print_annotation_line): Likewise.
3721 (layout::print_leading_fixits): Likewise.
3722 (layout::annotation_line_showed_range_p): Likewise.
3723 (struct line_corrections): Likewise for field m_row.
3724 (line_corrections::line_corrections): Likewise for param "row".
3725 (layout::print_trailing_fixits): Likewise.
3726 (layout::get_state_at_point): Likewise.
3727 (layout::get_x_bound_for_row): Likewise.
3728 (layout::print_line): Likewise.
3729 (diagnostic_show_locus): Likewise for locals "last_line" and
3731 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3732 * input.c (selftest::test_linenum_comparisons): New function.
3733 (selftest::input_c_tests): Call it.
3734 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3735 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3736 * selftest.h (ASSERT_GT): New macro.
3737 (ASSERT_GT_AT): New macro.
3738 (ASSERT_LT): New macro.
3739 (ASSERT_LT_AT): New macro.
3741 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3743 PR rtl-optimization/84780
3744 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3746 2018-03-14 Martin Liska <mliska@suse.cz>
3748 * tree.c (record_node_allocation_statistics): Use
3749 get_stats_node_kind.
3750 (get_stats_node_kind): New function extracted from
3751 record_node_allocation_statistics.
3752 (free_node): Use get_stats_node_kind.
3754 2018-03-14 Richard Biener <rguenther@suse.de>
3756 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3757 that the value-set of ANTIC_IN doesn't grow.
3760 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3762 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3763 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3765 2018-03-14 Julia Koval <julia.koval@intel.com>
3767 * config.gcc (icelake-client, icelake-server): New.
3769 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3770 (initial_ix86_arch_features): Ditto.
3771 (PTA_SKYLAKE): Add SGX.
3772 (PTA_ICELAKE): Remove.
3773 (PTA_ICELAKE_CLIENT): New.
3774 (PTA_ICELAKE_SERVER): New.
3775 (ix86_option_override_internal): Split up icelake on icelake client and
3777 (get_builtin_code_for_version): Ditto.
3778 (fold_builtin_cpu): Ditto.
3779 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3780 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3781 * config/i386/i386.h (processor_type): Ditto.
3782 * doc/invoke.texi: Ditto.
3784 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3787 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3788 INTEGER_CST offset, add it together with bitpos / 8 and
3789 sign extend based on POINTER_SIZE.
3793 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3796 * config/i386/constraints.md (Yc): New register constraint.
3797 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3798 Use Yc constraint for alternative 2 of operand 0. Remove
3799 preferred_for_speed attribute.
3801 2018-03-14 Richard Biener <rguenther@suse.de>
3803 PR tree-optimization/84830
3804 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3805 with the old one to avoid oscillations.
3807 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3810 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3812 (assign_by_spills): Return a flag of reload assignment failure.
3813 Do not process the reload assignment failures. Do not spill other
3814 reload pseudos if they has the same reg class. Update n if
3816 (lra_assign): Add a return arg. Set up from the result of
3817 assign_by_spills call.
3818 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3819 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3820 usage_insns if it is not NULL.
3821 (spill_hard_reg_in_range): New function.
3822 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3823 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3824 function prototypes.
3825 (lra_assign): Change prototype.
3826 * lra.c (lra): Add code to deal with fails by splitting hard reg
3829 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3831 * config/riscv/riscv.opt (mrelax): New option.
3832 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3833 "norelax" when riscv_mrelax is disabled.
3834 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3836 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3839 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3840 reassociation for int modes.
3842 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3844 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3845 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3847 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3848 * config/aarch64/aarch64-sve.md
3849 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3850 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3851 (*extend<mode><Vwide>2): Rename to...
3852 (aarch64_sve_extend<mode><Vwide>2): ...this.
3853 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3854 renaming the old pattern to...
3855 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3857 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3858 define_expand, renaming the old pattern to...
3859 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3860 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3861 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3862 account when deciding which SVE instruction the optab should use.
3863 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3865 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3867 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3868 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3869 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3870 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3871 (tlsdesc_small_<mode>): Turn a define_expand and use
3872 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3873 (tlsdesc_small_advsimd_<mode>): ...this.
3874 (tlsdesc_small_sve_<mode>): New pattern.
3876 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3878 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3879 (UNSPEC_UMUL_HIGHPART): New constants.
3880 (MUL_HIGHPART): New int iteraor.
3881 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3882 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3884 (*<su>mul<mode>3_highpart): New define_insn.
3886 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3889 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3892 2018-03-13 Martin Liska <mliska@suse.cz>
3895 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3897 (sem_item_optimizer::~sem_item_optimizer): Release it.
3898 (sem_item_optimizer::merge_classes): Register variable aliases.
3899 (sem_item_optimizer::fixup_pt_set): New function.
3900 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3901 * ipa-icf.h: Declare new variables and functions.
3903 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3906 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3907 integer_pow2p@2 and test integer_pow2p in condition.
3908 (A < 0 ? C : 0): Similarly for @1.
3911 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3912 characters starting at p contain '\0' character, don't look beyond
3916 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3917 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3920 * reg-stack.c (change_stack): Change update_end var from int to
3921 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3922 also call set_block_for_insn on the newly added insns and rescan.
3925 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3926 on the last operand.
3929 * tree.c (stabilize_reference_1): Return save_expr (e) for
3930 STATEMENT_LIST even if it doesn't have side-effects.
3932 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3934 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3936 2018-03-12 Renlin Li <renlin.li@arm.com>
3938 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3939 aarch64_output_scalar_simd_mov_immediate.
3941 2018-03-12 Martin Sebor <msebor@redhat.com>
3943 PR tree-optimization/83456
3944 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3945 for perfectly overlapping calls to memcpy.
3946 (gimple_fold_builtin_memory_chk): Same.
3947 (gimple_fold_builtin_strcpy): Handle no-warning.
3948 (gimple_fold_builtin_stxcpy_chk): Same.
3949 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3951 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3953 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3954 parameter. Use it for SFmode.
3955 (rs6000_function_arg_advance_1): Adjust.
3956 (rs6000_function_arg): Adjust.
3957 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3959 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3961 PR rtl-optimization/84169
3962 PR rtl-optimization/84780
3963 * combine.c (can_combine_p): Check for a 2-insn combination whether
3964 the destination register is used between the two insns, too.
3966 2018-03-12 Richard Biener <rguenther@suse.de>
3968 PR tree-optimization/84803
3969 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3970 for refs DR analysis didn't process.
3972 2018-03-12 Richard Biener <rguenther@suse.de>
3974 PR tree-optimization/84777
3975 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3976 force-vectorize loops ignore whether we are optimizing for size.
3978 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3980 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3981 (TARGET_MD_ASM_ADJUST): Define.
3983 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3984 Kito Cheng <kito.cheng@gmail.com>
3985 Chung-Ju Wu <jasonwucj@gmail.com>
3987 * config/nds32/nds32.c (nds32_compute_stack_frame,
3988 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3989 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3990 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3991 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3992 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3993 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3994 * config/nds32/nds32.md (prologue, epilogue): Use macro
3995 NDS32_V3PUSH_AVAILABLE_P to do checking.
3997 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4000 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4001 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4002 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4003 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4004 addition of most attributes on !orig_type_die or the attribute not
4005 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4007 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4008 Chung-Ju Wu <jasonwucj@gmail.com>
4010 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4012 * config/nds32/nds32.opt (mvh): New option.
4014 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4015 Chung-Ju Wu <jasonwucj@gmail.com>
4017 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4019 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4020 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4023 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4024 Chung-Ju Wu <jasonwucj@gmail.com>
4026 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4028 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4029 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4031 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4032 Kito Cheng <kito.cheng@gmail.com>
4033 Chung-Ju Wu <jasonwucj@gmail.com>
4035 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4036 UNSPEC_FFMISM and UNSPEC_FLMISM.
4037 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4038 for ffb, ffmism and flmism.
4039 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4040 (unspec_ffmism): Ditto.
4041 (unspec_flmism): Ditto.
4042 (nds32_expand_builtin_impl): Check if string extension is available.
4043 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4044 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4046 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4049 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4052 * lra-assigns.c (assign_by_spills): Return a flag of reload
4053 assignment failure. Do not process the reload assignment
4054 failures. Do not spill other reload pseudos if they has the same
4056 (lra_assign): Add a return arg. Set up from the result of
4057 assign_by_spills call.
4058 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4059 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4060 usage_insns if it is not NULL.
4061 (spill_hard_reg_in_range): New function.
4062 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4063 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4064 function prototypes.
4065 (lra_assign): Change prototype.
4066 * lra.c (lra): Add code to deal with fails by splitting hard reg
4069 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4072 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4074 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4077 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4078 (dw_val_node): Add val_symbolic_view.
4079 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4080 (symview_upper_bound): New.
4081 (new_line_info_table): Initialize symviews_since_reset.
4082 (dwarf2out_source_line): Count symviews_since_reset and set
4083 symview_upper_bound.
4084 (dw_val_equal_p): Handle symview.
4085 (add_AT_symview): New.
4086 (print_dw_val): Handle symview.
4087 (attr_checksum, attr_checksum_ordered): Likewise.
4088 (same_dw_val_p, size_of_die): Likewise.
4089 (value_format, output_die): Likewise.
4090 (add_high_low_attributes): Use add_AT_symview for entry_view.
4091 (dwarf2out_finish): Reset symview_upper_bound, clear
4094 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4097 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4098 Add strict argument and use it.
4099 (rs6000_split_multireg_move): Update for new strict argument.
4100 (mem_operand_gpr): Disallow all non-offsettable addresses.
4101 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4103 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4106 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4107 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4108 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4111 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4112 decl, use remap_type if we want to use the type.
4114 2018-03-09 Martin Sebor <msebor@redhat.com>
4116 PR tree-optimization/84526
4117 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4119 (builtin_access::generic_overlap): Be prepared to handle non-array
4122 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4124 PR rtl-optimization/84682
4125 * lra-constraints.c (process_address_1): Check is_address flag
4126 for address constraints.
4127 (process_alt_operands): Likewise.
4128 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4129 preprocess_constraints.
4130 * recog.h (preprocess_constraints): Add oploc parameter.
4132 * recog.c (preprocess_constraints): Test address_operand for
4133 CT_ADDRESS constraints.
4135 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4138 * lra-assigns.c (assign_by_spills): Return a flag of reload
4139 assignment failure. Do not process the reload assignment
4140 failures. Do not spill other reload pseudos if they has the same
4142 (lra_assign): Add a return arg. Set up from the result of
4143 assign_by_spills call.
4144 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4145 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4146 usage_insns if it is not NULL.
4147 (spill_hard_reg_in_range): New function.
4148 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4149 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4150 function prototypes.
4151 (lra_assign): Change prototype.
4152 * lra.c (lra): Add code to deal with fails by splitting hard reg
4155 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4158 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4159 Accept complain bool parameter. Only emit errors if it is true.
4160 (arm_parse_cpu_option_name): Likewise.
4161 (arm_target_thumb_only): Adjust callers of the above.
4162 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4163 prototype to take a default true bool parameter.
4164 (arm_parse_arch_option_name): Likewise.
4166 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4167 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4171 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4172 * configure: Regenerate.
4173 * configure.ac ("linker --version-script option"): New.
4174 ("linker soname option"): New.
4176 2018-03-09 Richard Biener <rguenther@suse.de>
4178 PR tree-optimization/84775
4179 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4180 immediate uses of predicate stmts and mark them modified.
4183 PR tree-optimization/84178
4184 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4186 (version_loop_for_if_conversion): Delay update_ssa call.
4187 (tree_if_conversion): Delay update_ssa until after predicate
4190 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4193 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4194 when the function accesses prior frames.
4196 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4199 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4200 gen_llsym, otherwise call maybe_gen_llsym.
4203 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4204 has ',' character inside of it.
4206 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4209 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4210 as clobbering CC_REGNUM.
4212 2018-03-08 Richard Biener <rguenther@suse.de>
4215 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4216 (follow_copies_to_constant): Do not follow SSA names registered
4219 2018-03-08 Richard Biener <rguenther@suse.de>
4221 PR tree-optimization/84178
4222 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4224 (version_loop_for_if_conversion): Delay update_ssa call.
4225 (tree_if_conversion): Delay update_ssa until after predicate
4228 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4230 PR tree-optimization/84178
4231 * tree-if-conv.c (release_bb_predicate): Remove the
4232 the assertion that the stmts have NULL use_ops.
4233 Discard the statements, asserting that they haven't
4234 yet been added to a BB.
4236 2018-03-08 Richard Biener <rguenther@suse.de>
4238 PR tree-optimization/84746
4239 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4240 (phi_translate): Pass in destination ANTIC_OUT set.
4241 (phi_translate_1): Likewise. For a simplified result lookup
4242 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4243 (phi_translate_set): Adjust.
4244 (do_pre_regular_insertion): Likewise.
4245 (do_pre_partial_partial_insertion): Likewise.
4247 2018-03-08 Martin Liska <mliska@suse.cz>
4249 PR gcov-profile/84735
4250 * doc/gcov.texi: Document usage of profile files.
4251 * gcov-io.h: Document changes in the format.
4253 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4257 * dwarf2out.c (struct dw_line_info_table): Update comments for
4259 (FORCE_RESET_NEXT_VIEW): New.
4260 (FORCE_RESETTING_VIEW_P): New.
4261 (RESETTING_VIEW_P): Check for -1 too.
4262 (ZERO_VIEW_P): Likewise.
4263 (new_line_info_table): Force-reset next view.
4264 (dwarf2out_begin_function): Likewise.
4265 (dwarf2out_source_line): Simplify zero_view_p initialization.
4266 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4267 view directly. Omit view when omitting .loc at line 0.
4269 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4271 PR tree-optimization/84740
4272 * tree-switch-conversion.c (process_switch): Call build_constructors
4273 only if info.phi_count is non-zero.
4275 PR tree-optimization/84739
4276 * tree-tailcall.c (find_tail_calls): Check call arguments against
4277 DECL_ARGUMENTS (current_function_decl) rather than
4278 DECL_ARGUMENTS (func) when checking for tail recursion.
4280 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4282 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4283 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4284 Volker Reichelt's entry and add entries for people that perform
4285 GCC fuzzy testing and report numerous bugs.
4287 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4290 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4291 readonly data in sdata, if that is disabled.
4292 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4293 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4294 -mreadonly-in-sdata option.
4296 2018-03-07 Martin Sebor <msebor@redhat.com>
4298 PR tree-optimization/84468
4299 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4300 basic block when looking for nul assignment.
4302 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4305 * except.h (output_function_exception_table): Adjust prototype.
4306 * except.c (output_function_exception_table): Remove FNNAME parameter
4307 and add SECTION parameter. Ouput one part of the table at a time.
4308 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4309 the first part of the exception table and emit unwind directives.
4310 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4311 (i386_pe_seh_cold_init): Likewise.
4312 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4313 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4314 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4315 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4316 * config/i386/winnt.c: Include except.h.
4317 (struct seh_frame_state): Add reg_offset, after_prologue and
4318 in_cold_section fields.
4319 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4320 (i386_pe_seh_cold_init): New function.
4321 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4322 to seh->in_cold_section.
4323 (seh_emit_push): Record the offset of the push.
4324 (seh_emit_save): Record the offet of the save.
4325 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4326 Test seh->after_prologue to disregard the epilogue.
4327 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4328 (i386_pe_end_cold_function): New function.
4330 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4333 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4334 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4337 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4339 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4340 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4343 * multiple_target.c: Include tree-inline.h and intl.h.
4344 (expand_target_clones): Diagnose and fail if node->definition and
4345 !tree_versionable_function_p (node->decl).
4347 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4349 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4351 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4352 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4353 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4355 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4358 * combine.c (try_combine): Use reg_or_subregno instead of handling
4359 just paradoxical SUBREGs and REGs.
4361 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4363 * config/arc/arc.c (arc_finalize_pic): Remove function.
4364 (arc_must_save_register): We use single base PIC register, remove
4365 checks to save/restore the PIC register.
4366 (arc_expand_prologue): Likewise.
4367 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4369 (arc_verify_short): Likewise.
4370 (arc_attr_type): Likewise.
4371 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4372 (walk_stores): Likewise.
4373 (arc_address_cost): Make it static.
4374 (arc_verify_short): Likewise.
4375 (branch_dest): Likewise.
4376 (arc_attr_type): Likewise.
4377 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4378 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4379 (arc_final_prescan_insn): Remove inserting the nops due to
4380 hardware hazards. It is done in reorg step.
4381 (insn_length_variant_t): Remove.
4382 (insn_length_parameters_t): Likewise.
4383 (arc_insn_length_parameters): Likewise.
4384 (arc_get_insn_variants): Likewise.
4385 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4387 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4390 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4393 PR tree-optimization/84687
4394 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4396 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4398 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4400 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4401 Rename to ppc_speculation_barrier.
4402 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4403 __builtin_ppc_speculation_barrier.
4405 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4408 * combine.c (combine_simplify_rtx): Don't try to simplify if
4409 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4412 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4414 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4415 to 32 bytes when compiling for POWER9.
4417 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4420 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4421 regparm >= 3 with no arg reg available also for calls with
4422 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4425 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4427 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4429 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4432 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4434 2018-03-05 Richard Biener <rguenther@suse.de>
4436 PR tree-optimization/84486
4437 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4438 When inserting a __builtin_assume_aligned call set the LHS
4439 SSA name alignment info accordingly.
4441 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4443 PR tree-optimization/84114
4444 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4445 Avoid reassociation of FLOAT_MODE addition.
4447 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4449 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4450 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4451 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4452 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4453 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4454 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4455 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4457 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4458 __builtin_ia32_wbinvd): New builtins.
4459 (SPECIAL_ARGS2): New.
4460 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4461 (SPECIAL_ARGS2): New.
4462 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4463 (ix86_valid_target_attribute_inner_p): Ditto.
4464 (ix86_init_mmx_sse_builtins): Add special_args2.
4465 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4466 TARGET_WBNOINVD_P): New.
4467 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4468 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4469 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4470 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4471 * config/i386/pconfigintrin.h: New file.
4472 * config/i386/wbnoinvdintrin.h: Ditto.
4473 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4474 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4476 2018-03-05 Richard Biener <rguenther@suse.de>
4478 PR tree-optimization/84670
4479 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4481 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4482 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4483 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4484 doesn't grow if all successors (recursively) were visited at least
4487 2018-03-05 Richard Biener <rguenther@suse.de>
4489 PR tree-optimization/84650
4490 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4491 if executed in the loop pipeline.
4493 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4495 * doc/configfiles.texi (Configuration Files): Move info about
4496 conditionalizing $target-protos.h to...
4497 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4498 differs from $target-protos.h.
4500 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4501 Chung-Ju Wu <jasonwucj@gmail.com>
4503 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4504 * config/nds32/nds32-multiple.md (setmemsi): Define.
4505 * config/nds32/nds32-memory-manipulation.c
4506 (nds32_gen_dup_4_byte_to_word_value): New.
4507 (emit_setmem_word_loop): New.
4508 (emit_setmem_byte_loop): New.
4509 (nds32_expand_setmem_loop): New.
4510 (nds32_expand_setmem_loop_v3m): New.
4511 (nds32_expand_setmem_unroll): New.
4512 (nds32_expand_setmem): New.
4514 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4515 Chung-Ju Wu <jasonwucj@gmail.com>
4517 * config/nds32/nds32-memory-manipulation.c
4518 (nds32_emit_load_store): New.
4519 (nds32_emit_post_inc_load_store): New.
4520 (nds32_emit_mem_move): New.
4521 (nds32_emit_mem_move_block): New.
4522 (nds32_expand_movmemsi_loop_unknown_size): New.
4523 (nds32_expand_movmemsi_loop_known_size): New.
4524 (nds32_expand_movmemsi_loop): New.
4525 (nds32_expand_movmemsi_unroll): New.
4526 (nds32_expand_movmemqi): Rename ...
4527 (nds32_expand_movmemsi): ... to this.
4528 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4529 (movmemsi): ... to this.
4530 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4531 (nds32_expand_movmemsi): ... to this.
4533 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4534 Monk Chiang <sh.chiang04@gmail.com>
4535 Chung-Ju Wu <jasonwucj@gmail.com>
4537 * config/nds32/nds32-protos.h
4538 (nds32_expand_load_multiple): New arguments.
4539 (nds32_expand_store_multiple): Ditto.
4540 (nds32_valid_multiple_load_store): Rename ...
4541 (nds32_valid_multiple_load_store_p): ... to this.
4542 * config/nds32/nds32-memory-manipulation.c
4543 (nds32_expand_load_multiple): Refine implementation.
4544 (nds32_expand_store_multiple): Ditto.
4545 * config/nds32/nds32-multiple.md
4546 (load_multiple): Update nds32_expand_load_multiple interface.
4547 (store_multiple): Update nds32_expand_store_multiple interface.
4548 * config/nds32/nds32-predicates.c
4549 (nds32_valid_multiple_load_store): Rename ...
4550 (nds32_valid_multiple_load_store_p): ... to this and refine
4552 * config/nds32/predicates.md
4553 (nds32_load_multiple_and_update_address_operation): New predicate.
4554 (nds32_store_multiple_and_update_address_operation): New predicate.
4556 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4557 Chung-Ju Wu <jasonwucj@gmail.com>
4559 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4560 (combo): New attribute.
4561 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4563 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4565 * config/nds32/nds32.opt: Change -mcmodel= default value.
4567 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4568 Monk Chiang <sh.chiang04@gmail.com>
4569 Chung-Ju Wu <jasonwucj@gmail.com>
4571 * config/nds32/constants.md (unspec_element): New enum.
4572 * config/nds32/constraints.md (Umw): New constraint.
4573 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4574 * config/nds32/nds32-intrinsic.md: Likewise.
4575 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4576 (nds32_valid_smw_lwm_base_p): New.
4577 (nds32_output_smw_single_word): New.
4578 (nds32_output_lmw_single_word): New.
4579 (nds32_expand_unaligned_load): New.
4580 (nds32_expand_unaligned_store): New.
4581 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4582 (nds32_output_smw_single_word): Declare.
4583 (nds32_output_lmw_single_word): Declare.
4584 (nds32_expand_unaligned_load): Declare.
4585 (nds32_expand_unaligned_store): Declare.
4586 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4587 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4588 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4589 NDS32_BUILTIN_UASTORE_DW.
4590 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4593 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4594 Kito Cheng <kito.cheng@gmail.com>
4595 Chung-Ju Wu <jasonwucj@gmail.com>
4597 * config/nds32/nds32-intrinsic.c
4598 (nds32_expand_builtin_null_ftype_reg): Delete.
4599 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4600 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4601 (nds32_read_argument): New.
4602 (nds32_legitimize_target): Ditto.
4603 (nds32_legitimize_argument): Ditto.
4604 (nds32_check_constant_argument): Ditto.
4605 (nds32_expand_unop_builtin): Ditto.
4606 (nds32_expand_unopimm_builtin): Ditto.
4607 (nds32_expand_binop_builtin): Ditto.
4608 (nds32_builtin_decl_impl): Ditto.
4609 (builtin_description): Ditto.
4610 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4611 (nds32_init_builtins_impl): Ditto.
4612 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4613 (nds32_builtin_decl): New.
4614 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4615 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4617 2018-03-02 Jeff Law <law@redhat.com>
4619 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4620 (redundant_insn, fill_simple_delay_slots): Likewise.
4621 (fill_slots_from_thread): Likewise.
4622 * resource.c (mark_referenced_resources): Likewise.
4623 (mark_set_resources, find_dead_or_set_registers): Likewise.
4625 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4627 * substring-locations.h (format_warning_va): Formatting fix for
4629 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4631 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4632 * substring-locations.c: Include intl.h.
4633 (format_warning_va): Turned into small wrapper around
4634 format_warning_n_va, renamed to ...
4635 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4636 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4638 (format_warning_at_substring_n): New function.
4639 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4640 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4641 format_warning_at_substring with just a shorter name instead of
4642 const function pointer.
4643 (fmtwarn_n): New function.
4644 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4645 appropriate, get rid of all the fmtstr temporaries, move conditionals
4646 with G_() wrapped string literals directly into fmtwarn arguments,
4647 cast dir.len to (int), formatting fixes.
4649 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4651 * doc/invoke.texi: Remove "Cilk Plus" references.
4653 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4654 Richard Biener <rguenther@suse.de>
4657 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4658 for error or warning attributes if CALL_FROM_THUNK_P is set.
4661 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4664 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4665 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4668 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4669 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4671 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4672 instead of -1U in last predictors element's probability member.
4674 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4677 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4678 arguments if they are comparable.
4680 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4682 PR tree-optimization/84634
4683 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4684 masks and masked_loop_p with a single loop_masks, making sure it's
4685 null for bb vectorization.
4687 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4689 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4690 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4691 loop->force_vectorize to check whether there is no alias.
4693 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4696 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4698 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4700 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4701 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4702 instead of a loop around prev_real_insn.
4703 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4707 * config/i386/i386.c (ix86_print_operand): Use conditional
4708 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4711 2018-03-02 Richard Biener <rguenther@suse.de>
4713 PR tree-optimization/84427
4714 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4715 (bitmap_set_subtract_values): Rewrite to handle multiple
4718 (prune_clobbered_mems): Likewise.
4719 (phi_translate): Take edge instead of pred/phiblock.
4720 (phi_translate_1): Likewise.
4721 (phi_translate_set): Likewise. Insert all translated
4722 exprs for a value into the set, keeping possibly multiple
4723 expressions per value.
4724 (compute_antic_aux): Adjust for phi_translate changes.
4725 When intersecting union the expressions and prune those
4726 not in the final value set, keeping possibly multiple
4727 expressions per value. Do not use value-insertion
4728 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4729 all expressions. Add verification that the value-sets
4730 only shrink during iteration.
4731 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4732 (do_pre_regular_insertion): Likewise.
4733 (do_pre_partial_partial_insertion): Likewise.
4735 2018-03-02 Richard Biener <rguenther@suse.de>
4738 * config/darwin.c (saved_debug_info_level): New static global.
4739 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4740 (darwin_asm_lto_end): Restore debug info generation settings.
4742 2018-03-01 Martin Liska <mliska@suse.cz>
4745 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4748 2018-03-01 Richard Biener <rguenther@suse.de>
4751 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4752 type annotation with LTO when debug was disabled at compile-time.
4754 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4756 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4758 (mips_final_postscan_insn): Likewise.
4760 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4762 PR rtl-optimization/84528
4763 * alias.c (init_alias_target): Add commentary.
4764 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4765 a unique base value if the frame pointer is not eliminated
4766 to the stack pointer.
4768 2018-03-01 Tom de Vries <tom@codesourcery.com>
4770 PR rtl-optimization/83327
4771 * lra-int.h (hard_regs_spilled_into): Declare.
4772 * lra.c (hard_regs_spilled_into): Define.
4773 (init_reg_info): Init hard_regs_spilled_into.
4774 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4775 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4776 (process_bb_lives): Handle hard_regs_spilled_into.
4777 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4778 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4780 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4782 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4783 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4784 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4785 * config/rs6000/aix72.h: New file.
4787 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4789 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4790 instead of warning_at with conditional singular and plural messages
4794 * stor-layout.c (update_alignment_for_field): For
4795 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4796 && !DECL_PACKED (field), do the alignment update, just use
4797 only desired_align instead of MAX (type_align, desired_align)
4799 (place_field): Don't do known_align < desired_align handling
4800 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4801 is non-NULL, instead do it after rli->prev_field handling and
4802 only if not within a bitfield word. For DECL_PACKED (field)
4803 use type_align of BITS_PER_UNIT.
4805 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4807 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4808 superfluous parentheses and trailing spaces.
4810 2018-02-28 Richard Biener <rguenther@suse.de>
4812 PR tree-optimization/84584
4813 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4814 SCoPs with fake exit edge.
4816 2018-02-28 Martin Liska <mliska@suse.cz>
4819 * timevar.c (timer::print): Fix format to properly print 100%
4822 2018-02-28 Richard Biener <rguenther@suse.de>
4825 * genmatch.c (capture_info::walk_match): Do not mark
4826 captured expressions without operands as expr_p given
4827 they act more like predicates and should be subject to
4828 "lost tail" side-effect preserving.
4830 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4832 PR rtl-optimization/81611
4833 * auto-inc-dec.c (attempt_change): Move dead note from
4834 mem_insn if it's the next use of regno
4835 (find_address): Take address use of reg holding
4836 non-incremented value. Add parm to limit search to the named
4838 (merge_in_block): Attempt to use a mem insn that is the next
4839 use of the original regno.
4841 2018-02-27 Martin Sebor <msebor@redhat.com>
4844 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4845 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4847 2018-02-27 Martin Sebor <msebor@redhat.com>
4849 PR translation/84207
4850 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4851 n argument to unsigned HOST_WIDE_INT.
4852 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4853 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4854 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4855 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4857 2018-02-27 Richard Biener <rguenther@suse.de>
4859 PR tree-optimization/84512
4860 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4861 Do not use the estimate returned from record_stmt_cost for
4862 the scalar iteration cost but sum properly using add_stmt_cost.
4864 2018-02-27 Richard Biener <rguenther@suse.de>
4866 PR tree-optimization/84466
4867 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4868 Adjust last change to less strictly validate use operands.
4870 2018-02-27 Martin Liska <mliska@suse.cz>
4872 PR gcov-profile/84548
4873 * gcov.c (process_file): Allow partial overlap and consider it
4874 also as group functions.
4875 (output_lines): Properly calculate range of lines for a group.
4877 2018-02-27 Martin Liska <mliska@suse.cz>
4879 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4880 'ggc' suffixes. Change first column width.
4881 (timer::print): Fix formatting of the column.
4883 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4885 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4886 preserve inline entry blocks for the sake of debug inline
4887 entry point markers alone.
4888 (remove_unused_locals): Suggest in comments a better place to
4889 force the preservation of inline entry blocks that are
4890 otherwise unused, but do not preserve them.
4892 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4894 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4896 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4899 * config/i386/constraints.md (Bs): Replace
4900 ix86_indirect_branch_register with
4901 TARGET_INDIRECT_BRANCH_REGISTER.
4903 * config/i386/i386.md (indirect_jump): Likewise.
4904 (tablejump): Likewise.
4905 (*sibcall_memory): Likewise.
4906 (*sibcall_value_memory): Likewise.
4907 Peepholes of indirect call and jump via memory: Likewise.
4908 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4909 (*sibcall_value_GOT_32): Likewise.
4910 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4911 (GOT_memory_operand): Likewise.
4912 (call_insn_operand): Likewise.
4913 (sibcall_insn_operand): Likewise.
4914 (GOT32_symbol_operand): Likewise.
4915 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4917 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4919 PR rtl-optimization/83496
4920 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4921 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4922 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4923 redundant insn, if any.
4924 (relax_delay_slots): Likewise.
4925 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4927 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4929 PR tree-optimization/83965
4930 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4931 that grouped statements are part of a reduction chain. Return
4932 true if the statement is not marked as a reduction itself but
4934 (vect_recog_dot_prod_pattern): Don't check whether the statement
4935 is part of a group here.
4936 (vect_recog_sad_pattern): Likewise.
4937 (vect_recog_widen_sum_pattern): Likewise.
4939 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4942 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4944 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4947 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4949 (ix86_output_indirect_function_return): New prototype.
4950 (ix86_split_simple_return_pop_internal): Likewise.
4951 * config/i386/i386.c (indirect_return_via_cx): New.
4952 (indirect_return_via_cx_bnd): Likewise.
4953 (indirect_thunk_name): Handle return va CX_REG.
4954 (output_indirect_thunk_function): Create alias for
4955 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4956 (ix86_output_indirect_jmp): Remove the bool argument.
4957 (ix86_output_indirect_function_return): New function.
4958 (ix86_split_simple_return_pop_internal): Likewise.
4959 * config/i386/i386.md (*indirect_jump): Don't pass false
4960 to ix86_output_indirect_jmp.
4961 (*tablejump_1): Likewise.
4962 (simple_return_pop_internal): Change it to define_insn_and_split.
4963 Call ix86_split_simple_return_pop_internal to split it for
4965 (simple_return_indirect_internal): Call
4966 ix86_output_indirect_function_return instead of
4967 ix86_output_indirect_jmp.
4969 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4972 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4973 memset and value initialization afterwards.
4975 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4977 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4979 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4982 * common/config/aarch64/aarch64-common.c
4983 (aarch_option_optimization_table[]): Switch
4984 off fomit-frame-pointer
4986 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4987 Chung-Ju Wu <jasonwucj@gmail.com>
4989 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4991 (store_multiple): Ditto.
4993 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4995 * config.gcc: Add --with-cpu support for nds32 target.
4996 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4997 * config/nds32/nds32.opt: Add -mcpu= option.
4999 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5001 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5002 isel=yes): Warn for these deprecated options.
5004 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5006 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5007 ISA_2_5_MASKS_EMBEDDED.
5009 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5011 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5012 p->max as pointers rather than using iterative_hash_expr.
5014 2018-02-23 Carl Love <cel@us.ibm.com>
5016 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5017 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5019 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5020 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5021 P8V_BUILTIN_VEC_VUNSIGNED2.
5023 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5026 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5027 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5029 * lra-constraints.c (curr_insn_transform): Set up
5030 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5032 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5033 * lra-eliminations.c (spill_pseudos): Ditto.
5034 (process_insn_for_elimination): Ditto.
5035 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5036 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5037 LRA_NON_CLOBBERED_ALT.
5039 2018-02-22 Martin Sebor <msebor@redhat.com>
5041 PR tree-optimization/84480
5042 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5043 to maybe_diag_stxncpy_trunc. Call it.
5044 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5045 from gimple_fold_builtin_strcpy. Print inlining stack.
5046 (handle_builtin_stxncpy): Print inlining stack.
5047 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5049 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5052 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5053 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5054 and -fcheck-pointer-bounds are used together.
5055 (indirect_thunk_prefix): New enum.
5056 (indirect_thunk_need_prefix): New function.
5057 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5058 "_nt" instead of "_bnd" for NOTRACK prefix.
5059 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5060 (output_indirect_thunk_function): Likewise.
5062 (ix86_code_end): Update output_indirect_thunk_function calls.
5063 (ix86_output_indirect_branch_via_reg): Replace
5064 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5065 (ix86_output_indirect_branch_via_push): Likewise.
5066 (ix86_output_function_return): Likewise.
5067 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5068 incompatible with -fcf-protection=branch and
5069 -fcheck-pointer-bounds.
5071 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5074 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5075 Change gcc_assert call to output_operand_lossage.
5077 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5079 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5081 2018-02-22 DJ Delorie <dj@redhat.com>
5082 Sebastian Perta <sebastian.perta@renesas.com>
5083 Oleg Endo <olegendo@gcc.gnu.org>
5085 * config/rx/rx.c (rx_rtx_costs): New function.
5086 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5088 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5090 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5092 2018-02-22 Martin Liska <mliska@suse.cz>
5095 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5096 Add "native" as a possible value.
5098 2018-02-22 Martin Liska <mliska@suse.cz>
5101 * config/i386/i386.c (ix86_option_override_internal):
5102 Add "native" as a possible value for -march and -mtune.
5104 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5107 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5108 to all type variants.
5110 PR tree-optimization/84503
5111 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5112 width as info->bitpos + info->bitsize - start.
5113 (merged_store_group::merge_overlapping): Simplify width computation.
5114 (check_no_overlap): New function.
5115 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5116 start + width and last_order of the group, fail if check_no_overlap
5118 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5119 to group if check_no_overlap fails.
5121 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5123 * config/rs6000/altivec.md: Delete contraint arguments to
5124 define_expand, define_split, and define_peephole2, and in
5125 define_insn_and_split if always unused.
5126 * config/rs6000/darwin.md: Ditto.
5127 * config/rs6000/dfp.md: Ditto.
5128 * config/rs6000/rs6000.md: Ditto.
5129 * config/rs6000/sync.md: Ditto.
5130 * config/rs6000/vector.md: Ditto.
5131 * config/rs6000/vsx.md: Ditto.
5133 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5135 * config/rs6000/altivec.md: Write output control strings as braced
5136 blocks instead of double-quoted strings.
5137 * config/rs6000/darwin.md: Ditto.
5138 * config/rs6000/rs6000.md: Ditto.
5139 * config/rs6000/vector.md: Ditto.
5140 * config/rs6000/vsx.md: Ditto.
5142 2018-02-21 Jason Merrill <jason@redhat.com>
5144 PR c++/84314 - ICE with templates and fastcall attribute.
5145 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5147 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5149 * ipa-cp.c (determine_versionability): Fix comment typos.
5151 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5154 * ipa-cp.c (determine_versionability): Do not version functions caling
5157 2018-02-21 Martin Liska <mliska@suse.cz>
5160 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5161 Add "native" as a possible value.
5162 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5163 the macro when native cpu detection is available.
5165 2018-02-21 Martin Liska <mliska@suse.cz>
5168 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5169 Add "native" as a possible value.
5170 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5171 when native cpu detection is available.
5173 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5174 Martin Sebor <msebor@redhat.com>
5176 PR tree-optimization/84478
5177 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5179 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5180 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5181 support which is conservatively correct, for 2 only stay conservative
5182 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5183 argument to the 2 argument get_range_strlen, adjust 6 arg
5184 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5186 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5187 (gimple_fold_builtin_strlen): Pass true as last argument to
5190 2018-02-20 Martin Sebor <msebor@redhat.com>
5193 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5194 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5195 (builtin_memref::builtin_memref): Factor out parts into
5196 set_base_and_offset and call it.
5198 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5201 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5202 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5203 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5204 search at the associated MODE_INT.
5206 2018-02-20 Jeff Law <law@redhat.com>
5209 PR tree-optimization/81592
5211 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5212 for range data rather than using global data.
5213 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5214 range data rather than using global data.
5215 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5216 pass it to children as needed.
5217 (struct directive::fmtresult): Similarly.
5218 (struct directive::set_width): Similarly.
5219 (struct directive::set_precision): Similarly.
5220 (format_integer, format_directive, parse_directive): Similarly.
5221 (format_none): Accept unnamed vr_values parameter.
5222 (format_percent, format_floating, format_character): Similarly.
5223 (format_string, format_plain): Similarly.
5224 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5225 the EVRP range analyzer for range data rather than using global data.
5226 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5227 gimple-ssa-evrp-analyze.h
5228 (class sprintf_dom_walker): Add after_dom_children member function.
5229 Add evrp_range_analyzer member.
5230 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5231 range analyzer as needed.
5232 (sprintf_dom_walker::after_dom_children): New member function.
5233 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5235 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5236 (evrp_range_analyzer::pop_to_marker): Likewise.
5238 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5240 PR tree-optimization/84419
5241 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5242 with the required type if its current type is compatible but
5245 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5248 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5249 after vectorization.
5251 2018-02-20 Martin Liska <mliska@suse.cz>
5254 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5255 possible values if we don't have a hint.
5257 2018-02-20 Martin Liska <mliska@suse.cz>
5261 * final.c (shorten_branches): Build align_tab array with one
5263 * opts.c (finish_options): Add alignment option limit check.
5264 (MAX_CODE_ALIGN): Likewise.
5265 (MAX_CODE_ALIGN_VALUE): Likewise.
5266 * doc/invoke.texi: Document maximum allowed option value for
5267 all -falign-* options.
5269 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5272 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5273 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5274 * var-tracking.c (emit_note_insn_var_location): Remove all references
5275 to NOTE_INSN_CALL_ARG_LOCATION.
5276 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5277 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5278 Use copy_rtx_if_shared.
5279 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5280 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5281 (dwarf2out_var_location): Remove handling of
5282 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5284 * final.c (final_scan_insn): Remove all references to
5285 NOTE_INSN_CALL_ARG_LOCATION.
5286 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5287 before dumping final insns.
5288 * except.c (emit_note_eh_region_end): Remove all references to
5289 NOTE_INSN_CALL_ARG_LOCATION.
5290 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5291 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5292 * config/arc/arc.c (hwloop_optimize): Likewise.
5293 * config/arm/arm.c (create_fix_barrier): Likewise.
5294 * config/s390/s390.c (s390_chunkify_start): Likewise.
5295 * config/sh/sh.c (find_barrier): Likewise.
5296 * config/i386/i386.c (rest_of_insert_endbranch,
5297 ix86_seh_fixup_eh_fallthru): Likewise.
5298 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5299 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5300 * config/frv/frv.c (frv_function_prologue): Likewise.
5301 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5303 (note_outside_basic_block_p): Remove all references to
5304 NOTE_INSN_CALL_ARG_LOCATION.
5305 * gengtype.c (adjust_field_rtx_def): Likewise.
5306 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5308 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5309 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5312 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5315 PR tree-optimization/84452
5316 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5317 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5320 2018-02-19 Martin Liska <mliska@suse.cz>
5323 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5325 2018-02-19 Martin Liska <mliska@suse.cz>
5326 Richard Sandiford <richard.sandiford@linaro.org>
5328 PR tree-optimization/82491
5329 * gimple-fold.c (get_base_constructor): Make earlier bail out
5332 2018-02-19 Carl Love <cel@us.ibm.com>
5334 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5335 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5337 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5338 P8V_BUILTIN_VEC_NEG.
5340 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5342 * config/rl78/rl78.md (movdf): New define expand.
5344 2018-02-19 Martin Liska <mliska@suse.cz>
5347 * doc/invoke.texi: Fix typo.
5348 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5350 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5352 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5353 handle rs6000_single_float and rs6000_double_float specially for
5356 2018-02-16 Jeff Law <law@redhat.com>
5358 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5361 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5364 * ipa-inline.c (inline_small_functions): Fix a typo.
5366 2018-02-16 Nathan Sidwell <nathan@acm.org>
5368 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5370 2018-02-16 Carl Love <cel@us.ibm.com>
5372 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5373 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5374 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5375 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5376 expansion to P8V_BUILTIN_VEC_FLOAT2.
5378 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5380 PR rtl-optimization/70023
5381 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5382 src_regno into account.
5384 2018-02-16 Carl Love <cel@us.ibm.com>
5386 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5387 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5388 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5389 * config/rs6000/rs6000.c: Remove case statements for
5390 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5391 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5392 and P9V_BUILTIN_VEC_VINSERT4B.
5393 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5394 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5395 * config/rs6000/vsx.md:
5396 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5399 2018-02-16 Carl Love <cel@us.ibm.com>
5401 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5403 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5405 * config/rs6000/rs6000-c.c: Add the definitions for
5406 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5407 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5408 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5409 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5410 definition for insert4b and define insn *insert3b_internal.
5411 * doc/extend.texi: Add documentation for vec_extract4b.
5413 2018-02-16 Nathan Sidwell <nathan@acm.org>
5415 * doc/extend.texi (Backwards Compatibility): Mention friend
5416 injection. Note for-scope is deprecated.
5417 * doc/invoke.texi (-ffriend-injection): Deprecate.
5419 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5421 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5422 that moved to I2, also allow destinations that are a paradoxical
5423 subreg (instead of a normal reg).
5425 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5428 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5431 2018-02-16 Richard Biener <rguenther@suse.de>
5433 PR tree-optimization/84037
5434 PR tree-optimization/84016
5436 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5437 Adjust vec_construct for the fact we need additional higher latency
5438 128bit inserts for AVX256 and AVX512 vector builds.
5439 (ix86_add_stmt_cost): Scale vector construction cost for
5442 2018-02-16 Richard Biener <rguenther@suse.de>
5444 PR tree-optimization/84417
5445 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5446 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5447 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5449 2018-02-16 Martin Liska <mliska@suse.cz>
5452 * internal-fn.def (ASAN_CHECK): Set proper flags.
5453 (ASAN_MARK): Likewise.
5455 2018-02-16 Julia Koval <julia.koval@intel.com>
5457 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5458 from PTA_CANNONLAKE.
5460 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5463 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5464 Use ++iter rather than iter++ for std::list iterators.
5465 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5466 defer deleting them until all nodes in the forest are processed. Do
5467 free even leaf nodes. Change to_process into auto_vec.
5470 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5471 * vec.h (vec_default_construct): Use memset instead of placement new
5472 if BROKEN_VALUE_INITIALIZATION is defined.
5473 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5474 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5477 PR rtl-optimization/83723
5478 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5479 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5480 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5482 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5484 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5486 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5488 PR rtl-optimization/81443
5489 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5490 from inner REGs to paradoxical SUBREGs.
5492 2018-02-16 Richard Biener <rguenther@suse.de>
5494 PR tree-optimization/84399
5495 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5496 For operands we can analyze at their definition make sure we can
5497 analyze them at each use as well.
5499 2018-02-16 Richard Biener <rguenther@suse.de>
5501 PR tree-optimization/84190
5502 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5503 volatile accesses if the decl isn't volatile.
5505 2018-02-15 Jason Merrill <jason@redhat.com>
5507 PR c++/84314 - ICE with templates and fastcall attribute.
5508 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5509 TYPE_CANONICAL on an existing type.
5511 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5513 PR tree-optimization/84383
5514 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5515 dstoff nor call operand_equal_p if dstbase is NULL.
5517 PR tree-optimization/84334
5518 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5519 also a CONSTANT_CLASS_P, punt.
5521 2018-02-14 Jim Wilson <jimw@sifive.com>
5523 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5524 first SMALL_OPERAND check. New local min_second_step. Move assert
5525 to where locals are set. Add TARGET_RVC support.
5526 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5528 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5530 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5532 2018-02-14 Martin Sebor <msebor@redhat.com>
5534 PR tree-optimization/83698
5535 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5536 arrays constrain the offset range to their bounds.
5537 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5538 (builtin_access::overlap): Avoid setting the size of overlap if it's
5540 (maybe_diag_overlap): Also consider arrays when deciding what values
5541 of offsets to include in diagnostics.
5543 2018-02-14 Martin Sebor <msebor@redhat.com>
5546 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5547 that correspond to the kind of a declaration.
5549 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5552 * config/pa/pa.md: Load address of PIC label using the linkage table
5553 if the label is nonlocal.
5555 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5557 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5558 warning message if user requests -maltivec=be.
5559 * doc/invoke.texi: Document deprecation of -maltivec=be.
5561 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5564 * config/rs6000/rs6000-c.c: Update definitions for
5565 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5566 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5568 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5571 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5572 add _get_ssp intrinsics. Remove argument from
5573 __builtin_ia32_rdssp[d|q].
5574 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5575 * config/i386/i386-builtin.def: Remove argument from
5576 __builtin_ia32_rdssp[d|q].
5577 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5578 ix86_expand_special_args_builtin for _rdssp[d|q].
5579 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5580 Clear register before usage.
5581 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5582 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5584 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5586 PR tree-optimization/84357
5587 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5588 operand 1 of an ARRAY_REF too.
5590 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5593 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5594 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5596 (set_of_reg): New struct.
5597 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5598 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5599 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5601 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5602 Split into bitclr, bitset, bitinvert patterns if appropriate.
5603 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5604 use rx_fuse_in_memory_bitop.
5605 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5606 to named insn, correct maximum insn length.
5608 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5611 * machmode.def: Define a complex mode for PARTIAL_INT.
5612 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5614 * doc/rtl.texi: Document CSPImode.
5615 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5617 (msp430_hard_regno_nregs_with_padding): Likewise.
5619 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5622 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5624 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5626 PR rtl-optimization/84169
5627 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5628 we generated a parallel as new i3 and we split that to new i2 and i3
5629 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5630 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5631 those to i2, not i1. Partially rewrite this scan code.
5633 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5636 * stor-layout.c (place_field): For variable length fields, adjust
5637 offset_align afterwards not just based on the field's alignment,
5638 but also on the size.
5641 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5642 of exps and logs in the use_exp2 case.
5644 2018-02-13 Jeff Law <law@redhat.com>
5646 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5649 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5652 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5656 * common.opt (gas-loc-support, gas-locview-support): New.
5657 (ginline-points, ginternal-reset-location-views): New.
5658 * doc/invoke.texi: Document them. Use @itemx where intended.
5659 (gvariable-location-views): Adjust.
5660 * target.def (reset_location_view): New.
5661 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5662 (TARGET_RESET_LOCATION_VIEW): New.
5663 * doc/tm.texi: Rebuilt.
5664 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5665 (dwarf2out_default_as_locview_support): New.
5666 (output_asm_line_debug_info): Use option variables.
5667 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5668 (output_loc_list): Likewise.
5669 (add_high_low_attributes): Check option variables.
5670 Don't output entry view attribute in strict mode.
5671 (gen_inlined_subroutine_die): Check option variables.
5672 (dwarf2out_inline_entry): Likewise.
5673 (init_sections_and_labels): Likewise.
5674 (dwarf2out_early_finish): Likewise.
5675 (maybe_reset_location_view): New, from...
5676 (dwarf2out_var_location): ... here. Call it.
5677 * debug.h (dwarf2out_default_as_loc_support): Declare.
5678 (dwarf2out_default_as_locview_support): Declare.
5679 * hooks.c (hook_int_rtx_insn_0): New.
5680 * hooks.h (hook_int_rtx_insn_0): Declare.
5681 * toplev.c (process_options): Take -gas-loc-support and
5682 -gas-locview-support from dwarf2out. Enable
5683 -gvariable-location-views by default only with locview
5684 assembler support. Enable -ginternal-reset-location-views by
5685 default only if the target defines the corresponding hook.
5686 Enable -ginline-points by default if location views are
5687 enabled; force it disabled if statement frontiers are
5689 * tree-inline.c (expand_call_inline): Check option variables.
5690 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5692 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5694 PR tree-optimization/84321
5695 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5696 handling. Also check whether the anti-range contains any values
5697 that satisfy the mask; switch to a VR_RANGE if not.
5699 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5702 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5704 2018-02-13 Martin Jambor <mjambor@suse.cz>
5707 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5708 of call statements, also set location of a load to a temporary.
5710 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5712 * config/rl78/rl78.c (add_vector_labels): New function.
5713 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5714 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5715 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5716 which checks that no arguments are passed.
5717 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5718 * doc/extend.texi: Documentation for the new attribute.
5720 2018-02-13 Andreas Schwab <schwab@suse.de>
5722 * config/riscv/linux.h (CPP_SPEC): Define.
5724 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5727 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5728 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5729 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5730 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5731 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5732 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5733 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5734 temporarily for AES and PCLMUL builtins.
5736 PR tree-optimization/84339
5737 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5738 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5742 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5743 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5744 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5746 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5748 * omp-simd-clone.h: New file.
5749 * omp-simd-clone.c: Include omp-simd-clone.h.
5750 (expand_simd_clones): No longer static.
5751 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5752 cgraph.h and omp-simd-clone.h.
5753 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5754 (vect_recog_widen_shift_pattern): Formatting fix.
5755 (vect_pattern_recog_1): Don't check optab for calls.
5758 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5759 operands[2] into a REG before using gen_lowpart on it.
5761 2018-02-12 Jeff Law <law@redhat.com>
5764 * config/sh/sh.c (find_barrier): Consider a sibling call
5767 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5768 successfully back substituting a reg.
5770 2018-02-12 Richard Biener <rguenther@suse.de>
5772 PR tree-optimization/84037
5773 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5774 parameter, move visited init to caller.
5775 (vect_slp_analyze_operations): Separate cost from validity
5776 check, initialize visited once for all instances.
5777 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5779 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5781 (vect_model_promotion_demotion_cost): Likewise.
5782 (vectorizable_bswap): Guard cost modeling with !slp_node
5783 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5785 (vectorizable_call): Likewise.
5786 (vectorizable_conversion): Likewise.
5787 (vectorizable_assignment): Likewise.
5788 (vectorizable_shift): Likewise.
5789 (vectorizable_operation): Likewise.
5790 (vectorizable_store): Likewise.
5791 (vectorizable_load): Likewise.
5792 (vectorizable_condition): Likewise.
5793 (vectorizable_comparison): Likewise.
5795 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5798 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5799 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5800 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5802 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5805 * params.def (inline-min-speedup): Increase from 8 to 15.
5806 (max-inline-insns-auto): Decrease from 40 to 30.
5807 * ipa-split.c (consider_split): Add some buffer for function to
5808 be considered inlining candidate.
5809 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5812 2018-02-12 Richard Biener <rguenther@suse.de>
5814 PR tree-optimization/84037
5815 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5816 matched stmts if we cannot swap the non-matched ones.
5818 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5820 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5821 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5822 _mm_maskz_scalef_round_ss): New intrinsics.
5823 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5824 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5825 __builtin_ia32_scalefss_round): Remove.
5826 (__builtin_ia32_scalefsd_mask_round,
5827 __builtin_ia32_scalefss_mask_round): New intrinsics.
5828 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5829 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5830 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5831 "<round_constraint>")): Changed to ...
5832 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5833 "<round_scalar_constraint>")): ... this.
5834 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5835 %0, %1, %2<round_op3>}"): Changed to ...
5836 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5837 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5838 %2<round_scalar_mask_op3>}"): ... this.
5839 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5841 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5843 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5844 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5845 (_mm_maskz_sqrt_round_ss): New intrinsics.
5846 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5847 (__builtin_ia32_sqrtsd_mask_round)
5848 (__builtin_ia32_sqrtss_mask_round): New builtins.
5849 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5850 (__builtin_ia32_sqrtss_round): Remove.
5851 (__builtin_ia32_sqrtsd_mask_round)
5852 (__builtin_ia32_sqrtss_mask_round): New builtins.
5853 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5854 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5855 ((match_operand:VF_128 1 "vector_operand"
5856 "xBm,<round_constraint>")): Changed to ...
5857 ((match_operand:VF_128 1 "vector_operand"
5858 "xBm,<round_scalar_constraint>")): ... this.
5859 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5860 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5861 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5862 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5863 %<iptr>1<round_scalar_mask_op3>}): ... this.
5864 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5865 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5867 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5870 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5871 Cast vec_cmpeq result to correct type.
5872 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5873 Cast vec_cmpgt result to correct type.
5875 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5877 * final.c (final_scan_insn_1): Renamed from...
5878 (final_scan_insn): ... this. New wrapper, to recover
5879 seen from the outermost call in recursive ones.
5880 * config/sparc/sparc.c (output_return): Drop seen from call.
5881 (output_sibcall): Likewise.
5882 * config/visium/visium.c (output_branch): Likewise.
5884 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5886 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5889 2018-02-10 Alan Modra <amodra@gmail.com>
5892 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5893 Specify LR as an input.
5895 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5898 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5899 remove_member_access_dummy_vars): New functions.
5900 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5901 lower_omp_1, execute_lower_omp): Use them.
5903 PR rtl-optimization/84308
5904 * shrink-wrap.c (spread_components): Release todo vector.
5906 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5908 PR rtl-optimization/57193
5909 * ira-color.c (struct allocno_color_data): Add member
5910 conflict_allocno_hard_prefs.
5911 (update_conflict_allocno_hard_prefs): New.
5912 (bucket_allocno_compare_func): Add a preference based on
5913 conflict_allocno_hard_prefs.
5914 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5915 (color_allocnos): Remove a dead code. Initiate
5916 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5918 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5921 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5922 constraint from =wa to wa. Avoid a subreg on the output operand,
5923 instead use a pseudo and subreg it in a move.
5924 (p9_xxbrd_<mode>): Changed to ...
5925 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5926 (p9_xxbrd_v2df): New expander.
5927 (p9_xxbrw_<mode>): Changed to ...
5928 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5929 (p9_xxbrw_v4sf): New expander.
5931 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5933 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5935 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5938 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5939 multiply in 32-bit mode.
5940 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5941 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5944 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5946 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5947 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5948 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5949 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5951 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5954 * dwarf2out.c (is_trivial_indirect_ref): New function.
5955 (dwarf2out_late_global_decl): Do not generate a location
5956 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5957 and that are not defined in the current unit.
5959 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5961 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5962 instead of a libcall for UNORDERED.
5964 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5967 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5968 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5970 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5973 * config/s390/s390.c (s390_set_current_function): Invoke
5974 s390_indirect_branch_settings also if fndecl didn't change.
5976 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5978 * config/rs6000/rs6000.md (blockage): Set length to zero.
5980 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5982 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5984 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5987 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5988 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5992 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5993 PARALLEL incoming that failed vt_get_decl_and_offset check.
5996 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5997 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5999 (get_variable_section): For decls in named .bss* sections pass true as
6000 second argument to bss_initializer_p.
6002 2018-02-09 Marek Polacek <polacek@redhat.com>
6003 Jakub Jelinek <jakub@redhat.com>
6006 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6007 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6008 Sync some changes from cxx_fold_indirect_ref.
6010 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6012 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6014 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6015 (BLOCK_INLINE_ENTRY_LABEL): New.
6016 (dwarf2out_var_location): Disregard inline entry markers.
6017 (inline_entry_data): New struct.
6018 (inline_entry_data_hasher): New hashtable type.
6019 (inline_entry_data_hasher::hash): New.
6020 (inline_entry_data_hasher::equal): New.
6021 (inline_entry_data_table): New variable.
6022 (add_high_low_attributes): Add DW_AT_entry_pc and
6023 DW_AT_GNU_entry_view attributes if a pending entry is found
6024 in inline_entry_data_table. Add old entry_pc attribute only
6025 if debug nonbinding markers are disabled.
6026 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6027 markers are enabled.
6028 (block_within_block_p, dwarf2out_inline_entry): New.
6029 (dwarf2out_finish): Check that no entries remained in
6030 inline_entry_data_table.
6031 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6032 (final_scan_insn, notice_source_line): Likewise.
6033 (rest_of_clean_state): Skip inline entry markers.
6034 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6036 * gimple.c (gimple_build_debug_inline_entry): New.
6037 * gimple.h (enum gimple_debug_subcode): Add
6038 GIMPLE_DEBUG_INLINE_ENTRY.
6039 (gimple_build_debug_inline_entry): Declare.
6040 (gimple_debug_inline_entry_p): New.
6041 (gimple_debug_nonbind_marker_p): Adjust.
6042 * insn-notes.def (INLINE_ENTRY): New.
6043 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6044 inline entry marker notes.
6045 (print_insn): Likewise.
6046 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6047 (INSN_DEBUG_MARKER_KIND): Likewise.
6048 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6049 * tree-inline.c (expand_call_inline): Build and insert
6050 debug_inline_entry stmt.
6051 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6052 inline entry blocks early, if nonbind markers are enabled.
6053 (dump_scope_block): Dump fragment info.
6054 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6055 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6056 (gimple_build_debug_inline_entry): New.
6057 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6058 Enable/disable inline entry points too.
6059 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6060 (DEBUG_INSN): Describe inline entry markers.
6062 * common.opt (gvariable-location-views): New.
6063 (gvariable-location-views=incompat5): New.
6064 * config.in: Rebuilt.
6065 * configure: Rebuilt.
6066 * configure.ac: Test assembler for view support.
6067 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6068 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6069 * dwarf2out.c (var_loc_view): New typedef.
6070 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6071 (dwarf2out_locviews_in_attribute): New.
6072 (dwarf2out_locviews_in_loclist): New.
6073 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6074 (enum dw_line_info_opcode): Add LI_adv_address.
6075 (struct dw_line_info_table): Add view.
6076 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6077 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6078 (zero_view_p): New variable.
6079 (ZERO_VIEW_P): New macro.
6080 (output_asm_line_debug_info): New.
6081 (struct var_loc_node): Add view.
6082 (add_AT_view_list, AT_loc_list): New.
6083 (add_var_loc_to_decl): Add view param. Test it against last.
6084 (new_loc_list): Add view params. Record them.
6085 (AT_loc_list_ptr): Handle loc and view lists.
6086 (view_list_to_loc_list_val_node): New.
6087 (print_dw_val): Handle dw_val_class_view_list.
6088 (size_of_die): Likewise.
6089 (value_format): Likewise.
6090 (loc_list_has_views): New.
6091 (gen_llsym): Set vl_symbol too.
6092 (maybe_gen_llsym, skip_loc_list_entry): New.
6093 (dwarf2out_maybe_output_loclist_view_pair): New.
6094 (output_loc_list): Output view list or entries too.
6095 (output_view_list_offset): New.
6096 (output_die): Handle dw_val_class_view_list.
6097 (output_dwarf_version): New.
6098 (output_compilation_unit_header): Use it.
6099 (output_skeleton_debug_sections): Likewise.
6100 (output_rnglists, output_line_info): Likewise.
6101 (output_pubnames, output_aranges): Update version comments.
6102 (output_one_line_info_table): Output view numbers in asm comments.
6103 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6104 Call maybe_gen_llsym.
6105 (loc_list_from_tree_1): Adjust.
6106 (add_AT_location_description): Create view list attribute if
6107 needed, check it's absent otherwise.
6108 (convert_cfa_to_fb_loc_list): Adjust.
6109 (maybe_emit_file): Call output_asm_line_debug_info for test.
6110 (dwarf2out_var_location): Reset views as needed. Precompute
6111 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6112 attribute. Set view.
6113 (new_line_info_table): Reset next view.
6114 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6115 (dwarf2out_source_line): Likewise. Output view resets and labels to
6116 the assembler, or select appropriate line info opcodes.
6117 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6118 (optimize_string_length): Catch it. Adjust.
6119 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6120 dw_val_class_view_list, and remove it if no longer needed.
6121 (hash_loc_list): Hash view numbers.
6122 (loc_list_hasher::equal): Compare them.
6123 (optimize_location_lists): Check whether a view list symbol is
6124 needed, and whether the locview attribute is present, and
6125 whether they match. Remove the locview attribute if no longer
6127 (index_location_lists): Call skip_loc_list_entry for test.
6128 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6129 Use output_dwarf_version.
6130 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6131 (struct dw_val_node): Add val_view_list.
6132 * final.c (SEEN_NEXT_VIEW): New.
6133 (set_next_view_needed): New.
6134 (clear_next_view_needed): New.
6135 (maybe_output_next_view): New.
6136 (final_start_function): Rename to...
6137 (final_start_function_1): ... this. Take pointer to FIRST,
6138 add SEEN parameter. Emit param bindings in the initial view.
6139 (final_start_function): Reintroduce SEEN-less interface.
6140 (final): Rename to...
6141 (final_1): ... this. Take SEEN parameter. Output final pending
6142 next view at the end.
6143 (final): Reintroduce seen-less interface.
6144 (final_scan_insn): Output pending next view before switching
6145 sections or ending a block. Mark the next view as needed when
6146 outputting variable locations. Notify debug backend of section
6147 changes, and of location view changes.
6148 (rest_of_handle_final): Adjust.
6149 * toplev.c (process_options): Autodetect value for debug variable
6150 location views option. Warn on incompat5 without -gdwarf-5.
6151 * doc/invoke.texi (gvariable-location-views): New.
6152 (gvariable-location-views=incompat5): New.
6153 (gno-variable-location-views): New.
6155 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6157 PR tree-optimization/84136
6158 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6159 that the result of find_edge is non-NULL.
6161 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6164 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6165 storing integer register in SImode. Fix cost of 256 and 512
6166 byte aligned SSE register store.
6168 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6170 * config/i386/i386.c (ix86_multiplication_cost): Fix
6171 multiplication cost for TARGET_AVX512DQ.
6173 2018-02-08 Marek Polacek <polacek@redhat.com>
6175 PR tree-optimization/84238
6176 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6179 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6181 PR tree-optimization/84265
6182 * tree-vect-stmts.c (vectorizable_store): Don't treat
6183 VMAT_CONTIGUOUS accesses as grouped.
6184 (vectorizable_load): Likewise.
6186 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6188 PR tree-optimization/81635
6189 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6190 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6191 (test_round_for_mask): New functions.
6192 (wide_int_cc_tests): Call test_round_for_mask.
6193 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6194 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6195 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6196 range returned by get_range_info.
6198 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6201 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6202 * symtab.c: Include builtins.h
6203 (symtab_node::output_to_lto_symbol_table_p): Move here
6204 from lto-streamer-out.c:output_symbol_p.
6205 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6206 (output_symbol_p): Move all logic to symtab.c
6207 (produce_symtab): Update.
6209 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6211 * config/s390/s390-opts.h (enum indirect_branch): Define.
6212 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6213 (s390_indirect_branch_via_thunk)
6214 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6215 (enum s390_indirect_branch_type): Define.
6216 * config/s390/s390.c (struct s390_frame_layout, struct
6217 machine_function): Remove.
6218 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6219 (indirect_branch_table_label_no, indirect_branch_table_name):
6221 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6222 (enum s390_indirect_branch_option): Define.
6223 (s390_return_addr_from_memory): New function.
6224 (s390_handle_string_attribute): New function.
6225 (s390_attribute_table): Add new attribute handler.
6226 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6227 (s390_indirect_branch_via_thunk): New function.
6228 (s390_indirect_branch_via_inline_thunk): New function.
6229 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6230 sibling call optimization for non z10 compiles.
6231 (s390_emit_call): Force indirect branch target to be a single
6232 register. Add r1 clobber for non-z10 compiles.
6233 (s390_emit_epilogue): Emit return jump via return_use expander.
6234 (s390_reorg): Handle JUMP_INSNs as execute targets.
6235 (s390_option_override_internal): Perform validity checks for the
6236 new command line options.
6237 (s390_indirect_branch_attrvalue): New function.
6238 (s390_indirect_branch_settings): New function.
6239 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6240 (s390_output_indirect_thunk_function): New function.
6241 (s390_code_end): Implement target hook.
6242 (s390_case_values_threshold): Implement target hook.
6243 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6245 * config/s390/s390.h (struct s390_frame_layout)
6246 (struct machine_function): Move here from s390.c.
6247 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6248 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6249 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6250 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6251 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6252 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6253 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6254 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6255 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6256 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6257 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6258 (mnemonic attribute): Add values which aren't recognized
6260 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6261 pattern for branch conversion. Fix mnemonic attribute.
6262 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6263 indirect branch via thunk if requested.
6264 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6265 ("*indirect_jump"): Disable for branch conversion using out of
6267 ("indirect_jump_via_thunk<mode>_z10")
6268 ("indirect_jump_via_thunk<mode>")
6269 ("indirect_jump_via_inlinethunk<mode>_z10")
6270 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6271 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6272 ("casesi_jump_via_inlinethunk<mode>_z10")
6273 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6274 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6275 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6276 ("*indirect2_jump"): Disable for branch conversion.
6277 ("casesi_jump"): Turn into expander and expand patterns for branch
6279 ("return_use"): New expander.
6280 ("*return"): Emit return via thunk and rename it to ...
6281 ("*return<mode>"): ... this one.
6282 * config/s390/s390.opt: Add new options and and enum for the
6285 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6287 * lra-constraints.c (match_reload): Unconditionally use
6288 gen_lowpart_SUBREG, rather than selecting between that
6289 and equivalent gen_rtx_SUBREG code.
6291 2018-02-08 Richard Biener <rguenther@suse.de>
6293 PR tree-optimization/84233
6294 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6295 changed flag instead of boguously re-using phi_inserted.
6297 2018-02-08 Martin Jambor <mjambor@suse.cz>
6299 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6300 static local variables.
6302 2018-02-08 Richard Biener <rguenther@suse.de>
6304 PR tree-optimization/84278
6305 * tree-vect-stmts.c (vectorizable_store): When looking for
6306 smaller vector types to perform grouped strided loads/stores
6307 make sure the mode is supported by the target.
6308 (vectorizable_load): Likewise.
6310 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6312 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6313 Increase LDP/STP opportunities by adding adjacent callee-saves.
6315 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6317 PR rtl-optimization/84068
6318 PR rtl-optimization/83459
6319 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6321 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6323 PR tree-optimization/84224
6324 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6325 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6328 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6331 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6332 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6333 count, remove one USE.
6335 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6337 * doc/install.texi (Configuration): Document the
6338 --with-long-double-format={ibm,ieee} PowerPC configuration
6342 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6343 Convert from define_expand to be define_insn_and_split. Rework
6344 float/double/_Float128 conversions to QI/HI/SImode to work with
6345 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6346 conversions to QI/HImode types did a store and then a load to
6347 truncate the value. For conversions to VSX registers, don't split
6348 the insn, instead emit the code directly. Use the code iterator
6349 any_fix to combine signed and unsigned conversions.
6350 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6351 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6352 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6353 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6354 (fix_<mode>di2_hw): Likewise.
6355 (fixuns_<mode>di2_hw): Likewise.
6356 (fix_<mode>si2_hw): Likewise.
6357 (fixuns_<mode>si2_hw): Likewise.
6358 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6359 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6360 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6361 fix<uns>_trunc<SFDF:mode>si2_p8.
6362 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6364 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6365 (fix<uns>_<mode>_mem): Likewise.
6366 (fctiw<u>z_<mode>_mem): Likewise.
6367 (fix<uns>_<mode>_mem): Likewise.
6368 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6369 the register allocator from doing a direct move to the GPRs to do
6370 a store, and instead use the ISA 3.0 store byte/half-word from
6371 vector register instruction. For IEEE 128-bit floating point,
6372 also optimize stores of 32-bit ints.
6373 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6375 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6377 * genextract.c (push_pathstr_operand): New function to support
6379 (walk_rtx): Call push_pathstr_operand.
6380 (print_path): Support [a-zA-Z].
6382 2018-02-07 Richard Biener <rguenther@suse.de>
6384 PR tree-optimization/84037
6385 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6386 (cse_and_gimplify_to_preheader): Declare.
6387 (vect_get_place_in_interleaving_chain): Likewise.
6388 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6390 (_loop_vec_info::~_loop_vec_info): Delete it.
6391 (cse_and_gimplify_to_preheader): New function.
6392 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6393 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6394 (vectorizable_load): Likewise. For grouped stores always base
6395 the IV on the first element.
6396 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6397 condition before gimplifying.
6399 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6401 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6402 *DIV_EXPR and *MOD_EXPR.
6404 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6407 * config/i386/i386.c (ix86_option_override_internal): Mask out
6408 the CF_SET bit when checking -fcf-protection.
6410 2018-02-07 Tom de Vries <tom@codesourcery.com>
6413 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6416 2018-02-07 Richard Biener <rguenther@suse.de>
6418 PR tree-optimization/84204
6419 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6422 PR tree-optimization/84205
6423 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6424 special-case isl_ast_op_zdiv_r.
6426 PR tree-optimization/84223
6427 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6428 Only add conditions from within the region.
6429 (gather_bbs::after_dom_children): Adjust.
6431 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6434 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6435 * config/avr/avr.md: Only post-reload split REG-REG moves if
6436 either register is GENERAL_REG_P.
6438 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6440 PR tree-optimization/84235
6441 * tree-ssa-scopedtables.c
6442 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6443 if the subtraction is performed in floating point type where NaNs are
6444 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6445 build 1. Formatting fix.
6447 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6450 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6451 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6452 and skip it regardless of bb boundaries. Use CALL_P macro,
6453 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6454 unnecessarily, formatting fix.
6456 2018-02-06 Michael Collison <michael.collison@arm.com>
6458 * config/arm/thumb2.md:
6459 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6460 (*thumb_mov_notscc): Ditto.
6462 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6465 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6466 unsigned_fix, not "s".
6468 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6470 * configure.ac (gcc_fn_eh_frame_ro): New function.
6471 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6472 correct .eh_frame permissions.
6473 * configure: Regenerate.
6475 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6477 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6480 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6482 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6483 Display warning message for -mno-speculate-indirect-jumps.
6485 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6487 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6489 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6491 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6493 PR tree-optimization/84225
6494 * tree-eh.c (find_trapping_overflow): Only call
6495 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6497 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6500 * config/i386/i386.c: Reimplement the check of possible options
6501 -mibt/-mshstk conbination. Change error messages.
6502 * doc/invoke.texi: Fix a typo: remove extra '='.
6504 2018-02-06 Marek Polacek <polacek@redhat.com>
6506 PR tree-optimization/84228
6507 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6509 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6512 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6513 emitted arch directives.
6514 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6515 __ARM_FEATURE_COPROC before changing architectures.
6517 2018-02-06 Richard Biener <rguenther@suse.de>
6519 * config/i386/i386.c (print_reg): Fix typo.
6520 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6522 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6524 * configure: Regenerate.
6526 2018-02-05 Martin Sebor <msebor@redhat.com>
6528 PR tree-optimization/83369
6529 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6532 2018-02-05 Martin Liska <mliska@suse.cz>
6534 * doc/invoke.texi: Cherry-pick upstream r323995.
6536 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6538 * ira.c (ira_init_register_move_cost): Adjust comment.
6540 2018-02-05 Martin Liska <mliska@suse.cz>
6542 PR gcov-profile/84137
6543 * doc/gcov.texi: Fix typo in documentation.
6545 2018-02-05 Martin Liska <mliska@suse.cz>
6547 PR gcov-profile/83879
6548 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6549 using dlopen functionality.
6551 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6553 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6554 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6555 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6556 _mm_maskz_range_round_ss): New intrinsics.
6557 (__builtin_ia32_rangesd128_round)
6558 (__builtin_ia32_rangess128_round): Remove.
6559 (__builtin_ia32_rangesd128_mask_round,
6560 __builtin_ia32_rangess128_mask_round): New builtins.
6561 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6562 __builtin_ia32_rangess128_round): Remove.
6563 (__builtin_ia32_rangesd128_mask_round,
6564 __builtin_ia32_rangess128_mask_round): New builtins.
6565 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6566 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6567 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6568 "<round_saeonly_constraint>")): Changed to ...
6569 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6570 "<round_saeonly_scalar_constraint>")): ... this.
6571 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6572 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6573 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6574 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6575 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6577 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6579 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6581 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6582 Remove all values except native, 8540 and 8548.
6584 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6586 * config/i386/i386.c (ix86_output_function_return): Pass
6587 INVALID_REGNUM, instead of -1, as invalid register number to
6588 indirect_thunk_name and output_indirect_thunk.
6590 2018-02-02 Julia Koval <julia.koval@intel.com>
6592 * config.gcc: Add -march=icelake.
6593 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6594 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6595 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6596 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6597 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6598 (processor_target_table): Add icelake.
6599 (ix86_option_override_internal): Handle new PTAs.
6600 (get_builtin_code_for_version): Handle icelake.
6601 (M_INTEL_COREI7_ICELAKE): New.
6602 (fold_builtin_cpu): Handle icelake.
6603 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6604 * doc/invoke.texi: Add -march=icelake.
6606 2018-02-02 Julia Koval <julia.koval@intel.com>
6608 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6609 to wide_int_bitmask.
6610 * wide-int-bitmask.h: New.
6612 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6615 * config/i386/i386.md: Replace Pmode with word_mode in
6616 builtin_setjmp_setup and builtin_longjmp to support x32.
6618 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6622 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6624 (rs6000_supported_cpu_names): New static variable.
6625 (linux_cpu_translation_table): Likewise.
6626 (elf_platform) <cpu>: Define new static variable and use it.
6627 Translate kernel AT_PLATFORM name to canonical name if needed.
6628 Error if platform name is unknown.
6630 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6633 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6635 2018-02-01 Jeff Law <law@redhat.com>
6638 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6639 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6640 the scratch if RELEASE_VIA_POP is false.
6641 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6642 If we have to save a temporary register, decrement SIZE appropriately.
6643 Pass new arguments to release_scratch_register_on_entry.
6644 (ix86_adjust_stack_and_probe): Likewise.
6645 (ix86_emit_probe_stack_range): Pass new arguments to
6646 release_scratch_register_on_entry.
6648 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6650 PR rtl-optimization/84157
6651 * combine.c (change_zero_ext): Use REG_P predicate in
6652 front of HARD_REGISTER_P predicate.
6654 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6656 * config/avr/avr.c (avr_option_override): Move disabling of
6657 -fdelete-null-pointer-checks to...
6658 * common/config/avr/avr-common.c (avr_option_optimization_table):
6661 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6663 PR tree-optimization/81635
6664 * tree-data-ref.c (split_constant_offset_1): For types that
6665 wrap on overflow, try to use range info to prove that wrapping
6668 2018-02-01 Renlin Li <renlin.li@arm.com>
6671 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6673 (aarch64_register_move_cost): Likewise.
6674 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6676 (REG_CLASS_NAMES): Likewise.
6677 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6678 TAILCALL_ADDR_REGS. Remove IP registers.
6679 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6681 2018-02-01 Richard Biener <rguenther@suse.de>
6683 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6684 for specifying RPO order and allow NULL for that.
6685 * domwalk.c (dom_walker::dom_walker): Likewise.
6686 (dom_walker::walk): Handle NULL RPO order.
6687 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6689 (rewrite_update_dom_walker): Likewise.
6690 (mark_def_dom_walker): Likewise.
6692 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6694 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6695 (aarch64_maybe_expand_sve_subreg_move): Declare.
6696 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6697 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6699 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6700 that are semantically a reverse operation.
6701 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6702 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6703 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6705 (aarch64_can_change_mode_class): For big-endian, forbid changes
6706 between two SVE modes if they have different element sizes.
6708 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6710 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6711 the TImode handling for big-endian targets.
6713 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6715 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6716 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6718 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6719 Remove BSWAP handing for big-endian targets and use the form of
6720 LD1RQ appropariate for the mode.
6722 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6724 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6725 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6728 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6731 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6732 check for operands that need to go through aarch64_sve_reload_be.
6734 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6736 PR tree-optimization/81661
6737 PR tree-optimization/84117
6738 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6739 * tree-eh.c: Include gimplify.h.
6740 (find_trapping_overflow, replace_trapping_overflow,
6741 rewrite_to_non_trapping_overflow): New functions.
6742 * tree-vect-loop.c: Include tree-eh.h.
6743 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6744 * tree-data-ref.c: Include tree-eh.h.
6745 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6747 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6749 PR rtl-optimization/84123
6750 * combine.c (change_zero_ext): Check if hard register satisfies
6751 can_change_dest_mode before calling gen_lowpart_SUBREG.
6753 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6756 * ira.c (ira_init_register_move_cost): Remove assert.
6758 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6760 PR rtl-optimization/84071
6761 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6762 * doc/tm.texi: Regenerate.
6764 2018-01-31 Richard Biener <rguenther@suse.de>
6766 PR tree-optimization/84132
6767 * tree-data-ref.c (analyze_miv_subscript): Properly
6768 check whether evolution_function_is_affine_multivariate_p
6769 before calling gcd_of_steps_may_divide_p.
6771 2018-01-31 Julia Koval <julia.koval@intel.com>
6774 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6775 * config/i386/i386.md (rdpid_rex64) New.
6776 (rdpid): Make 32bit only.
6778 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6781 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6782 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6784 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6787 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6789 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6791 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6793 PR rtl-optimization/84071
6794 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6795 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6797 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6799 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6800 (arc_attribute_table): Add 'aux' attribute.
6801 (arc_in_small_data_p): Consider aux like variables.
6802 (arc_is_aux_reg_p): New function.
6803 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6804 (arc_get_aux_arg): New function.
6805 (prepare_move_operands): Handle aux-register access.
6806 (arc_handle_aux_attribute): New function.
6807 * doc/extend.texi (ARC Variable attributes): Add subsection.
6809 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6811 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6812 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6813 (arc_attribute_table): Add 'uncached' attribute.
6814 (arc_print_operand): Print '.di' flag for uncached memory
6816 (arc_in_small_data_p): Do not consider for small data the uncached
6818 (arc_is_uncached_mem_p): New function.
6819 * config/arc/predicates.md (compact_store_memory_operand): Check
6820 for uncached memory accesses.
6821 (nonvol_nonimm_operand): Likewise.
6822 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6824 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6827 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6828 falign-loops=): Add Optimization flag.
6830 2018-01-30 Jeff Law <law@redhat.com>
6833 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6834 INT_REGISTERS_SAVED. Check it prior to calling
6835 get_scratch_register_on_entry.
6836 (ix86_adjust_stack_and_probe): Similarly.
6837 (ix86_emit_probe_stack_range): Similarly.
6838 (ix86_expand_prologue): Corresponding changes.
6840 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6843 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6844 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6846 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6849 * lra-constraints.c (curr_insn_transform): Process AND in the
6852 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6854 PR rtl-optimization/83986
6855 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6856 dependence against last_pending_memory_flush in addition to
6859 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6861 PR tree-optimization/81611
6862 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6865 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6868 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6871 2018-01-30 Richard Biener <rguenther@suse.de>
6872 Jakub Jelinek <jakub@redhat.com>
6874 PR tree-optimization/84111
6875 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6876 inner loops added during recursion, as they don't have up-to-date
6879 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6882 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6883 (can_inline_edge_by_limits_p): ... here.
6884 (can_early_inline_edge_p, check_callers,
6885 update_caller_keys, update_callee_keys, recursive_inlining,
6886 add_new_edges_to_heap, speculation_useful_p,
6887 inline_small_functions,
6888 inline_small_functions, flatten_function,
6889 inline_to_all_callers_1): Update.
6891 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6893 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6896 2018-01-30 Richard Biener <rguenther@suse.de>
6898 PR tree-optimization/83008
6899 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6900 invariant and constant vector uses in stmts when they need
6903 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6906 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6907 * configure: Regenerate.
6909 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6911 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6913 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6914 Use gen_rtx_REG rather than gen_lowpart.
6916 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6918 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6919 rather than 0 when creating partial subregs.
6921 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6923 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6926 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6929 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6930 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6931 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6932 flags. This restores the settings used before the 2017-07-24.
6933 Turning off pre increment/decrement/modify allows IVOPTS to
6934 optimize DF/SF loops where the index is an int.
6936 2018-01-29 Richard Biener <rguenther@suse.de>
6937 Kelvin Nilsen <kelvin@gcc.gnu.org>
6940 * tree-vect-stmts.c (vectorizable_call): Don't call
6941 targetm.vectorize_builtin_md_vectorized_function if callee is
6944 2018-01-22 Carl Love <cel@us.ibm.com>
6946 * doc/extend.tex: Fix typo in second arg in
6947 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6949 2018-01-29 Richard Biener <rguenther@suse.de>
6951 PR tree-optimization/84086
6952 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6953 (flush_ssaname_freelist): When SSA names were released reset
6954 the SCEV hash table.
6956 2018-01-29 Richard Biener <rguenther@suse.de>
6958 PR tree-optimization/84057
6959 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6960 removed paths when removing edges.
6962 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6964 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6965 -mfunction-return=@var{choice}.
6967 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6970 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6972 (layout::print_source_line): Likewise.
6973 (test_get_line_width_without_trailing_whitespace): Add test cases.
6975 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6978 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6981 2018-01-26 Jim Wilson <jimw@sifive.com>
6983 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6985 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6988 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6990 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6991 and CMP + SUB-immediate -> SUBS.
6993 2018-01-26 Martin Sebor <msebor@redhat.com>
6995 PR tree-optimization/83896
6996 * tree-ssa-strlen.c (get_string_len): Rename...
6997 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6998 Avoid assuming length is constant.
6999 (handle_char_store): Use HOST_WIDE_INT for string length.
7001 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7004 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7005 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7007 2018-01-26 Richard Biener <rguenther@suse.de>
7009 PR rtl-optimization/84003
7010 * dse.c (record_store): Only record redundant stores when
7011 the earlier store aliases at least all accesses the later one does.
7013 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7015 PR rtl-optimization/83985
7016 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7017 REG_CFA_RESTORE insns.
7018 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7019 REG_CFA_RESTORE insns here.
7022 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7023 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7025 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7027 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7028 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7029 (arc_init): Likewise.
7030 (arc_override_options): Likewise.
7031 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7033 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7035 * config/arc/arc.h (TARGET_DBNZ): Define.
7036 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7037 properly set the tune attribute.
7038 (dbnz): Use TARGET_DBNZ guard.
7039 * config/arc/arc.opt (mtune): Add core3 option.
7041 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7043 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7044 recognize new pic like addresses.
7045 (arc_delegitimize_address): Clean up.
7047 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7049 * config/arc/arc-arches.def: Option mrf16 valid for all
7051 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7052 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7053 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7054 * config/arc/arc-tables.opt: Regenerate.
7055 * config/arc/arc.c (arc_conditional_register_usage): Handle
7056 reduced register file case.
7057 (arc_file_start): Set must have build attributes.
7058 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7060 * config/arc/arc.opt (mrf16): Add new option.
7061 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7062 * config/arc/genmultilib.awk: Handle new mrf16 option.
7063 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7064 * config/arc/t-multilib: Regenerate.
7065 * doc/invoke.texi (ARC Options): Document mrf16 option.
7067 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7069 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7070 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7071 (arc_attribute_table): Add 'secure_call' attribute.
7072 (arc_print_operand): Print secure call operand.
7073 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7075 (arc_is_secure_call_p): New function. * config/arc/arc.md
7076 (call_i): Add support for sjli instruction.
7077 (call_value_i): Likewise.
7078 * config/arc/constraints.md (Csc): New constraint.
7080 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7081 John Eric Martin <John.Martin@emmicro-us.com>
7083 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7084 * config/arc/arc.c (_arc_jli_section): New struct.
7085 (arc_jli_section): New type.
7086 (rc_jli_sections): New static variable.
7087 (arc_handle_jli_attribute): New function.
7088 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7089 (arc_file_end): New function.
7090 (TARGET_ASM_FILE_END): Define.
7091 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7092 (arc_add_jli_section): New function.
7093 (jli_call_scan): Likewise.
7094 (arc_reorg): Call jli_call_scan.
7095 (arc_output_addsi): Remove 'S' from printing asm operand.
7096 (arc_is_jli_call_p): New function.
7097 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7099 (movhi_insn): Likewise.
7100 (movsi_insn): Likewise.
7101 (movsi_set_cc_insn): Likewise.
7102 (loadqi_update): Likewise.
7103 (load_zeroextendqisi_update): Likewise.
7104 (load_signextendqisi_update): Likewise.
7105 (loadhi_update): Likewise.
7106 (load_zeroextendhisi_update): Likewise.
7107 (load_signextendhisi_update): Likewise.
7108 (loadsi_update): Likewise.
7109 (loadsf_update): Likewise.
7110 (movsicc_insn): Likewise.
7111 (bset_insn): Likewise.
7112 (bxor_insn): Likewise.
7113 (bclr_insn): Likewise.
7114 (bmsk_insn): Likewise.
7115 (bicsi3_insn): Likewise.
7116 (cmpsi_cc_c_insn): Likewise.
7117 (movsi_ne): Likewise.
7118 (movsi_cond_exec): Likewise.
7119 (clrsbsi2): Likewise.
7130 (call_i): Remove 'S' asm letter, add jli instruction.
7131 (call_value_i): Likewise.
7132 * config/arc/arc.op (mjli-always): New option.
7133 * config/arc/constraints.md (Cji): New constraint.
7134 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7136 (subsf3_fpx): Likewise.
7137 (mulsf3_fpx): Likewise.
7138 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7140 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7142 * doc/invoke.texi (ARC): Document mjli-always option.
7144 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7146 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7147 avoid addition with 0 and use incw and decw where possible.
7149 2018-01-26 Richard Biener <rguenther@suse.de>
7151 PR tree-optimization/81082
7152 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7153 association if it requires casting to unsigned.
7154 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7155 from fold_plusminus_mult_expr to catch important cases late when
7156 range info is available.
7158 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7160 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7161 * configure.ac (hidden_linkonce): New test.
7162 * configure: Regenerate.
7163 * config.in: Regenerate.
7165 2018-01-26 Julia Koval <julia.koval@intel.com>
7167 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7168 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7169 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7170 _mm_mask_bitshuffle_epi64_mask): Fix type.
7171 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7172 USI_FTYPE_V4DI_V4DI_USI): Remove.
7173 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7174 __builtin_ia32_vpshufbitqmb256_mask,
7175 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7176 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7177 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7179 2018-01-26 Alan Modra <amodra@gmail.com>
7182 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7183 UNSPEC_VBPERMQ. Sort other unspecs.
7185 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7187 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7189 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7192 * predict.c (drop_profile): Do not push/pop cfun; update also
7194 (handle_missing_profiles): Fix logic looking for zero profiles.
7196 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7199 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7200 on functions with #pragma omp declare simd or functions with simd
7202 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7203 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7204 Remove trailing \n from warning_at calls.
7206 2018-01-25 Tom de Vries <tom@codesourcery.com>
7209 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7210 for neutered workers.
7212 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7215 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7216 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7218 2018-01-24 Jeff Law <law@redhat.com>
7221 * i386.c (get_probe_interval): Move to earlier point.
7222 (ix86_compute_frame_layout): If -fstack-clash-protection and
7223 the frame is larger than the probe interval, then use pushes
7224 to save registers rather than reg->mem moves.
7225 (ix86_expand_prologue): Remove conditional for int_registers_saved
7228 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7231 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7232 min/max for never referenced object.
7234 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7237 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7239 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7240 attributes from DECL_ATTRIBUTES (decl) without affecting
7241 DECL_ATTRIBUTES (current_function_decl).
7242 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7243 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7245 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7247 PR tree-optimization/83979
7248 * fold-const.c (fold_comparison): Use constant_boolean_node
7249 instead of boolean_{true,false}_node.
7251 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7253 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7256 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7258 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7259 Simplify the clause that sets the length attribute.
7260 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7261 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7262 clause that sets the length attribute.
7263 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7265 2018-01-24 Tom de Vries <tom@codesourcery.com>
7268 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7269 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7270 Add strict parameter.
7271 (prevent_branch_around_nothing): Insert dummy insn between branch to
7272 label and label with no ptx insn inbetween.
7273 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7275 2018-01-24 Tom de Vries <tom@codesourcery.com>
7278 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7279 for neutered threads in warp.
7280 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7282 2018-01-24 Richard Biener <rguenther@suse.de>
7284 PR tree-optimization/83176
7285 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7288 2018-01-24 Richard Biener <rguenther@suse.de>
7290 PR tree-optimization/82819
7291 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7292 code generating pluses that are no-ops in the target precision.
7294 2018-01-24 Richard Biener <rguenther@suse.de>
7297 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7299 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7301 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7302 to merge probabilities.
7303 * predict.c (probably_never_executed): Also mark as cold functions
7304 with global 0 profile and guessed local profile.
7305 * profile-count.c (profile_probability::combine_with_count): New
7307 * profile-count.h (profile_probability::operator*,
7308 profile_probability::operator*=, profile_probability::operator/,
7309 profile_probability::operator/=): Reduce precision to adjusted
7310 and set value to guessed on contradictory divisions.
7311 (profile_probability::combine_with_freq): Remove.
7312 (profile_probability::combine_wiht_count): Declare.
7313 (profile_count::force_nonzero):: Set to adjusted.
7314 (profile_count::probability_in):: Set quality to adjusted.
7315 * tree-ssa-tail-merge.c (replace_block_by): Use
7318 2018-01-23 Andrew Waterman <andrew@sifive.com>
7319 Jim Wilson <jimw@sifive.com>
7321 * config/riscv/riscv.c (riscv_stack_boundary): New.
7322 (riscv_option_override): Set riscv_stack_boundary. Handle
7323 riscv_preferred_stack_boundary_arg.
7324 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7325 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7326 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7327 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7328 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7329 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7331 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7334 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7335 of struct ix86_frame.
7336 (ix86_expand_epilogue): Likewise. Add a local variable for
7337 the reg_save_offset field in struct ix86_frame.
7339 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7341 PR tree-optimization/82604
7342 * tree-loop-distribution.c (enum partition_kind): New enum item
7343 PKIND_PARTIAL_MEMSET.
7344 (partition_builtin_p): Support above new enum item.
7345 (generate_code_for_partition): Ditto.
7346 (compute_access_range): Differentiate cases that equality can be
7347 proven at all loops, the innermost loops or no loops.
7348 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7349 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7350 (finalize_partitions, distribute_loop): Don't fuse partition of
7351 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7352 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7355 2018-01-23 Martin Liska <mliska@suse.cz>
7357 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7358 order to ignore the predictor.
7359 (PRED_POLYMORPHIC_CALL): Likewise.
7360 (PRED_RECURSIVE_CALL): Likewise.
7362 2018-01-23 Martin Liska <mliska@suse.cz>
7364 * tree-profile.c (tree_profiling): Print function header to
7365 aware reader which function we are working on.
7366 * value-prof.c (gimple_find_values_to_profile): Do not print
7367 not interesting value histograms.
7369 2018-01-23 Martin Liska <mliska@suse.cz>
7371 * profile-count.h (enum profile_quality): Add
7372 profile_uninitialized as the first value. Do not number values
7373 as they are zero based.
7374 (profile_count::verify): Update sanity check.
7375 (profile_probability::verify): Likewise.
7377 2018-01-23 Nathan Sidwell <nathan@acm.org>
7379 * doc/invoke.texi (ffor-scope): Deprecate.
7381 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7383 PR tree-optimization/83510
7384 * domwalk.c (set_all_edges_as_executable): New function.
7385 (dom_walker::dom_walker): Convert bool param
7386 "skip_unreachable_blocks" to enum reachability. Move setup of
7387 edge flags to set_all_edges_as_executable and only do it when
7388 reachability is REACHABLE_BLOCKS.
7389 * domwalk.h (enum dom_walker::reachability): New enum.
7390 (dom_walker::dom_walker): Convert bool param
7391 "skip_unreachable_blocks" to enum reachability.
7392 (set_all_edges_as_executable): New decl.
7393 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7394 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7396 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7397 but converting true to REACHABLE_BLOCKS.
7398 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7400 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7401 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7402 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7404 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7405 if check_all_array_refs will be called.
7407 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7409 * tree.c (selftest::test_location_wrappers): Add more test
7412 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7414 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7415 (selftest::test_bit_in_range): Likewise.
7417 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7420 * doc/sourcebuild.texi (vect_float): Say that the selector
7421 only describes the situation when -funsafe-math-optimizations is on.
7422 (vect_float_strict): Document.
7424 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7426 PR tree-optimization/83965
7427 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7428 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7429 instead of checking only for a reduction.
7430 (vect_recog_widen_sum_pattern): Likewise.
7432 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7434 * predict.c (probably_never_executed): Only use precise profile info.
7435 (compute_function_frequency): Skip after inlining hack since we now
7436 have quality checking.
7438 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7440 * profile-count.h (profile_probability::very_unlikely,
7441 profile_probability::unlikely, profile_probability::even): Set
7442 precision to guessed.
7444 2018-01-23 Richard Biener <rguenther@suse.de>
7446 PR tree-optimization/83963
7447 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7448 Properly terminate dominator walk when crossing the exit edge not
7449 when visiting its source block.
7451 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7454 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7455 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7457 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7459 PR tree-optimization/83957
7460 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7461 semicolon after for body surrounded by braces.
7463 PR tree-optimization/83081
7464 * profile-count.h (profile_probability::split): New method.
7465 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7466 Use profile_probability::split.
7467 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7468 when splitting a single conditional jump into 2.
7470 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7472 PR tree-optimization/69452
7473 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7476 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7478 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7479 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7480 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7482 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7484 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7485 * config/rl78/rl78.md (movdi): New define_expand.
7486 * config/rl78/rl78.c (rl78_split_movdi): New function.
7488 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7491 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7493 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7494 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7495 128-bit to produce an UNSPEC move to get the double word with the
7496 signbit and then a shift directly to do signbit.
7497 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7498 implementation with a new version that just does either a direct
7499 move or a regular move. Move memory interface to separate insns.
7500 Move insns so they are next to the expander.
7501 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7502 with signbit move. Split big and little endian case.
7503 (signbit<mode>2_dm_mem_le): Likewise.
7504 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7505 (signbit<mode>2_dm2): Likewise.
7507 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7509 * config/rl78/rl78.md (anddi3): New define_expand.
7511 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7513 * config/rl78/rl78.md (umindi3): New define_expand.
7515 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7517 * config/rl78/rl78.md (smindi3): New define_expand.
7519 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7521 * config/rl78/rl78.md (smaxdi3): New define_expand.
7523 2018-01-22 Carl Love <cel@us.ibm.com>
7525 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7526 LVX_V1TI): Add macro expansion.
7527 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7528 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7529 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7530 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7531 Change check to determine if the instruction is a byte reversing
7532 entry. Fix typo in comment.
7533 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7534 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7535 Add def_builtin calls for new builtins.
7536 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7537 Add define_insn expansion.
7539 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7541 * config/rl78/rl78.md (umaxdi3): New define_expand.
7543 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7545 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7546 for non-QImode registers.
7548 2018-01-22 Richard Biener <rguenther@suse.de>
7550 PR tree-optimization/83963
7551 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7552 including the loop exit block.
7553 (scop_detection::merge_sese): Likewise.
7554 (scop_detection::add_scop): Do it here instead.
7556 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7558 * doc/sourcebuild.texi (arm_softfloat): Document.
7560 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7563 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7564 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7565 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7567 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7568 David Edelsohn <dje.gcc@gmail.com>
7571 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7572 Change "crset eq" to "crset 2".
7573 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7574 (*call_indirect_aix<mode>_nospec): Likewise.
7575 (*call_value_indirect_aix<mode>_nospec): Likewise.
7576 (*call_indirect_elfv2<mode>_nospec): Likewise.
7577 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7578 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7579 change assembly output from . to $.
7580 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7581 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7582 (*tablejump<mode>_internal1_nospec): Likewise.
7584 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7587 * config/sh/sh_optimize_sett_clrt.cc:
7588 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7590 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7592 PR tree-optimization/83940
7593 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7594 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7595 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7596 use it to pass back the definition type.
7597 (vect_check_store_rhs): Likewise rhs_dt_out.
7598 (vect_build_gather_load_calls): Add a mask_dt argument and use
7599 it instead of a call to vect_is_simple_use.
7600 (vectorizable_store): Update calls to vect_check_load_store_mask
7601 and vect_check_store_rhs. Use the dt returned by the latter instead
7602 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7603 instead of calls to vect_is_simple_use. Pass the scalar rather
7604 than the vector operand to vect_is_simple_use when handling
7605 second and subsequent copies of an rhs value.
7606 (vectorizable_load): Update calls to vect_check_load_store_mask
7607 and vect_build_gather_load_calls. Use the cached mask_dt and
7608 gs_info.offset_dt instead of calls to vect_is_simple_use.
7610 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7613 * tree-emutls.c: Include gimplify.h.
7614 (lower_emutls_2): New function.
7615 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7616 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7617 it before further processing.
7620 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7621 UINTVAL (trueop1) instead of INTVAL (op1).
7623 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7627 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7628 INCOMING_FRAME_SP_OFFSET if not defined.
7629 (scan_trace): Add ENTRY argument. If true and
7630 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7631 emit a note to adjust the CFA offset.
7632 (create_cfi_notes): Adjust scan_trace callers.
7633 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7634 INCOMING_FRAME_SP_OFFSET in the CIE.
7635 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7636 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7638 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7639 * doc/tm.texi: Regenerated.
7641 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7643 PR rtl-optimization/83147
7644 * lra-constraints.c (remove_inheritance_pseudos): Use
7645 lra_substitute_pseudo_within_insn.
7647 2018-01-19 Tom de Vries <tom@codesourcery.com>
7648 Cesar Philippidis <cesar@codesourcery.com>
7651 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7653 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7656 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7657 spaces for function labels.
7659 2018-01-19 Martin Liska <mliska@suse.cz>
7661 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7662 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7663 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7664 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7665 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7666 (PRED_CONST_RETURN): Change from 69 to 65.
7667 (PRED_NULL_RETURN): Change from 91 to 71.
7668 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7669 (PRED_LOOP_GUARD): Change from 66 to 73.
7671 2018-01-19 Martin Liska <mliska@suse.cz>
7673 * predict.c (predict_insn_def): Add new assert.
7674 (struct branch_predictor): Change type to signed integer.
7675 (test_prediction_value_range): Amend test to cover
7677 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7678 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7679 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7680 (PRED_LOOP_IV_COMPARE): Likewise.
7681 * predict.h (PROB_UNINITIALIZED): Define new constant.
7683 2018-01-19 Martin Liska <mliska@suse.cz>
7685 * predict.c (dump_prediction): Add new format for
7686 analyze_brprob.py script which is enabled with -details
7688 * profile-count.h (precise_p): New function.
7690 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7692 PR tree-optimization/83922
7693 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7694 there are no statements that need masking.
7695 (vect_active_double_reduction_p): New function.
7696 (vect_analyze_loop_operations): Use it when handling phis that
7697 are not in the loop header.
7699 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7701 PR tree-optimization/83914
7702 * tree-vect-loop.c (vectorizable_induction): Don't convert
7703 init_expr or apply the peeling adjustment for inductions
7704 that are nested within the vectorized loop.
7706 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7708 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7711 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7715 * function.h (gimplify_parameters): Add gimple_seq * argument.
7716 * function.c: Include gimple.h and options.h.
7717 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7718 for the added local temporaries if needed.
7719 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7720 if there are any parameter cleanups, wrap whole body into a
7721 try/finally with the cleanups.
7723 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7726 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7727 Use GET_MODE_CLASS for scalar floating point.
7729 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7733 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7734 Fix call of call_cgraph_insertion_hooks.
7736 2018-01-18 Martin Sebor <msebor@redhat.com>
7738 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7740 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7743 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7746 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7749 * common.opt: (-ffile-prefix-map): New option.
7750 * opts.c (common_handle_option): Defer it.
7751 * opts-global.c (handle_common_deferred_options): Handle it.
7752 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7753 * file-prefix-map.h: New file.
7754 (remap_debug_filename, add_debug_prefix_map): ...here.
7755 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7756 * final.c (debug_prefix_map, add_debug_prefix_map
7757 remap_debug_filename): Move to...
7758 * file-prefix-map.c: New file.
7759 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7760 generalize, get rid of alloca(), use strrchr() instead of strchr().
7761 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7762 Implement in terms of add_prefix_map().
7763 (remap_macro_filename, remap_debug_filename): Implement in term of
7765 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7766 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7767 * dbxout.c: Include file-prefix-map.h.
7768 * varasm.c: Likewise.
7769 * vmsdbgout.c: Likewise.
7770 * xcoffout.c: Likewise.
7771 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7772 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7773 * doc/invoke.texi (-ffile-prefix-map): Document.
7774 (-fdebug-prefix-map): Update description.
7776 2018-01-18 Martin Liska <mliska@suse.cz>
7778 * config/i386/i386.c (indirect_thunk_name): Document that also
7780 (output_indirect_thunk): Document why both instructions
7781 (pause and lfence) are generated.
7783 2018-01-18 Richard Biener <rguenther@suse.de>
7785 PR tree-optimization/83887
7786 * graphite-scop-detection.c
7787 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7788 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7789 (scop_detection::merge_sese): Re-implement with a flood-fill
7790 algorithm that properly finds a SESE region if it exists.
7792 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7795 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7796 pointer_diff optimizations use view_convert instead of convert.
7798 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7800 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7801 Generate different code for -mno-speculate-indirect-jumps.
7802 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7803 (*call_indirect_aix<mode>): Disable for
7804 -mno-speculate-indirect-jumps.
7805 (*call_indirect_aix<mode>_nospec): New define_insn.
7806 (*call_value_indirect_aix<mode>): Disable for
7807 -mno-speculate-indirect-jumps.
7808 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7809 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7810 -mno-speculate-indirect-jumps.
7811 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7813 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7815 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7816 long double type, set the flags for noting the default long double
7817 type, even if we don't pass or return a long double type.
7819 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7822 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7825 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7827 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7828 support for merge[hl].
7829 (fold_mergehl_helper): New helper function.
7830 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7831 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7832 (altivec_vmrglw_direct): Add xxmrglw insn.
7834 2018-01-17 Andrew Waterman <andrew@sifive.com>
7836 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7837 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7839 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7842 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7843 call the lto_location_cache before reading the
7844 DECL_SOURCE_LOCATION of the types.
7846 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7847 Richard Sandiford <richard.sandiford@linaro.org>
7849 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7850 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7851 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7852 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7853 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7855 * config/aarch64/constraints.md (aarch64_movti_operand):
7857 * config/aarch64/predicates.md (Uti): Add new constraint.
7859 2018-01-17 Carl Love <cel@us.ibm.com>
7861 * config/rs6000/vsx.md (define_expand xl_len_r,
7862 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7863 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7865 (define_expand, define_insn): Move the shift left from the
7866 define_insn to the define_expand for lxvl and stxvl instructions.
7867 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7868 and XL_LEN_R definitions to PURE.
7870 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7872 * config/i386/i386.c (indirect_thunk_name): Declare regno
7873 as unsigned int. Compare regno with INVALID_REGNUM.
7874 (output_indirect_thunk): Ditto.
7875 (output_indirect_thunk_function): Ditto.
7876 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7877 in the call to output_indirect_thunk_function.
7879 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7882 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7883 rather than the size of inner_type to determine the stack slot size
7884 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7886 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7889 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7892 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7894 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7895 endian Linux systems to optionally enable multilibs for selecting
7896 the long double type if the user configured an explicit type.
7897 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7898 have no long double multilibs if not defined.
7899 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7900 warn if the user used -mabi={ieee,ibm}longdouble and we built
7901 multilibs for long double.
7902 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7903 appropriate multilib option.
7904 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7906 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7907 for building long double multilibs.
7908 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7910 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7912 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7915 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7917 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7920 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7923 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7926 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7928 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7929 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7931 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7933 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7934 different rtl trees depending on TARGET_64BIT.
7935 (rs6000_gen_lvx): Likewise.
7937 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7939 * config/visium/visium.md (nop): Tweak comment.
7940 (hazard_nop): Likewise.
7942 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7944 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7945 -mspeculate-indirect-jumps.
7946 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7947 for -mno-speculate-indirect-jumps.
7948 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7949 (*call_value_indirect_elfv2<mode>): Disable for
7950 -mno-speculate-indirect-jumps.
7951 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7952 (indirect_jump): Emit different RTL for
7953 -mno-speculate-indirect-jumps.
7954 (*indirect_jump<mode>): Disable for
7955 -mno-speculate-indirect-jumps.
7956 (*indirect_jump<mode>_nospec): New define_insn.
7957 (tablejump): Emit different RTL for
7958 -mno-speculate-indirect-jumps.
7959 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7960 (tablejumpsi_nospec): New define_expand.
7961 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7962 (tablejumpdi_nospec): New define_expand.
7963 (*tablejump<mode>_internal1): Disable for
7964 -mno-speculate-indirect-jumps.
7965 (*tablejump<mode>_internal1_nospec): New define_insn.
7966 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7969 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7971 * caller-save.c (insert_save): Drop unnecessary parameter. All
7974 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7975 Richard Biener <rguenth@suse.de>
7978 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7979 return early, inline manually is_gimple_sizepos. Make sure if we
7980 call gimplify_expr we don't end up with a gimple constant.
7981 * tree.c (variably_modified_type_p): Don't return true for
7982 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7983 * gimplify.h (is_gimple_sizepos): Remove.
7985 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7987 PR tree-optimization/83857
7988 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7989 vectorizable_live_operation for pure SLP statements.
7990 (vectorizable_live_operation): Handle PHIs.
7992 2018-01-16 Richard Biener <rguenther@suse.de>
7994 PR tree-optimization/83867
7995 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7996 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7998 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8001 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8002 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8003 If off is not INTEGER_CST, issue a may not be aligned warning
8004 rather than isn't aligned. Use isn%'t rather than isn't.
8005 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8007 <case MULT_EXPR>: Improve the case when bottom and one of the
8008 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8009 operand, in that case check if the other operand is multiple of
8010 bottom divided by the INTEGER_CST operand.
8012 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8015 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8016 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8017 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8018 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8019 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8020 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8021 (pa_function_arg_size): New function.
8023 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8025 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8026 in a separate statement.
8028 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8030 PR tree-optimization/83847
8031 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8032 group gathers and scatters.
8034 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8036 PR rtl-optimization/86620
8037 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8039 PR rtl-optimization/83213
8040 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8041 to last if both are JUMP_INSNs.
8043 PR tree-optimization/83843
8044 * gimple-ssa-store-merging.c
8045 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8046 store_immediate_info for bswap/nop orig_stores.
8048 2018-01-15 Andrew Waterman <andrew@sifive.com>
8050 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8052 <UDIV>: Increase cost if !TARGET_DIV.
8054 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8056 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8057 (define_attr "cr_logical_3op"): New.
8058 (cceq_ior_compare): Adjust.
8059 (cceq_ior_compare_complement): Adjust.
8060 (*cceq_rev_compare): Adjust.
8061 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8062 (is_cracked_insn): Adjust.
8063 (insn_must_be_first_in_group): Adjust.
8064 * config/rs6000/40x.md: Adjust.
8065 * config/rs6000/440.md: Adjust.
8066 * config/rs6000/476.md: Adjust.
8067 * config/rs6000/601.md: Adjust.
8068 * config/rs6000/603.md: Adjust.
8069 * config/rs6000/6xx.md: Adjust.
8070 * config/rs6000/7450.md: Adjust.
8071 * config/rs6000/7xx.md: Adjust.
8072 * config/rs6000/8540.md: Adjust.
8073 * config/rs6000/cell.md: Adjust.
8074 * config/rs6000/e300c2c3.md: Adjust.
8075 * config/rs6000/e500mc.md: Adjust.
8076 * config/rs6000/e500mc64.md: Adjust.
8077 * config/rs6000/e5500.md: Adjust.
8078 * config/rs6000/e6500.md: Adjust.
8079 * config/rs6000/mpc.md: Adjust.
8080 * config/rs6000/power4.md: Adjust.
8081 * config/rs6000/power5.md: Adjust.
8082 * config/rs6000/power6.md: Adjust.
8083 * config/rs6000/power7.md: Adjust.
8084 * config/rs6000/power8.md: Adjust.
8085 * config/rs6000/power9.md: Adjust.
8086 * config/rs6000/rs64.md: Adjust.
8087 * config/rs6000/titan.md: Adjust.
8089 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8091 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8092 ix86_indirect_branch_register logic.
8094 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8096 * config/i386/constraints.md (Bs): Update
8097 ix86_indirect_branch_register check. Don't check
8098 ix86_indirect_branch_register with GOT_memory_operand.
8100 * config/i386/predicates.md (GOT_memory_operand): Don't check
8101 ix86_indirect_branch_register here.
8102 (GOT32_symbol_operand): Likewise.
8104 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8106 * config/i386/predicates.md (constant_call_address_operand):
8107 Rewrite ix86_indirect_branch_register logic.
8108 (sibcall_insn_operand): Likewise.
8110 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8112 * config/i386/constraints.md (Bs): Replace
8113 ix86_indirect_branch_thunk_register with
8114 ix86_indirect_branch_register.
8116 * config/i386/i386.md (indirect_jump): Likewise.
8117 (tablejump): Likewise.
8118 (*sibcall_memory): Likewise.
8119 (*sibcall_value_memory): Likewise.
8120 Peepholes of indirect call and jump via memory: Likewise.
8121 * config/i386/i386.opt: Likewise.
8122 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8123 (GOT_memory_operand): Likewise.
8124 (call_insn_operand): Likewise.
8125 (sibcall_insn_operand): Likewise.
8126 (GOT32_symbol_operand): Likewise.
8128 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8131 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8132 type rather than type addr's type points to.
8133 (expand_omp_atomic_mutex): Likewise.
8134 (expand_omp_atomic): Likewise.
8136 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8139 * config/i386/i386.c (output_indirect_thunk_function): Use
8140 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8141 for __x86_return_thunk.
8143 2018-01-15 Richard Biener <rguenther@suse.de>
8146 * expmed.c (extract_bit_field_1): Fix typo.
8148 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8151 * config/arm/iterators.md (VF): New mode iterator.
8152 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8153 Remove integer-related logic from pattern.
8154 (neon_vabd<mode>_3): Likewise.
8156 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8159 * common.opt (fstrict-overflow): No longer an alias.
8160 (fwrapv-pointer): New option.
8161 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8162 also for pointer types based on flag_wrapv_pointer.
8163 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8164 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8165 opts->x_flag_wrapv got set.
8166 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8167 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8168 POINTER_TYPE_OVERFLOW_UNDEFINED.
8169 * match.pd: Likewise in address comparison pattern.
8170 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8172 2018-01-15 Richard Biener <rguenther@suse.de>
8175 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8176 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8177 Reset type names to their identifier if their TYPE_DECL doesn't
8178 have linkage (and thus is used for ODR and devirt).
8179 (save_debug_info_for_decl): Remove.
8180 (save_debug_info_for_type): Likewise.
8181 (add_tree_to_fld_list): Adjust.
8182 * tree-pretty-print.c (dump_generic_node): Make dumping of
8183 type names more robust.
8185 2018-01-15 Richard Biener <rguenther@suse.de>
8187 * BASE-VER: Bump to 8.0.1.
8189 2018-01-14 Martin Sebor <msebor@redhat.com>
8192 * builtins.c (check_access): Avoid warning when the no-warning bit
8195 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8197 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8198 * ira-color (allocno_hard_regs_compare): Likewise.
8200 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8203 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8204 Use .pushsection/.popsection.
8206 2018-01-14 Martin Sebor <msebor@redhat.com>
8209 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8211 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8213 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8214 entry from extra_headers.
8215 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8216 extra_headers, make the list bitwise identical to the i?86-*-* one.
8218 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8220 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8221 -mcmodel=large with -mindirect-branch=thunk,
8222 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8223 -mfunction-return=thunk-extern.
8224 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8225 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8226 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8228 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8230 * config/i386/i386.c (print_reg): Print the name of the full
8231 integer register without '%'.
8232 (ix86_print_operand): Handle 'V'.
8233 * doc/extend.texi: Document 'V' modifier.
8235 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8237 * config/i386/constraints.md (Bs): Disallow memory operand for
8238 -mindirect-branch-register.
8240 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8241 (GOT_memory_operand): Likewise.
8242 (call_insn_operand): Likewise.
8243 (sibcall_insn_operand): Likewise.
8244 (GOT32_symbol_operand): Likewise.
8245 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8246 for -mindirect-branch-register.
8247 (tablejump): Likewise.
8248 (*sibcall_memory): Likewise.
8249 (*sibcall_value_memory): Likewise.
8250 Disallow peepholes of indirect call and jump via memory for
8251 -mindirect-branch-register.
8252 (*call_pop): Replace m with Bw.
8253 (*call_value_pop): Likewise.
8254 (*sibcall_pop_memory): Replace m with Bs.
8255 * config/i386/i386.opt (mindirect-branch-register): New option.
8256 * doc/invoke.texi: Document -mindirect-branch-register option.
8258 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8260 * config/i386/i386-protos.h (ix86_output_function_return): New.
8261 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8262 set function_return_type.
8263 (indirect_thunk_name): Add ret_p to indicate thunk for function
8265 (output_indirect_thunk_function): Pass false to
8266 indirect_thunk_name.
8267 (ix86_output_indirect_branch_via_reg): Likewise.
8268 (ix86_output_indirect_branch_via_push): Likewise.
8269 (output_indirect_thunk_function): Create alias for function
8270 return thunk if regno < 0.
8271 (ix86_output_function_return): New function.
8272 (ix86_handle_fndecl_attribute): Handle function_return.
8273 (ix86_attribute_table): Add function_return.
8274 * config/i386/i386.h (machine_function): Add
8275 function_return_type.
8276 * config/i386/i386.md (simple_return_internal): Use
8277 ix86_output_function_return.
8278 (simple_return_internal_long): Likewise.
8279 * config/i386/i386.opt (mfunction-return=): New option.
8280 (indirect_branch): Mention -mfunction-return=.
8281 * doc/extend.texi: Document function_return function attribute.
8282 * doc/invoke.texi: Document -mfunction-return= option.
8284 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8286 * config/i386/i386-opts.h (indirect_branch): New.
8287 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8288 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8289 with local indirect jump when converting indirect call and jump.
8290 (ix86_set_indirect_branch_type): New.
8291 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8292 (indirectlabelno): New.
8293 (indirect_thunk_needed): Likewise.
8294 (indirect_thunk_bnd_needed): Likewise.
8295 (indirect_thunks_used): Likewise.
8296 (indirect_thunks_bnd_used): Likewise.
8297 (INDIRECT_LABEL): Likewise.
8298 (indirect_thunk_name): Likewise.
8299 (output_indirect_thunk): Likewise.
8300 (output_indirect_thunk_function): Likewise.
8301 (ix86_output_indirect_branch_via_reg): Likewise.
8302 (ix86_output_indirect_branch_via_push): Likewise.
8303 (ix86_output_indirect_branch): Likewise.
8304 (ix86_output_indirect_jmp): Likewise.
8305 (ix86_code_end): Call output_indirect_thunk_function if needed.
8306 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8308 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8309 (ix86_attribute_table): Add indirect_branch.
8310 * config/i386/i386.h (machine_function): Add indirect_branch_type
8311 and has_local_indirect_jump.
8312 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8314 (tablejump): Likewise.
8315 (*indirect_jump): Use ix86_output_indirect_jmp.
8316 (*tablejump_1): Likewise.
8317 (simple_return_indirect_internal): Likewise.
8318 * config/i386/i386.opt (mindirect-branch=): New option.
8319 (indirect_branch): New.
8322 (thunk-inline): Likewise.
8323 (thunk-extern): Likewise.
8324 * doc/extend.texi: Document indirect_branch function attribute.
8325 * doc/invoke.texi: Document -mindirect-branch= option.
8327 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8330 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8332 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8334 * ipa-inline.c (want_inline_small_function_p): Return false if
8335 inlining has already failed with CIF_FINAL_ERROR.
8336 (update_caller_keys): Call want_inline_small_function_p before
8338 (update_callee_keys): Likewise.
8340 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8342 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8344 (rs6000_quadword_masked_address_p): Likewise.
8345 (quad_aligned_load_p): Likewise.
8346 (quad_aligned_store_p): Likewise.
8347 (const_load_sequence_p): Add comment to describe the outer-most loop.
8348 (mimic_memory_attributes_and_flags): New function.
8349 (rs6000_gen_stvx): Likewise.
8350 (replace_swapped_aligned_store): Likewise.
8351 (rs6000_gen_lvx): Likewise.
8352 (replace_swapped_aligned_load): Likewise.
8353 (replace_swapped_load_constant): Capitalize argument name in
8354 comment describing this function.
8355 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8356 and stores that access quad-word aligned addresses and replace
8357 with stvx or lvx instructions when appropriate.
8358 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8359 New function prototype.
8360 (rs6000_quadword_masked_address_p): Likewise.
8361 (rs6000_gen_lvx): Likewise.
8362 (rs6000_gen_stvx): Likewise.
8363 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8364 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8365 when memory address is aligned.
8366 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8367 this split to select lvx instruction when memory address is aligned.
8368 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8369 instruction when memory address is aligned.
8370 (*vsx_le_perm_load_v16qi): Likewise.
8371 (four unnamed splitters): Modify to select the stvx instruction
8372 when memory is aligned.
8374 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8376 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8377 which appears in the queue multiple times.
8379 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8380 Alan Hayward <alan.hayward@arm.com>
8381 David Sherwood <david.sherwood@arm.com>
8383 * tree-vectorizer.h (vec_lower_bound): New structure.
8384 (_loop_vec_info): Add check_nonzero and lower_bounds.
8385 (LOOP_VINFO_CHECK_NONZERO): New macro.
8386 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8387 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8388 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8389 fields. Make seg_len the distance travelled, not including the
8391 (dr_direction_indicator): Declare.
8392 (dr_zero_step_indicator): Likewise.
8393 (dr_known_forward_stride_p): Likewise.
8394 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8396 (runtime_alias_check_p): Allow runtime alias checks with
8398 (operator ==): Compare access_size and align.
8399 (prune_runtime_alias_test_list): Rework for new distinction between
8400 the access_size and seg_len.
8401 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8403 (get_segment_min_max): New function.
8404 (create_intersect_range_checks): Use it.
8405 (dr_step_indicator): New function.
8406 (dr_direction_indicator): Likewise.
8407 (dr_zero_step_indicator): Likewise.
8408 (dr_known_forward_stride_p): Likewise.
8409 * tree-loop-distribution.c (data_ref_segment_size): Return
8410 DR_STEP * (niters - 1).
8411 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8413 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8414 (vect_preserves_scalar_order_p): New function, split out from...
8415 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8416 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8417 (vect_vfa_access_size): New function.
8418 (vect_vfa_align): Likewise.
8419 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8420 (dump_lower_bound): New function.
8421 (vect_check_lower_bound): Likewise.
8422 (vect_small_gap_p): Likewise.
8423 (vectorizable_with_step_bound_p): Likewise.
8424 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8425 depencies if the vectorization factor is 1. Convert the checks
8426 for nonzero steps into checks on the bounds of DR_STEP. Try using
8427 a bunds check for variable steps if the minimum required step is
8428 relatively small. Update calls to the dr_with_seg_len
8429 constructor and to vect_compile_time_alias.
8430 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8432 (vect_loop_versioning): Call it.
8433 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8435 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8437 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8438 Alan Hayward <alan.hayward@arm.com>
8439 David Sherwood <david.sherwood@arm.com>
8441 * doc/sourcebuild.texi (vect_scatter_store): Document.
8442 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8444 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8446 * genopinit.c (main): Add supports_vec_scatter_store and
8447 supports_vec_scatter_store_cached to target_optabs.
8448 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8449 IFN_MASK_SCATTER_STORE.
8450 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8452 * internal-fn.h (internal_store_fn_p): Declare.
8453 (internal_fn_stored_value_index): Likewise.
8454 * internal-fn.c (scatter_store_direct): New macro.
8455 (expand_scatter_store_optab_fn): New function.
8456 (direct_scatter_store_optab_supported_p): New macro.
8457 (internal_store_fn_p): New function.
8458 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8459 IFN_MASK_SCATTER_STORE.
8460 (internal_fn_mask_index): Likewise.
8461 (internal_fn_stored_value_index): New function.
8462 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8464 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8465 * optabs-query.c (supports_vec_scatter_store_p): New function.
8466 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8467 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8468 true for scatter stores.
8469 (vect_gather_scatter_fn_p): Handle scatter stores too.
8470 (vect_check_gather_scatter): Consider using scatter stores if
8471 supports_vec_scatter_store_p.
8472 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8474 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8475 internal_fn_stored_value_index.
8476 (check_load_store_masking): Handle scatter stores too.
8477 (vect_get_store_rhs): Make public.
8478 (vectorizable_call): Use internal_store_fn_p.
8479 (vectorizable_store): Handle scatter store internal functions.
8480 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8481 when deciding whether the end of the group has been reached.
8482 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8483 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8484 (mask_scatter_store<mode>): New insns.
8486 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8487 Alan Hayward <alan.hayward@arm.com>
8488 David Sherwood <david.sherwood@arm.com>
8490 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8491 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8492 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8494 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8495 Use vect_truncate_gather_scatter_offset if we can't treat the
8496 operation as a normal gather load or scatter store.
8497 (get_group_load_store_type): Take the gather_scatter_info
8498 as argument. Try using a gather load or scatter store for
8499 single-element groups.
8500 (get_load_store_type): Update calls to get_group_load_store_type
8501 and vect_use_strided_gather_scatters_p.
8503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8504 Alan Hayward <alan.hayward@arm.com>
8505 David Sherwood <david.sherwood@arm.com>
8507 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8508 optional tree argument.
8509 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8511 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8512 but continue to use the current value as a fallback.
8513 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8514 to compare the updates.
8515 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8516 (get_load_store_type): Use it when handling a strided access.
8517 (vect_get_strided_load_store_ops): New function.
8518 (vect_get_data_ptr_increment): Likewise.
8519 (vectorizable_load): Handle strided gather loads. Always pass
8520 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8523 Alan Hayward <alan.hayward@arm.com>
8524 David Sherwood <david.sherwood@arm.com>
8526 * doc/md.texi (gather_load@var{m}): Document.
8527 (mask_gather_load@var{m}): Likewise.
8528 * genopinit.c (main): Add supports_vec_gather_load and
8529 supports_vec_gather_load_cached to target_optabs.
8530 * optabs-tree.c (init_tree_optimization_optabs): Use
8531 ggc_cleared_alloc to allocate target_optabs.
8532 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8533 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8535 * internal-fn.h (internal_load_fn_p): Declare.
8536 (internal_gather_scatter_fn_p): Likewise.
8537 (internal_fn_mask_index): Likewise.
8538 (internal_gather_scatter_fn_supported_p): Likewise.
8539 * internal-fn.c (gather_load_direct): New macro.
8540 (expand_gather_load_optab_fn): New function.
8541 (direct_gather_load_optab_supported_p): New macro.
8542 (direct_internal_fn_optab): New function.
8543 (internal_load_fn_p): Likewise.
8544 (internal_gather_scatter_fn_p): Likewise.
8545 (internal_fn_mask_index): Likewise.
8546 (internal_gather_scatter_fn_supported_p): Likewise.
8547 * optabs-query.c (supports_at_least_one_mode_p): New function.
8548 (supports_vec_gather_load_p): Likewise.
8549 * optabs-query.h (supports_vec_gather_load_p): Declare.
8550 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8551 and memory_type field.
8552 (NUM_PATTERNS): Bump to 15.
8553 * tree-vect-data-refs.c: Include internal-fn.h.
8554 (vect_gather_scatter_fn_p): New function.
8555 (vect_describe_gather_scatter_call): Likewise.
8556 (vect_check_gather_scatter): Try using internal functions for
8557 gather loads. Recognize existing calls to a gather load function.
8558 (vect_analyze_data_refs): Consider using gather loads if
8559 supports_vec_gather_load_p.
8560 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8561 (vect_get_gather_scatter_offset_type): Likewise.
8562 (vect_convert_mask_for_vectype): Likewise.
8563 (vect_add_conversion_to_patterm): Likewise.
8564 (vect_try_gather_scatter_pattern): Likewise.
8565 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8566 (vect_vect_recog_func_ptrs): Add it.
8567 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8568 internal_fn_mask_index and internal_gather_scatter_fn_p.
8569 (check_load_store_masking): Take the gather_scatter_info as an
8570 argument and handle gather loads.
8571 (vect_get_gather_scatter_ops): New function.
8572 (vectorizable_call): Check internal_load_fn_p.
8573 (vectorizable_load): Likewise. Handle gather load internal
8575 (vectorizable_store): Update call to check_load_store_masking.
8576 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8577 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8578 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8579 (aarch64_gather_scale_operand_d): New predicates.
8580 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8581 (mask_gather_load<mode>): New insns.
8583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8584 Alan Hayward <alan.hayward@arm.com>
8585 David Sherwood <david.sherwood@arm.com>
8587 * optabs.def (fold_left_plus_optab): New optab.
8588 * doc/md.texi (fold_left_plus_@var{m}): Document.
8589 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8590 * internal-fn.c (fold_left_direct): Define.
8591 (expand_fold_left_optab_fn): Likewise.
8592 (direct_fold_left_optab_supported_p): Likewise.
8593 * fold-const-call.c (fold_const_fold_left): New function.
8594 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8595 * tree-parloops.c (valid_reduction_p): New function.
8596 (gather_scalar_reductions): Use it.
8597 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8598 (vect_finish_replace_stmt): Declare.
8599 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8600 (needs_fold_left_reduction_p): New function, split out from...
8601 (vect_is_simple_reduction): ...here. Accept reductions that
8602 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8603 (vect_force_simple_reduction): Also store the reduction type in
8604 the assignment's STMT_VINFO_REDUC_TYPE.
8605 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8606 (merge_with_identity): New function.
8607 (vect_expand_fold_left): Likewise.
8608 (vectorize_fold_left_reduction): Likewise.
8609 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8610 scalar phi in place for it. Check for target support and reject
8611 cases that would reassociate the operation. Defer the transform
8612 phase to vectorize_fold_left_reduction.
8613 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8614 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8615 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8617 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8619 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8620 call to ifc_temp_var.
8622 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8623 Alan Hayward <alan.hayward@arm.com>
8624 David Sherwood <david.sherwood@arm.com>
8626 * target.def (legitimize_address_displacement): Take the original
8627 offset as a poly_int.
8628 * targhooks.h (default_legitimize_address_displacement): Update
8630 * targhooks.c (default_legitimize_address_displacement): Likewise.
8631 * doc/tm.texi: Regenerate.
8632 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8633 as an argument, moving assert of ad->disp == ad->disp_term to...
8634 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8635 Try calling targetm.legitimize_address_displacement before expanding
8636 the address rather than afterwards, and adjust for the new interface.
8637 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8638 Match the new hook interface. Handle SVE addresses.
8639 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8642 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8644 * Makefile.in (OBJS): Add early-remat.o.
8645 * target.def (select_early_remat_modes): New hook.
8646 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8647 * doc/tm.texi: Regenerate.
8648 * targhooks.h (default_select_early_remat_modes): Declare.
8649 * targhooks.c (default_select_early_remat_modes): New function.
8650 * timevar.def (TV_EARLY_REMAT): New timevar.
8651 * passes.def (pass_early_remat): New pass.
8652 * tree-pass.h (make_pass_early_remat): Declare.
8653 * early-remat.c: New file.
8654 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8656 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8659 Alan Hayward <alan.hayward@arm.com>
8660 David Sherwood <david.sherwood@arm.com>
8662 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8663 vfm1 with a bound_epilog parameter.
8664 (vect_do_peeling): Update calls accordingly, and move the prologue
8665 call earlier in the function. Treat the base bound_epilog as 0 for
8666 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8667 this base when peeling for gaps.
8668 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8669 with fully-masked loops.
8670 (vect_estimate_min_profitable_iters): Handle the single peeled
8671 iteration in that case.
8673 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8674 Alan Hayward <alan.hayward@arm.com>
8675 David Sherwood <david.sherwood@arm.com>
8677 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8678 single-element interleaving even if the size is not a power of 2.
8679 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8680 accesses for single-element interleaving if the group size is
8683 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8684 Alan Hayward <alan.hayward@arm.com>
8685 David Sherwood <david.sherwood@arm.com>
8687 * doc/md.texi (fold_extract_last_@var{m}): Document.
8688 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8689 * optabs.def (fold_extract_last_optab): New optab.
8690 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8691 * internal-fn.c (fold_extract_direct): New macro.
8692 (expand_fold_extract_optab_fn): Likewise.
8693 (direct_fold_extract_optab_supported_p): Likewise.
8694 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8695 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8696 EXTRACT_LAST_REDUCTION.
8697 (get_initial_def_for_reduction): Do not create an initial vector
8698 for EXTRACT_LAST_REDUCTION reductions.
8699 (vectorizable_reduction): Leave the scalar phi in place for
8700 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8701 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8702 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8703 transform phase to vectorizable_condition.
8704 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8706 (vect_finish_stmt_generation): ...here.
8707 (vect_finish_replace_stmt): New function.
8708 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8709 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8711 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8714 Alan Hayward <alan.hayward@arm.com>
8715 David Sherwood <david.sherwood@arm.com>
8717 * doc/md.texi (extract_last_@var{m}): Document.
8718 * optabs.def (extract_last_optab): New optab.
8719 * internal-fn.def (EXTRACT_LAST): New internal function.
8720 * internal-fn.c (cond_unary_direct): New macro.
8721 (expand_cond_unary_optab_fn): Likewise.
8722 (direct_cond_unary_optab_supported_p): Likewise.
8723 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8724 loops using EXTRACT_LAST.
8725 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8726 (extract_last_<mode>): ...this optab.
8727 (vec_extract<mode><Vel>): Update accordingly.
8729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8730 Alan Hayward <alan.hayward@arm.com>
8731 David Sherwood <david.sherwood@arm.com>
8733 * target.def (empty_mask_is_expensive): New hook.
8734 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8735 * doc/tm.texi: Regenerate.
8736 * targhooks.h (default_empty_mask_is_expensive): Declare.
8737 * targhooks.c (default_empty_mask_is_expensive): New function.
8738 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8739 if the target says that empty masks are expensive.
8740 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8742 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8744 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8745 Alan Hayward <alan.hayward@arm.com>
8746 David Sherwood <david.sherwood@arm.com>
8748 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8749 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8750 (vect_use_loop_mask_for_alignment_p): New function.
8751 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8752 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8753 niters_skip argument. Make sure that the first niters_skip elements
8754 of the first iteration are inactive.
8755 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8756 Update call to vect_set_loop_masks_directly.
8757 (get_misalign_in_elems): New function, split out from...
8758 (vect_gen_prolog_loop_niters): ...here.
8759 (vect_update_init_of_dr): Take a code argument that specifies whether
8760 the adjustment should be added or subtracted.
8761 (vect_update_init_of_drs): Likewise.
8762 (vect_prepare_for_masked_peels): New function.
8763 (vect_do_peeling): Skip prologue peeling if we're using a mask
8764 instead. Update call to vect_update_inits_of_drs.
8765 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8767 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8768 alignment. Do not include the number of peeled iterations in
8769 the minimum threshold in that case.
8770 (vectorizable_induction): Adjust the start value down by
8771 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8772 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8773 Take the number of skipped iterations into account when calculating
8775 * tree-vect-stmts.c (vect_gen_while_not): New function.
8777 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8778 Alan Hayward <alan.hayward@arm.com>
8779 David Sherwood <david.sherwood@arm.com>
8781 * doc/sourcebuild.texi (vect_fully_masked): Document.
8782 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8784 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8786 (vect_analyze_loop_2): ...here. Don't check the vectorization
8787 factor against the number of loop iterations if the loop is
8790 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8791 Alan Hayward <alan.hayward@arm.com>
8792 David Sherwood <david.sherwood@arm.com>
8794 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8795 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8796 (dump_groups): Update accordingly.
8797 (iv_use::mem_type): New member variable.
8798 (address_p): New function.
8799 (record_use): Add a mem_type argument and initialize the new
8801 (record_group_use): Add a mem_type argument. Use address_p.
8802 Remove obsolete null checks of base_object. Update call to record_use.
8803 (find_interesting_uses_op): Update call to record_group_use.
8804 (find_interesting_uses_cond): Likewise.
8805 (find_interesting_uses_address): Likewise.
8806 (get_mem_type_for_internal_fn): New function.
8807 (find_address_like_use): Likewise.
8808 (find_interesting_uses_stmt): Try find_address_like_use before
8809 calling find_interesting_uses_op.
8810 (addr_offset_valid_p): Use the iv mem_type field as the type
8811 of the addressed memory.
8812 (add_autoinc_candidates): Likewise.
8813 (get_address_cost): Likewise.
8814 (split_small_address_groups_p): Use address_p.
8815 (split_address_groups): Likewise.
8816 (add_iv_candidate_for_use): Likewise.
8817 (autoinc_possible_for_pair): Likewise.
8818 (rewrite_groups): Likewise.
8819 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8820 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8821 (get_alias_ptr_type_for_ptr_address): New function.
8822 (rewrite_use_address): Rewrite address uses in calls that were
8823 identified by find_address_like_use.
8825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8826 Alan Hayward <alan.hayward@arm.com>
8827 David Sherwood <david.sherwood@arm.com>
8829 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8831 * gimple-expr.h (is_gimple_addressable: Likewise.
8832 * gimple-expr.c (is_gimple_address): Likewise.
8833 * internal-fn.c (expand_call_mem_ref): New function.
8834 (expand_mask_load_optab_fn): Use it.
8835 (expand_mask_store_optab_fn): Likewise.
8837 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8838 Alan Hayward <alan.hayward@arm.com>
8839 David Sherwood <david.sherwood@arm.com>
8841 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8842 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8843 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8844 (cond_umax@var{mode}): Document.
8845 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8846 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8847 (cond_umin_optab, cond_umax_optab): New optabs.
8848 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8849 (COND_IOR, COND_XOR): New internal functions.
8850 * internal-fn.h (get_conditional_internal_fn): Declare.
8851 * internal-fn.c (cond_binary_direct): New macro.
8852 (expand_cond_binary_optab_fn): Likewise.
8853 (direct_cond_binary_optab_supported_p): Likewise.
8854 (get_conditional_internal_fn): New function.
8855 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8856 Cope with reduction statements that are vectorized as calls rather
8858 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8859 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8860 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8861 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8862 (UNSPEC_COND_EOR): New unspecs.
8863 (optab): Add mappings for them.
8864 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8865 (sve_int_op, sve_fp_op): New int attributes.
8867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8868 Alan Hayward <alan.hayward@arm.com>
8869 David Sherwood <david.sherwood@arm.com>
8871 * optabs.def (while_ult_optab): New optab.
8872 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8873 * internal-fn.def (WHILE_ULT): New internal function.
8874 * internal-fn.h (direct_internal_fn_supported_p): New override
8875 that takes two types as argument.
8876 * internal-fn.c (while_direct): New macro.
8877 (expand_while_optab_fn): New function.
8878 (convert_optab_supported_p): Likewise.
8879 (direct_while_optab_supported_p): New macro.
8880 * wide-int.h (wi::udiv_ceil): New function.
8881 * tree-vectorizer.h (rgroup_masks): New structure.
8882 (vec_loop_masks): New typedef.
8883 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8885 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8886 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8887 (vect_max_vf): New function.
8888 (slpeel_make_loop_iterate_ntimes): Delete.
8889 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8890 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8891 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8892 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8893 internal-fn.h, stor-layout.h and optabs-query.h.
8894 (vect_set_loop_mask): New function.
8895 (add_preheader_seq): Likewise.
8896 (add_header_seq): Likewise.
8897 (interleave_supported_p): Likewise.
8898 (vect_maybe_permute_loop_masks): Likewise.
8899 (vect_set_loop_masks_directly): Likewise.
8900 (vect_set_loop_condition_masked): Likewise.
8901 (vect_set_loop_condition_unmasked): New function, split out from
8902 slpeel_make_loop_iterate_ntimes.
8903 (slpeel_make_loop_iterate_ntimes): Rename to..
8904 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8905 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8906 (vect_do_peeling): Update call accordingly.
8907 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8909 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8910 mask_compare_type, can_fully_mask_p and fully_masked_p.
8911 (release_vec_loop_masks): New function.
8912 (_loop_vec_info): Use it to free the loop masks.
8913 (can_produce_all_loop_masks_p): New function.
8914 (vect_get_max_nscalars_per_iter): Likewise.
8915 (vect_verify_full_masking): Likewise.
8916 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8917 retries, and free the mask rgroups before retrying. Check loop-wide
8918 reasons for disallowing fully-masked loops. Make the final decision
8919 about whether use a fully-masked loop or not.
8920 (vect_estimate_min_profitable_iters): Do not assume that peeling
8921 for the number of iterations will be needed for fully-masked loops.
8922 (vectorizable_reduction): Disable fully-masked loops.
8923 (vectorizable_live_operation): Likewise.
8924 (vect_halve_mask_nunits): New function.
8925 (vect_double_mask_nunits): Likewise.
8926 (vect_record_loop_mask): Likewise.
8927 (vect_get_loop_mask): Likewise.
8928 (vect_transform_loop): Handle the case in which the final loop
8929 iteration might handle a partial vector. Call vect_set_loop_condition
8930 instead of slpeel_make_loop_iterate_ntimes.
8931 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8932 (check_load_store_masking): New function.
8933 (prepare_load_store_mask): Likewise.
8934 (vectorizable_store): Handle fully-masked loops.
8935 (vectorizable_load): Likewise.
8936 (supportable_widening_operation): Use vect_halve_mask_nunits for
8938 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8939 (vect_gen_while): New function.
8940 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8941 (aarch64_uqdec<mode>): New insn.
8943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8944 Alan Hayward <alan.hayward@arm.com>
8945 David Sherwood <david.sherwood@arm.com>
8947 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8948 (reduc_xor_scal_optab): New optabs.
8949 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8950 (reduc_xor_scal_@var{m}): Document.
8951 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8952 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8954 * fold-const-call.c (fold_const_call): Handle them.
8955 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8956 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8957 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8958 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8959 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8960 (UNSPEC_XORV): New unspecs.
8961 (optab): Add entries for them.
8962 (BITWISEV): New int iterator.
8963 (bit_reduc_op): New int attributes.
8965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8966 Alan Hayward <alan.hayward@arm.com>
8967 David Sherwood <david.sherwood@arm.com>
8969 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8970 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8971 * optabs.def (vec_shl_insert_optab): New optab.
8972 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8973 (duplicate_and_interleave): Likewise.
8974 * tree-vect-loop.c: Include internal-fn.h.
8975 (neutral_op_for_slp_reduction): New function, split out from
8976 get_initial_defs_for_reduction.
8977 (get_initial_def_for_reduction): Handle option 2 for variable-length
8978 vectors by loading the neutral value into a vector and then shifting
8979 the initial value into element 0.
8980 (get_initial_defs_for_reduction): Replace the code argument with
8981 the neutral value calculated by neutral_op_for_slp_reduction.
8982 Use gimple_build_vector for constant-length vectors.
8983 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8984 but the first group_size elements have a neutral value.
8985 Use duplicate_and_interleave otherwise.
8986 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8987 Update call to get_initial_defs_for_reduction. Handle SLP
8988 reductions for variable-length vectors by creating one vector
8989 result for each scalar result, with the elements associated
8990 with other scalar results stubbed out with the neutral value.
8991 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8992 Require IFN_VEC_SHL_INSERT for double reductions on
8993 variable-length vectors, or SLP reductions that have
8994 a neutral value. Require can_duplicate_and_interleave_p
8995 support for variable-length unchained SLP reductions if there
8996 is no neutral value, such as for MIN/MAX reductions. Also require
8997 the number of vector elements to be a multiple of the number of
8998 SLP statements when doing variable-length unchained SLP reductions.
8999 Update call to vect_create_epilog_for_reduction.
9000 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9001 and remove initial values.
9002 (duplicate_and_interleave): Make public.
9003 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9004 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9006 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9007 Alan Hayward <alan.hayward@arm.com>
9008 David Sherwood <david.sherwood@arm.com>
9010 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9011 (can_duplicate_and_interleave_p): New function.
9012 (vect_get_and_check_slp_defs): Take the vector of statements
9013 rather than just the current one. Remove excess parentheses.
9014 Restriction rejectinon of vect_constant_def and vect_external_def
9015 for variable-length vectors to boolean types, or types for which
9016 can_duplicate_and_interleave_p is false.
9017 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9018 (duplicate_and_interleave): New function.
9019 (vect_get_constant_vectors): Use gimple_build_vector for
9020 constant-length vectors and suitable variable-length constant
9021 vectors. Use duplicate_and_interleave for other variable-length
9022 vectors. Don't defer the update when inserting new statements.
9024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9025 Alan Hayward <alan.hayward@arm.com>
9026 David Sherwood <david.sherwood@arm.com>
9028 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9029 min_profitable_iters doesn't go negative.
9031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9032 Alan Hayward <alan.hayward@arm.com>
9033 David Sherwood <david.sherwood@arm.com>
9035 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9036 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9037 * optabs.def (vec_mask_load_lanes_optab): New optab.
9038 (vec_mask_store_lanes_optab): Likewise.
9039 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9040 (MASK_STORE_LANES): Likewise.
9041 * internal-fn.c (mask_load_lanes_direct): New macro.
9042 (mask_store_lanes_direct): Likewise.
9043 (expand_mask_load_optab_fn): Handle masked operations.
9044 (expand_mask_load_lanes_optab_fn): New macro.
9045 (expand_mask_store_optab_fn): Handle masked operations.
9046 (expand_mask_store_lanes_optab_fn): New macro.
9047 (direct_mask_load_lanes_optab_supported_p): Likewise.
9048 (direct_mask_store_lanes_optab_supported_p): Likewise.
9049 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9051 (vect_load_lanes_supported): Likewise.
9052 * tree-vect-data-refs.c (strip_conversion): New function.
9053 (can_group_stmts_p): Likewise.
9054 (vect_analyze_data_ref_accesses): Use it instead of checking
9055 for a pair of assignments.
9056 (vect_store_lanes_supported): Take a masked_p parameter.
9057 (vect_load_lanes_supported): Likewise.
9058 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9059 vect_store_lanes_supported and vect_load_lanes_supported.
9060 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9061 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9062 parameter. Don't allow gaps for masked accesses.
9063 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9064 and vect_load_lanes_supported.
9065 (get_load_store_type): Take a masked_p parameter and update
9066 call to get_group_load_store_type.
9067 (vectorizable_store): Update call to get_load_store_type.
9068 Handle IFN_MASK_STORE_LANES.
9069 (vectorizable_load): Update call to get_load_store_type.
9070 Handle IFN_MASK_LOAD_LANES.
9072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9073 Alan Hayward <alan.hayward@arm.com>
9074 David Sherwood <david.sherwood@arm.com>
9076 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9078 * config/aarch64/aarch64-protos.h
9079 (aarch64_sve_struct_memory_operand_p): Declare.
9080 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9081 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9082 (VPRED, vpred): Handle SVE structure modes.
9083 * config/aarch64/constraints.md (Utx): New constraint.
9084 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9085 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9086 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9087 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9088 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9089 structure modes. Split into pieces after RA.
9090 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9091 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9093 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9094 SVE structure modes.
9095 (aarch64_classify_address): Likewise.
9096 (sizetochar): Move earlier in file.
9097 (aarch64_print_operand): Handle SVE register lists.
9098 (aarch64_array_mode): New function.
9099 (aarch64_sve_struct_memory_operand_p): Likewise.
9100 (TARGET_ARRAY_MODE): Redefine.
9102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9103 Alan Hayward <alan.hayward@arm.com>
9104 David Sherwood <david.sherwood@arm.com>
9106 * target.def (array_mode): New target hook.
9107 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9108 * doc/tm.texi: Regenerate.
9109 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9110 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9111 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9113 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9116 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9117 Alan Hayward <alan.hayward@arm.com>
9118 David Sherwood <david.sherwood@arm.com>
9120 * fold-const.c (fold_binary_loc): Check the argument types
9121 rather than the result type when testing for a vector operation.
9123 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9125 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9126 * doc/tm.texi: Regenerate.
9128 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9129 Alan Hayward <alan.hayward@arm.com>
9130 David Sherwood <david.sherwood@arm.com>
9132 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9133 (sve): Document new AArch64 extension.
9134 * doc/md.texi (w): Extend the description of the AArch64
9135 constraint to include SVE vectors.
9136 (Upl, Upa): Document new AArch64 predicate constraints.
9137 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9139 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9140 (msve-vector-bits=): New option.
9141 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9142 SVE when these are disabled.
9143 (sve): New extension.
9144 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9145 modes. Adjust their number of units based on aarch64_sve_vg.
9146 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9147 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9148 aarch64_addr_query_type.
9149 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9150 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9151 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9152 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9153 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9154 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9155 (aarch64_simd_imm_zero_p): Delete.
9156 (aarch64_check_zero_based_sve_index_immediate): Declare.
9157 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9158 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9159 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9160 (aarch64_sve_float_mul_immediate_p): Likewise.
9161 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9163 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9164 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9165 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9166 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9167 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9168 (aarch64_regmode_natural_size): Likewise.
9169 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9170 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9172 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9173 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9174 for VG and the SVE predicate registers.
9175 (V_ALIASES): Add a "z"-prefixed alias.
9176 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9177 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9178 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9179 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9180 (REG_CLASS_NAMES): Add entries for them.
9181 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9182 and the predicate registers.
9183 (aarch64_sve_vg): Declare.
9184 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9185 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9186 (REGMODE_NATURAL_SIZE): Define.
9187 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9189 * config/aarch64/aarch64.c: Include cfgrtl.h.
9190 (simd_immediate_info): Add a constructor for series vectors,
9191 and an associated step field.
9192 (aarch64_sve_vg): New variable.
9193 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9194 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9195 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9196 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9197 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9198 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9199 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9200 (aarch64_get_mask_mode): New functions.
9201 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9202 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9203 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9204 predicate modes and predicate registers. Explicitly restrict
9205 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9206 to store a vector mode if it is recognized by
9207 aarch64_classify_vector_mode.
9208 (aarch64_regmode_natural_size): New function.
9209 (aarch64_hard_regno_caller_save_mode): Return the original mode
9211 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9212 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9213 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9214 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9216 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9217 does not overlap dest if the function is frame-related. Handle
9219 (aarch64_split_add_offset): New function.
9220 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9221 them aarch64_add_offset.
9222 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9223 and update call to aarch64_sub_sp.
9224 (aarch64_add_cfa_expression): New function.
9225 (aarch64_expand_prologue): Pass extra temporary registers to the
9226 functions above. Handle the case in which we need to emit new
9227 DW_CFA_expressions for registers that were originally saved
9228 relative to the stack pointer, but now have to be expressed
9229 relative to the frame pointer.
9230 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9232 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9233 IP0 and IP1 values for SVE frames.
9234 (aarch64_expand_vec_series): New function.
9235 (aarch64_expand_sve_widened_duplicate): Likewise.
9236 (aarch64_expand_sve_const_vector): Likewise.
9237 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9238 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9239 into the register, rather than emitting a SET directly.
9240 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9241 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9242 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9243 (offset_9bit_signed_scaled_p): New functions.
9244 (aarch64_replicate_bitmask_imm): New function.
9245 (aarch64_bitmask_imm): Use it.
9246 (aarch64_cannot_force_const_mem): Reject expressions involving
9247 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9248 (aarch64_classify_index): Handle SVE indices, by requiring
9249 a plain register index with a scale that matches the element size.
9250 (aarch64_classify_address): Handle SVE addresses. Assert that
9251 the mode of the address is VOIDmode or an integer mode.
9252 Update call to aarch64_classify_symbol.
9253 (aarch64_classify_symbolic_expression): Update call to
9254 aarch64_classify_symbol.
9255 (aarch64_const_vec_all_in_range_p): New function.
9256 (aarch64_print_vector_float_operand): Likewise.
9257 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9258 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9259 and the FP immediates 1.0 and 0.5.
9260 (aarch64_print_address_internal): Handle SVE addresses.
9261 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9262 (aarch64_regno_regclass): Handle predicate registers.
9263 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9265 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9266 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9267 (aarch64_convert_sve_vector_bits): New function.
9268 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9269 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9271 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9272 Handle SVE vector and predicate modes. Accept VL-based constants
9273 that need only one temporary register, and VL offsets that require
9274 no temporary registers.
9275 (aarch64_conditional_register_usage): Mark the predicate registers
9276 as fixed if SVE isn't available.
9277 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9278 Return true for SVE vector and predicate modes.
9279 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9280 rather than an unsigned int. Handle SVE modes.
9281 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9283 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9285 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9286 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9287 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9288 (aarch64_sve_float_mul_immediate_p): New functions.
9289 (aarch64_sve_valid_immediate): New function.
9290 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9291 Explicitly reject structure modes. Check for INDEX constants.
9292 Handle PTRUE and PFALSE constants.
9293 (aarch64_check_zero_based_sve_index_immediate): New function.
9294 (aarch64_simd_imm_zero_p): Delete.
9295 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9296 vector modes. Accept constants in the range of CNT[BHWD].
9297 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9298 ask for an Advanced SIMD mode.
9299 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9300 (aarch64_simd_vector_alignment): Handle SVE predicates.
9301 (aarch64_vectorize_preferred_vector_alignment): New function.
9302 (aarch64_simd_vector_alignment_reachable): Use it instead of
9304 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9305 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9307 (MAX_VECT_LEN): Delete.
9308 (expand_vec_perm_d): Add a vec_flags field.
9309 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9310 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9311 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9313 (aarch64_evpc_rev): Rename to...
9314 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9315 (aarch64_evpc_rev_global): New function.
9316 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9317 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9319 (aarch64_evpc_sve_tbl): New function.
9320 (aarch64_expand_vec_perm_const_1): Update after rename of
9321 aarch64_evpc_rev. Handle SVE permutes too, trying
9322 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9323 than aarch64_evpc_tbl.
9324 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9325 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9326 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9327 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9328 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9329 (aarch64_expand_sve_vcond): New functions.
9330 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9331 of aarch64_vector_mode_p.
9332 (aarch64_dwarf_poly_indeterminate_value): New function.
9333 (aarch64_compute_pressure_classes): Likewise.
9334 (aarch64_can_change_mode_class): Likewise.
9335 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9336 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9337 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9338 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9339 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9340 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9341 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9342 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9344 (Dn, Dl, Dr): Accept const as well as const_vector.
9345 (Dz): Likewise. Compare against CONST0_RTX.
9346 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9347 of "vector" where appropriate.
9348 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9349 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9350 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9351 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9352 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9353 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9354 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9355 (v_int_equiv): Extend to SVE modes.
9356 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9358 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9359 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9360 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9361 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9362 (SVE_COND_FP_CMP): New int iterators.
9363 (perm_hilo): Handle the new unpack unspecs.
9364 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9366 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9367 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9368 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9369 (aarch64_equality_operator, aarch64_constant_vector_operand)
9370 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9371 (aarch64_sve_nonimmediate_operand): Likewise.
9372 (aarch64_sve_general_operand): Likewise.
9373 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9374 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9375 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9376 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9377 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9378 (aarch64_sve_float_arith_immediate): Likewise.
9379 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9380 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9381 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9382 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9383 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9384 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9385 (aarch64_sve_float_arith_operand): Likewise.
9386 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9387 (aarch64_sve_float_mul_operand): Likewise.
9388 (aarch64_sve_vec_perm_operand): Likewise.
9389 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9390 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9391 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9392 as well as const_vector.
9393 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9394 in file. Use CONST0_RTX and CONSTM1_RTX.
9395 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9396 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9397 Use aarch64_simd_imm_zero.
9398 * config/aarch64/aarch64-sve.md: New file.
9399 * config/aarch64/aarch64.md: Include it.
9400 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9401 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9402 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9403 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9404 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9405 (sve): New attribute.
9406 (enabled): Disable instructions with the sve attribute unless
9408 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9409 aarch64_expand_mov_immediate.
9410 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9411 CNT[BHSD] immediates.
9412 (movti): Split CONST_POLY_INT moves into two halves.
9413 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9414 Split additions that need a temporary here if the destination
9415 is the stack pointer.
9416 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9417 (*add<mode>3_poly_1): New instruction.
9418 (set_clobber_cc): New expander.
9420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9422 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9423 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9424 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9425 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9426 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9427 Change innermode from fixed_mode_size to machine_mode.
9428 (simplify_subreg): Update call accordingly. Handle a constant-sized
9429 subreg of a variable-length CONST_VECTOR.
9431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9432 Alan Hayward <alan.hayward@arm.com>
9433 David Sherwood <david.sherwood@arm.com>
9435 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9436 (add_offset_to_base): New function, split out from...
9437 (create_mem_ref): ...here. When handling a scale other than 1,
9438 check first whether the address is valid without the offset.
9439 Add it into the base if so, leaving the index and scale as-is.
9441 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9444 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9445 fold_for_warn before checking if arg2 is INTEGER_CST.
9447 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9449 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9450 (store_multiple_operation): Delete.
9451 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9452 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9453 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9454 guarded by TARGET_STRING.
9455 (rs6000_output_load_multiple): Delete.
9456 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9457 OPTION_MASK_STRING / TARGET_STRING handling.
9458 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9459 (const rs6000_opt_masks) <"string">: Change mask to 0.
9460 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9461 (MASK_STRING): Delete.
9462 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9464 (load_multiple): Delete.
9471 (store_multiple): Delete.
9478 (movmemsi_8reg): Delete.
9479 (corresponding unnamed define_insn): Delete.
9480 (movmemsi_6reg): Delete.
9481 (corresponding unnamed define_insn): Delete.
9482 (movmemsi_4reg): Delete.
9483 (corresponding unnamed define_insn): Delete.
9484 (movmemsi_2reg): Delete.
9485 (corresponding unnamed define_insn): Delete.
9486 (movmemsi_1reg): Delete.
9487 (corresponding unnamed define_insn): Delete.
9488 * config/rs6000/rs6000.opt (mno-string): New.
9489 (mstring): Replace by deprecation warning stub.
9490 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9492 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9494 * regrename.c (regrename_do_replace): If replacing the same
9495 reg multiple times, try to reuse last created gen_raw_REG.
9498 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9499 main to workaround a bug in GDB.
9501 2018-01-12 Tom de Vries <tom@codesourcery.com>
9504 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9506 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9508 PR rtl-optimization/80481
9509 * ira-color.c (get_cap_member): New function.
9510 (allocnos_conflict_by_live_ranges_p): Use it.
9511 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9512 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9514 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9517 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9518 (*saddl_se_1): Ditto.
9520 (*ssubl_se_1): Ditto.
9522 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9524 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9525 rather than wi::to_widest for DR_INITs.
9526 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9527 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9528 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9530 (vect_analyze_group_access_1): Note that here.
9532 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9534 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9535 polynomial type sizes.
9537 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9539 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9540 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9541 (gimple_add_tmp_var): Likewise.
9543 2018-01-12 Martin Liska <mliska@suse.cz>
9545 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9546 (gimple_alloc_sizes): Likewise.
9547 (dump_gimple_statistics): Use PRIu64 in printf format.
9548 * gimple.h: Change uint64_t to int.
9550 2018-01-12 Martin Liska <mliska@suse.cz>
9552 * tree-core.h: Use uint64_t instead of int.
9553 * tree.c (tree_node_counts): Likewise.
9554 (tree_node_sizes): Likewise.
9555 (dump_tree_statistics): Use PRIu64 in printf format.
9557 2018-01-12 Martin Liska <mliska@suse.cz>
9559 * Makefile.in: As qsort_chk is implemented in vec.c, add
9560 vec.o to linkage of gencfn-macros.
9561 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9562 passing the info to record_node_allocation_statistics.
9563 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9565 * ggc-common.c (struct ggc_usage): Add operator== and use
9566 it in operator< and compare function.
9567 * mem-stats.h (struct mem_usage): Likewise.
9568 * vec.c (struct vec_usage): Remove operator< and compare
9569 function. Can be simply inherited.
9571 2018-01-12 Martin Jambor <mjambor@suse.cz>
9574 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9575 * tree-ssa-math-opts.c: Include domwalk.h.
9576 (convert_mult_to_fma_1): New function.
9577 (fma_transformation_info): New type.
9578 (fma_deferring_state): Likewise.
9579 (cancel_fma_deferring): New function.
9580 (result_of_phi): Likewise.
9581 (last_fma_candidate_feeds_initial_phi): Likewise.
9582 (convert_mult_to_fma): Added deferring logic, split actual
9583 transformation to convert_mult_to_fma_1.
9584 (math_opts_dom_walker): New type.
9585 (math_opts_dom_walker::after_dom_children): New method, body moved
9586 here from pass_optimize_widening_mul::execute, added deferring logic
9588 (pass_optimize_widening_mul::execute): Moved most of code to
9589 math_opts_dom_walker::after_dom_children.
9590 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9591 * config/i386/i386.c (ix86_option_override_internal): Added
9592 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9594 2018-01-12 Richard Biener <rguenther@suse.de>
9597 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9598 inline instance vars.
9600 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9603 * config/rx/rx.c (rx_is_restricted_memory_address):
9606 2018-01-12 Richard Biener <rguenther@suse.de>
9608 PR tree-optimization/80846
9609 * target.def (split_reduction): New target hook.
9610 * targhooks.c (default_split_reduction): New function.
9611 * targhooks.h (default_split_reduction): Declare.
9612 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9613 target requests first reduce vectors by combining low and high
9615 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9616 (get_vectype_for_scalar_type_and_size): Export.
9617 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9618 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9619 * doc/tm.texi: Regenerate.
9620 * config/i386/i386.c (ix86_split_reduction): Implement
9621 TARGET_VECTORIZE_SPLIT_REDUCTION.
9623 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9626 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9627 in PIC mode except for TARGET_VXWORKS_RTP.
9628 * config/sparc/sparc.c: Include cfgrtl.h.
9629 (TARGET_INIT_PIC_REG): Define.
9630 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9631 (sparc_pic_register_p): New predicate.
9632 (sparc_legitimate_address_p): Use it.
9633 (sparc_legitimize_pic_address): Likewise.
9634 (sparc_delegitimize_address): Likewise.
9635 (sparc_mode_dependent_address_p): Likewise.
9636 (gen_load_pcrel_sym): Remove 4th parameter.
9637 (load_got_register): Adjust call to above. Remove obsolete stuff.
9638 (sparc_expand_prologue): Do not call load_got_register here.
9639 (sparc_flat_expand_prologue): Likewise.
9640 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9641 (sparc_use_pseudo_pic_reg): New function.
9642 (sparc_init_pic_reg): Likewise.
9643 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9644 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9646 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9648 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9649 Add item for branch_cost.
9651 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9653 PR rtl-optimization/83565
9654 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9655 not extend the result to a larger mode for rotate operations.
9656 (num_sign_bit_copies1): Likewise.
9658 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9661 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9663 Use values-Xc.o for -pedantic.
9664 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9666 2018-01-12 Martin Liska <mliska@suse.cz>
9669 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9671 (possible_polymorphic_call_targets): Use it.
9672 (ipa_devirt): Likewise.
9674 2018-01-12 Martin Liska <mliska@suse.cz>
9676 * profile-count.h (enum profile_quality): Use 0 as invalid
9677 enum value of profile_quality.
9679 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9681 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9682 -mext-string options.
9684 2018-01-12 Richard Biener <rguenther@suse.de>
9686 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9687 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9688 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9690 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9692 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9694 * configure.ac (--with-long-double-format): Add support for the
9695 configuration option to change the default long double format on
9697 * config.gcc (powerpc*-linux*-*): Likewise.
9698 * configure: Regenerate.
9699 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9700 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9701 used without modification.
9703 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9705 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9706 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9707 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9708 MISC_BUILTIN_SPEC_BARRIER.
9709 (rs6000_init_builtins): Likewise.
9710 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9712 (speculation_barrier): New define_insn.
9713 * doc/extend.texi: Document __builtin_speculation_barrier.
9715 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9718 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9719 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9720 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9722 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9723 integral modes instead of "ss" and "sd".
9724 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9725 vectors with 32-bit and 64-bit elements.
9726 (vecdupssescalarmodesuffix): New mode attribute.
9727 (vec_dup<mode>): Use it.
9729 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9732 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9733 frame if argument is passed on stack.
9735 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9738 * ree.c (combine_reaching_defs): Optimize also
9739 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9740 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9742 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9745 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9747 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9750 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9751 after they are computed.
9753 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9755 PR tree-optimization/83695
9756 * gimple-loop-linterchange.cc
9757 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9758 reset cached scev information after interchange.
9759 (pass_linterchange::execute): Remove call to scev_reset_htab.
9761 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9763 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9764 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9765 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9766 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9767 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9768 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9769 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9770 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9771 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9772 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9773 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9774 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9775 (V_lane_reg): Likewise.
9776 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9778 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9779 (vfmal_lane_low<mode>_intrinsic,
9780 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9781 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9782 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9783 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9784 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9785 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9787 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9789 * config/arm/arm-cpus.in (fp16fml): New feature.
9790 (ALL_SIMD): Add fp16fml.
9791 (armv8.2-a): Add fp16fml as an option.
9792 (armv8.3-a): Likewise.
9793 (armv8.4-a): Add fp16fml as part of fp16.
9794 * config/arm/arm.h (TARGET_FP16FML): Define.
9795 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9797 * config/arm/arm-modes.def (V2HF): Define.
9798 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9799 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9800 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9801 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9802 vfmsl_low, vfmsl_high): New set of builtins.
9803 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9804 (vfml_op): New code attribute.
9805 (VFMLHALVES): New int iterator.
9806 (VFML, VFMLSEL): New mode attributes.
9807 (V_reg): Define mapping for V2HF.
9808 (V_hi, V_lo): New mode attributes.
9809 (VF_constraint): Likewise.
9810 (vfml_half, vfml_half_selector): New int attributes.
9811 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9813 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9814 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9816 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9817 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9818 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9819 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9821 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9822 Document new effective target and option set.
9824 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9826 * config/arm/arm-cpus.in (armv8_4): New feature.
9827 (ARMv8_4a): New fgroup.
9828 (armv8.4-a): New arch.
9829 * config/arm/arm-tables.opt: Regenerate.
9830 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9831 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9832 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9833 Add matching rules for -march=armv8.4-a and extensions.
9834 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9836 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9839 * config/rx/rx.md (BW): New mode attribute.
9840 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9842 2018-01-11 Richard Biener <rguenther@suse.de>
9844 PR tree-optimization/83435
9845 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9846 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9847 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9849 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9850 Alan Hayward <alan.hayward@arm.com>
9851 David Sherwood <david.sherwood@arm.com>
9853 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9855 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9856 (aarch64_print_address_internal): Use it to check for a zero offset.
9858 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9859 Alan Hayward <alan.hayward@arm.com>
9860 David Sherwood <david.sherwood@arm.com>
9862 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9863 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9864 Return a poly_int64 rather than a HOST_WIDE_INT.
9865 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9866 rather than a HOST_WIDE_INT.
9867 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9868 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9869 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9870 final_offset from HOST_WIDE_INT to poly_int64.
9871 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9872 to_constant when getting the number of units in an Advanced SIMD
9874 (aarch64_builtin_vectorized_function): Check for a constant number
9876 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9878 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9879 attribute instead of GET_MODE_NUNITS.
9880 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9881 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9882 GET_MODE_SIZE for fixed-size registers.
9883 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9884 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9885 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9886 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9887 (aarch64_print_operand, aarch64_print_address_internal)
9888 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9889 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9890 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9891 Handle polynomial GET_MODE_SIZE.
9892 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9893 wider than SImode without modification.
9894 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9895 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9896 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9897 passing and returning SVE modes.
9898 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9899 rather than GEN_INT.
9900 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9901 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9902 (aarch64_allocate_and_probe_stack_space): Likewise.
9903 (aarch64_layout_frame): Cope with polynomial offsets.
9904 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9905 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9907 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9908 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9909 poly_int64 rather than a HOST_WIDE_INT.
9910 (aarch64_get_separate_components, aarch64_process_components)
9911 (aarch64_expand_prologue, aarch64_expand_epilogue)
9912 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9913 (aarch64_anchor_offset): New function, split out from...
9914 (aarch64_legitimize_address): ...here.
9915 (aarch64_builtin_vectorization_cost): Handle polynomial
9916 TYPE_VECTOR_SUBPARTS.
9917 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9919 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9920 number of elements from the PARALLEL rather than the mode.
9921 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9922 rather than GET_MODE_BITSIZE.
9923 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9924 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9925 (aarch64_expand_vec_perm_const_1): Handle polynomial
9926 d->perm.length () and d->perm elements.
9927 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9928 Apply to_constant to d->perm elements.
9929 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9930 polynomial CONST_VECTOR_NUNITS.
9931 (aarch64_move_pointer): Take amount as a poly_int64 rather
9933 (aarch64_progress_pointer): Avoid temporary variable.
9934 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9935 the mode attribute instead of GET_MODE.
9937 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9938 Alan Hayward <alan.hayward@arm.com>
9939 David Sherwood <david.sherwood@arm.com>
9941 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9942 x exists before using it.
9943 (aarch64_add_constant_internal): Rename to...
9944 (aarch64_add_offset_1): ...this. Replace regnum with separate
9945 src and dest rtxes. Handle the case in which they're different,
9946 including when the offset is zero. Replace scratchreg with an rtx.
9947 Use 2 additions if there is no spare register into which we can
9948 move a 16-bit constant.
9949 (aarch64_add_constant): Delete.
9950 (aarch64_add_offset): Replace reg with separate src and dest
9951 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9952 Use aarch64_add_offset_1.
9953 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9954 an rtx rather than an int. Take the delta as a poly_int64
9955 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9956 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9957 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9958 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9959 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9961 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9962 aarch64_add_constant.
9964 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9966 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9967 Use scalar_float_mode.
9969 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9971 * config/aarch64/aarch64-simd.md
9972 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9973 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9974 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9975 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9976 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9977 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9978 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9979 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9980 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9981 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9983 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9986 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9987 targ_options->x_arm_arch_string is non NULL.
9989 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9991 * config/aarch64/aarch64.h
9992 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9994 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9997 * expmed.c (emit_store_flag_force): Swap if const op0
9998 and change VOIDmode to mode of op0.
10000 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10002 PR rtl-optimization/83761
10003 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10004 than bytes to mode_for_size.
10006 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10008 PR middle-end/83189
10009 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10010 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10013 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10015 PR middle-end/83575
10016 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10017 when in layout mode.
10018 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10019 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10022 2018-01-10 Michael Collison <michael.collison@arm.com>
10024 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10025 * config/aarch64/aarch64-option-extension.def: Add
10026 AARCH64_OPT_EXTENSION of 'fp16fml'.
10027 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10028 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10029 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10030 * config/aarch64/constraints.md (Ui7): New constraint.
10031 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10032 (VFMLA_SEL_W): Ditto.
10035 (VFMLA16_LOW): New int iterator.
10036 (VFMLA16_HIGH): Ditto.
10037 (UNSPEC_FMLAL): New unspec.
10038 (UNSPEC_FMLSL): Ditto.
10039 (UNSPEC_FMLAL2): Ditto.
10040 (UNSPEC_FMLSL2): Ditto.
10041 (f16mac): New code attribute.
10042 * config/aarch64/aarch64-simd-builtins.def
10043 (aarch64_fmlal_lowv2sf): Ditto.
10044 (aarch64_fmlsl_lowv2sf): Ditto.
10045 (aarch64_fmlalq_lowv4sf): Ditto.
10046 (aarch64_fmlslq_lowv4sf): Ditto.
10047 (aarch64_fmlal_highv2sf): Ditto.
10048 (aarch64_fmlsl_highv2sf): Ditto.
10049 (aarch64_fmlalq_highv4sf): Ditto.
10050 (aarch64_fmlslq_highv4sf): Ditto.
10051 (aarch64_fmlal_lane_lowv2sf): Ditto.
10052 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10053 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10054 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10055 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10056 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10057 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10058 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10059 (aarch64_fmlal_lane_highv2sf): Ditto.
10060 (aarch64_fmlsl_lane_highv2sf): Ditto.
10061 (aarch64_fmlal_laneq_highv2sf): Ditto.
10062 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10063 (aarch64_fmlalq_lane_highv4sf): Ditto.
10064 (aarch64_fmlsl_lane_highv4sf): Ditto.
10065 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10066 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10067 * config/aarch64/aarch64-simd.md:
10068 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10069 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10070 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10071 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10072 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10073 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10074 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10075 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10076 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10077 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10078 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10079 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10080 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10081 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10082 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10083 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10084 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10085 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10086 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10087 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10088 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10089 (vfmlsl_low_u32): Ditto.
10090 (vfmlalq_low_u32): Ditto.
10091 (vfmlslq_low_u32): Ditto.
10092 (vfmlal_high_u32): Ditto.
10093 (vfmlsl_high_u32): Ditto.
10094 (vfmlalq_high_u32): Ditto.
10095 (vfmlslq_high_u32): Ditto.
10096 (vfmlal_lane_low_u32): Ditto.
10097 (vfmlsl_lane_low_u32): Ditto.
10098 (vfmlal_laneq_low_u32): Ditto.
10099 (vfmlsl_laneq_low_u32): Ditto.
10100 (vfmlalq_lane_low_u32): Ditto.
10101 (vfmlslq_lane_low_u32): Ditto.
10102 (vfmlalq_laneq_low_u32): Ditto.
10103 (vfmlslq_laneq_low_u32): Ditto.
10104 (vfmlal_lane_high_u32): Ditto.
10105 (vfmlsl_lane_high_u32): Ditto.
10106 (vfmlal_laneq_high_u32): Ditto.
10107 (vfmlsl_laneq_high_u32): Ditto.
10108 (vfmlalq_lane_high_u32): Ditto.
10109 (vfmlslq_lane_high_u32): Ditto.
10110 (vfmlalq_laneq_high_u32): Ditto.
10111 (vfmlslq_laneq_high_u32): Ditto.
10112 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10113 (AARCH64_FL_FOR_ARCH8_4): New.
10114 (AARCH64_ISA_F16FML): New ISA flag.
10115 (TARGET_F16FML): New feature flag for fp16fml.
10116 (doc/invoke.texi): Document new fp16fml option.
10118 2018-01-10 Michael Collison <michael.collison@arm.com>
10120 * config/aarch64/aarch64-builtins.c:
10121 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10122 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10123 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10124 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10125 (AARCH64_ISA_SHA3): New ISA flag.
10126 (TARGET_SHA3): New feature flag for sha3.
10127 * config/aarch64/iterators.md (sha512_op): New int attribute.
10128 (CRYPTO_SHA512): New int iterator.
10129 (UNSPEC_SHA512H): New unspec.
10130 (UNSPEC_SHA512H2): Ditto.
10131 (UNSPEC_SHA512SU0): Ditto.
10132 (UNSPEC_SHA512SU1): Ditto.
10133 * config/aarch64/aarch64-simd-builtins.def
10134 (aarch64_crypto_sha512hqv2di): New builtin.
10135 (aarch64_crypto_sha512h2qv2di): Ditto.
10136 (aarch64_crypto_sha512su0qv2di): Ditto.
10137 (aarch64_crypto_sha512su1qv2di): Ditto.
10138 (aarch64_eor3qv8hi): Ditto.
10139 (aarch64_rax1qv2di): Ditto.
10140 (aarch64_xarqv2di): Ditto.
10141 (aarch64_bcaxqv8hi): Ditto.
10142 * config/aarch64/aarch64-simd.md:
10143 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10144 (aarch64_crypto_sha512su0qv2di): Ditto.
10145 (aarch64_crypto_sha512su1qv2di): Ditto.
10146 (aarch64_eor3qv8hi): Ditto.
10147 (aarch64_rax1qv2di): Ditto.
10148 (aarch64_xarqv2di): Ditto.
10149 (aarch64_bcaxqv8hi): Ditto.
10150 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10151 (vsha512h2q_u64): Ditto.
10152 (vsha512su0q_u64): Ditto.
10153 (vsha512su1q_u64): Ditto.
10154 (veor3q_u16): Ditto.
10155 (vrax1q_u64): Ditto.
10156 (vxarq_u64): Ditto.
10157 (vbcaxq_u16): Ditto.
10158 * config/arm/types.md (crypto_sha512): New type attribute.
10159 (crypto_sha3): Ditto.
10160 (doc/invoke.texi): Document new sha3 option.
10162 2018-01-10 Michael Collison <michael.collison@arm.com>
10164 * config/aarch64/aarch64-builtins.c:
10165 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10166 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10167 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10168 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10169 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10170 (AARCH64_ISA_SM4): New ISA flag.
10171 (TARGET_SM4): New feature flag for sm4.
10172 * config/aarch64/aarch64-simd-builtins.def
10173 (aarch64_sm3ss1qv4si): Ditto.
10174 (aarch64_sm3tt1aq4si): Ditto.
10175 (aarch64_sm3tt1bq4si): Ditto.
10176 (aarch64_sm3tt2aq4si): Ditto.
10177 (aarch64_sm3tt2bq4si): Ditto.
10178 (aarch64_sm3partw1qv4si): Ditto.
10179 (aarch64_sm3partw2qv4si): Ditto.
10180 (aarch64_sm4eqv4si): Ditto.
10181 (aarch64_sm4ekeyqv4si): Ditto.
10182 * config/aarch64/aarch64-simd.md:
10183 (aarch64_sm3ss1qv4si): Ditto.
10184 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10185 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10186 (aarch64_sm4eqv4si): Ditto.
10187 (aarch64_sm4ekeyqv4si): Ditto.
10188 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10189 (sm3part_op): Ditto.
10190 (CRYPTO_SM3TT): Ditto.
10191 (CRYPTO_SM3PART): Ditto.
10192 (UNSPEC_SM3SS1): New unspec.
10193 (UNSPEC_SM3TT1A): Ditto.
10194 (UNSPEC_SM3TT1B): Ditto.
10195 (UNSPEC_SM3TT2A): Ditto.
10196 (UNSPEC_SM3TT2B): Ditto.
10197 (UNSPEC_SM3PARTW1): Ditto.
10198 (UNSPEC_SM3PARTW2): Ditto.
10199 (UNSPEC_SM4E): Ditto.
10200 (UNSPEC_SM4EKEY): Ditto.
10201 * config/aarch64/constraints.md (Ui2): New constraint.
10202 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10203 * config/arm/types.md (crypto_sm3): New type attribute.
10204 (crypto_sm4): Ditto.
10205 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10206 (vsm3tt1aq_u32): Ditto.
10207 (vsm3tt1bq_u32): Ditto.
10208 (vsm3tt2aq_u32): Ditto.
10209 (vsm3tt2bq_u32): Ditto.
10210 (vsm3partw1q_u32): Ditto.
10211 (vsm3partw2q_u32): Ditto.
10212 (vsm4eq_u32): Ditto.
10213 (vsm4ekeyq_u32): Ditto.
10214 (doc/invoke.texi): Document new sm4 option.
10216 2018-01-10 Michael Collison <michael.collison@arm.com>
10218 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10219 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10220 (AARCH64_FL_FOR_ARCH8_4): New.
10221 (AARCH64_FL_V8_4): New flag.
10222 (doc/invoke.texi): Document new armv8.4-a option.
10224 2018-01-10 Michael Collison <michael.collison@arm.com>
10226 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10227 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10228 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10229 * config/aarch64/aarch64-option-extension.def: Add
10230 AARCH64_OPT_EXTENSION of 'sha2'.
10231 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10232 (crypto): Disable sha2 and aes if crypto disabled.
10233 (crypto): Enable aes and sha2 if enabled.
10234 (simd): Disable sha2 and aes if simd disabled.
10235 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10237 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10238 (TARGET_SHA2): New feature flag for sha2.
10239 (TARGET_AES): New feature flag for aes.
10240 * config/aarch64/aarch64-simd.md:
10241 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10242 conditional on TARGET_AES.
10243 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10244 (aarch64_crypto_sha1hsi): Make pattern conditional
10246 (aarch64_crypto_sha1hv4si): Ditto.
10247 (aarch64_be_crypto_sha1hv4si): Ditto.
10248 (aarch64_crypto_sha1su1v4si): Ditto.
10249 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10250 (aarch64_crypto_sha1su0v4si): Ditto.
10251 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10252 (aarch64_crypto_sha256su0v4si): Ditto.
10253 (aarch64_crypto_sha256su1v4si): Ditto.
10254 (doc/invoke.texi): Document new aes and sha2 options.
10256 2018-01-10 Martin Sebor <msebor@redhat.com>
10258 PR tree-optimization/83781
10259 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10262 2018-01-11 Martin Sebor <msebor@gmail.com>
10263 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10265 PR tree-optimization/83501
10266 PR tree-optimization/81703
10268 * tree-ssa-strlen.c (get_string_cst): Rename...
10269 (get_string_len): ...to this. Handle global constants.
10270 (handle_char_store): Adjust.
10272 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10273 Jim Wilson <jimw@sifive.com>
10275 * config/riscv/riscv-protos.h (riscv_output_return): New.
10276 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10277 (riscv_attribute_table, riscv_output_return),
10278 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10279 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10280 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10282 (riscv_expand_prologue): Add early return for naked function.
10283 (riscv_expand_epilogue): Likewise.
10284 (riscv_function_ok_for_sibcall): Return false for naked function.
10285 (riscv_set_current_function): New.
10286 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10287 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10288 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10289 * doc/extend.texi (RISC-V Function Attributes): New.
10291 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10293 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10294 check for 128-bit long double before checking TCmode.
10295 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10296 128-bit long doubles before checking TFmode or TCmode.
10297 (FLOAT128_IBM_P): Likewise.
10299 2018-01-10 Martin Sebor <msebor@redhat.com>
10301 PR tree-optimization/83671
10302 * builtins.c (c_strlen): Unconditionally return zero for the empty
10304 Use -Warray-bounds for warnings.
10305 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10306 for non-constant array indices with COMPONENT_REF, arrays of
10307 arrays, and pointers to arrays.
10308 (gimple_fold_builtin_strlen): Determine and set length range for
10309 non-constant character arrays.
10311 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10313 PR middle-end/81897
10314 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10317 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10319 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10321 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10324 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10325 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10326 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10327 indexed_or_indirect_operand predicate.
10328 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10329 (*vsx_le_perm_load_v8hi): Likewise.
10330 (*vsx_le_perm_load_v16qi): Likewise.
10331 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10332 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10333 (*vsx_le_perm_store_v8hi): Likewise.
10334 (*vsx_le_perm_store_v16qi): Likewise.
10335 (eight unnamed splitters): Likewise.
10337 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10339 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10340 * config/rs6000/emmintrin.h: Likewise.
10341 * config/rs6000/mmintrin.h: Likewise.
10342 * config/rs6000/xmmintrin.h: Likewise.
10344 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10347 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10349 * tree.c (tree_nop_conversion): Return true for location wrapper
10351 (maybe_wrap_with_location): New function.
10352 (selftest::check_strip_nops): New function.
10353 (selftest::test_location_wrappers): New function.
10354 (selftest::tree_c_tests): Call it.
10355 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10356 (maybe_wrap_with_location): New decl.
10357 (EXPR_LOCATION_WRAPPER_P): New macro.
10358 (location_wrapper_p): New inline function.
10359 (tree_strip_any_location_wrapper): New inline function.
10361 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10364 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10365 stack_realign_offset for the largest alignment of stack slot
10367 (ix86_find_max_used_stack_alignment): New function.
10368 (ix86_finalize_stack_frame_flags): Use it. Set
10369 max_used_stack_alignment if we don't realign stack.
10370 * config/i386/i386.h (machine_function): Add
10371 max_used_stack_alignment.
10373 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10375 * config/arm/arm.opt (-mbranch-cost): New option.
10376 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10379 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10382 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10383 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10385 2018-01-10 Richard Biener <rguenther@suse.de>
10388 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10389 early out so it also covers the case where we have a non-NULL
10392 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10394 PR tree-optimization/83753
10395 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10396 for non-strided grouped accesses if the number of elements is 1.
10398 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10401 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10402 * i386.h (TARGET_USE_GATHER): Define.
10403 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10405 2018-01-10 Martin Liska <mliska@suse.cz>
10408 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10409 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10411 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10412 CLEANUP_NO_PARTITIONING is not set.
10414 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10416 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10417 for vectors, as a partial revert of r254296.
10418 * rtl.h (const_vec_p): Delete.
10419 (const_vec_duplicate_p): Don't test for vector CONSTs.
10420 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10421 * expmed.c (make_tree): Likewise.
10424 * common.md (E, F): Use CONSTANT_P instead of checking for
10426 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10427 checking for CONST_VECTOR.
10429 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10431 PR middle-end/83575
10432 * predict.c (force_edge_cold): Handle in more sane way edges
10433 with no prediction.
10435 2018-01-09 Carl Love <cel@us.ibm.com>
10437 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10439 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10440 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10441 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10442 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10443 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10444 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10445 * config/rs6000/rs6000-protos.h: Add extern defition for
10446 rs6000_generate_float2_double_code.
10447 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10449 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10450 (float2_v2df): Add define_expand.
10452 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10455 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10456 op_mode in the force_to_mode call.
10458 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10460 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10461 instead of checking each element individually.
10462 (aarch64_evpc_uzp): Likewise.
10463 (aarch64_evpc_zip): Likewise.
10464 (aarch64_evpc_ext): Likewise.
10465 (aarch64_evpc_rev): Likewise.
10466 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10467 instead of checking each element individually. Return true without
10469 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10470 whether all selected elements come from the same input, instead of
10471 checking each element individually. Remove calls to gen_rtx_REG,
10472 start_sequence and end_sequence and instead assert that no rtl is
10475 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10477 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10478 order of HIGH and CONST checks.
10480 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10482 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10483 if the destination isn't an SSA_NAME.
10485 2018-01-09 Richard Biener <rguenther@suse.de>
10487 PR tree-optimization/83668
10488 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10490 (canonicalize_loop_form): ... here, renamed from ...
10491 (canonicalize_loop_closed_ssa_form): ... this and amended to
10492 swap successor edges for loop exit blocks to make us use
10493 the RPO order we need for initial schedule generation.
10495 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10497 PR tree-optimization/64811
10498 * match.pd: When optimizing comparisons with Inf, avoid
10499 introducing or losing exceptions from comparisons with NaN.
10501 2018-01-09 Martin Liska <mliska@suse.cz>
10504 * asan.c (shadow_mem_size): Add gcc_assert.
10506 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10508 Don't save registers in main().
10511 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10512 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10513 * config/avr/avr.c (avr_set_current_function): Don't error if
10514 naked, OS_task or OS_main are specified at the same time.
10515 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10517 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10519 * common/config/avr/avr-common.c (avr_option_optimization_table):
10520 Switch on -mmain-is-OS_task for optimizing compilations.
10522 2018-01-09 Richard Biener <rguenther@suse.de>
10524 PR tree-optimization/83572
10525 * graphite.c: Include cfganal.h.
10526 (graphite_transform_loops): Connect infinite loops to exit
10527 and remove fake edges at the end.
10529 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10531 * ipa-inline.c (edge_badness): Revert accidental checkin.
10533 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10536 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10537 symbols; not inline clones.
10539 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10542 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10543 hard registers. Formatting fixes.
10545 PR preprocessor/83722
10546 * gcc.c (try_generate_repro): Pass
10547 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10548 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10551 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10552 Kito Cheng <kito.cheng@gmail.com>
10554 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10555 (riscv_leaf_function_p): Delete.
10556 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10558 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10560 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10562 (do_ifelse): New function.
10563 (do_isel): New function.
10564 (do_sub3): New function.
10565 (do_add3): New function.
10566 (do_load_mask_compare): New function.
10567 (do_overlap_load_compare): New function.
10568 (expand_compare_loop): New function.
10569 (expand_block_compare): Call expand_compare_loop() when appropriate.
10570 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10571 option description.
10572 (-mblock-compare-inline-loop-limit): New option.
10574 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10577 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10578 Reverse order of second and third operands in first alternative.
10579 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10580 of first and second elements in UNSPEC_VPERMR vector.
10581 (altivec_expand_vec_perm_le): Likewise.
10583 2018-01-08 Jeff Law <law@redhat.com>
10585 PR rtl-optimizatin/81308
10586 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10587 (process_switch): If group_case_labels makes a change, then set
10589 (pass_convert_switch::execute): If a switch is converted, then
10590 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10592 PR rtl-optimization/81308
10593 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10596 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10598 PR target/83663 - Revert r255946
10599 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10600 generation for cases where splatting a value is not useful.
10601 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10602 across a vec_duplicate and a paradoxical subreg forming a vector
10603 mode to a vec_concat.
10605 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10607 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10608 -march=armv8.3-a variants.
10609 * config/arm/t-multilib: Likewise.
10610 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10612 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10614 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10616 (cceq_ior_compare_complement): Give it a name so I can use it, and
10617 change boolean_or_operator predicate to boolean_operator so it can
10618 be used to generate a crand.
10619 (eqne): New code iterator.
10620 (bd/bd_neg): New code_attrs.
10621 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10622 a single define_insn.
10623 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10624 decrement (bdnzt/bdnzf/bdzt/bdzf).
10625 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10626 with the new names of the branch decrement patterns, and added the
10627 names of the branch decrement conditional patterns.
10629 2018-01-08 Richard Biener <rguenther@suse.de>
10631 PR tree-optimization/83563
10632 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10635 2018-01-08 Richard Biener <rguenther@suse.de>
10637 PR middle-end/83713
10638 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10640 2018-01-08 Richard Biener <rguenther@suse.de>
10642 PR tree-optimization/83685
10643 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10644 references to abnormals.
10646 2018-01-08 Richard Biener <rguenther@suse.de>
10649 * dwarf2out.c (output_indirect_strings): Handle empty
10650 skeleton_debug_str_hash.
10651 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10653 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10655 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10656 (emit_store_direct): Likewise.
10657 (arc_trampoline_adjust_address): Likewise.
10658 (arc_asm_trampoline_template): New function.
10659 (arc_initialize_trampoline): Use asm_trampoline_template.
10660 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10661 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10662 * config/arc/arc.md (flush_icache): Delete pattern.
10664 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10666 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10667 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10670 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10673 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10674 by not USED_FOR_TARGET.
10675 (make_pass_resolve_sw_modes): Likewise.
10677 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10679 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10682 2018-01-08 Richard Biener <rguenther@suse.de>
10684 PR middle-end/83580
10685 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10687 2018-01-08 Richard Biener <rguenther@suse.de>
10689 PR middle-end/83517
10690 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10692 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10694 PR middle-end/81897
10695 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10696 basic blocks with a small number of successors.
10697 (convert_control_dep_chain_into_preds): Improve handling of
10699 (dump_predicates): Split apart into...
10700 (dump_pred_chain): ...here...
10701 (dump_pred_info): ...and here.
10702 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10703 (can_chain_union_be_invalidated_p): Improve check for invalidation
10705 (uninit_uses_cannot_happen): Avoid unnecessary if
10706 convert_control_dep_chain_into_preds yielded nothing.
10708 2018-01-06 Martin Sebor <msebor@redhat.com>
10710 PR tree-optimization/83640
10711 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10712 subtracting negative offset from size.
10713 (builtin_access::overlap): Adjust offset bounds of the access to fall
10714 within the size of the object if possible.
10716 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10718 PR rtl-optimization/83699
10719 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10720 extract_bit_field_as_subreg to cases in which the extracted
10721 value is also a vector.
10723 * lra-constraints.c (process_alt_operands): Test for the equivalence
10724 substitutions when detecting a possible reload cycle.
10726 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10729 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10730 by default if flag_selective_schedling{,2}. Formatting fixes.
10732 PR rtl-optimization/83682
10733 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10734 if it has non-VECTOR_MODE element mode.
10735 (vec_duplicate_p): Likewise.
10737 PR middle-end/83694
10738 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10739 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10741 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10744 * config/i386/i386-builtin.def
10745 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10746 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10747 Require also OPTION_MASK_ISA_AVX512F in addition to
10748 OPTION_MASK_ISA_GFNI.
10749 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10750 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10751 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10752 to OPTION_MASK_ISA_GFNI.
10753 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10754 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10755 OPTION_MASK_ISA_AVX512BW.
10756 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10757 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10758 addition to OPTION_MASK_ISA_GFNI.
10759 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10760 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10761 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10762 to OPTION_MASK_ISA_GFNI.
10763 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10764 a requirement for all ISAs rather than any of them with a few
10766 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10768 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10769 bitmasks to be enabled with 3 exceptions, instead of requiring any
10770 enabled ISA with lots of exceptions.
10771 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10772 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10773 Change avx512bw in isa attribute to avx512f.
10774 * config/i386/sgxintrin.h: Add license boilerplate.
10775 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10776 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10777 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10778 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10780 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10781 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10782 temporarily sse2 rather than sse if not enabled already.
10785 * config/i386/sse.md (VI248_VLBW): Rename to ...
10786 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10787 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10788 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10789 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10790 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10791 mode iterator instead of VI248_VLBW.
10793 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10795 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10796 (record_modified): Skip clobbers; add debug output.
10797 (param_change_prob): Use sreal frequencies.
10799 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10801 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10802 punt for user-aligned variables.
10804 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10806 * tree-chrec.c (chrec_contains_symbols): Return true for
10809 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10812 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10813 of (x|y) == x for BICS pattern.
10815 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10817 PR tree-optimization/83605
10818 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10819 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10822 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10824 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10825 * config/epiphany/rtems.h: New file.
10827 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10828 Uros Bizjak <ubizjak@gmail.com>
10831 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10832 QIreg_operand instead of register_operand predicate.
10833 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10834 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10835 comments instead of -fmitigate[-_]rop.
10837 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10840 * cgraphunit.c (symbol_table::compile): Switch to text_section
10841 before calling assembly_start debug hook.
10842 * run-rtl-passes.c (run_rtl_passes): Likewise.
10845 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10847 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10848 range_int_cst_p rather than !symbolic_range_p before calling
10849 extract_range_from_multiplicative_op_1.
10851 2018-01-04 Jeff Law <law@redhat.com>
10853 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10854 redundant test in assertion.
10856 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10858 * doc/rtl.texi: Document machine_mode wrapper classes.
10860 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10862 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10863 using tree_to_uhwi.
10865 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10867 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10868 the VEC_PERM_EXPR fold to fail.
10870 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10873 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10874 to switched_sections.
10876 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10879 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10880 test for d.testing.
10882 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10885 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10886 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10888 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10891 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10892 is BLKmode and bitpos not zero or mode change is needed.
10894 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10897 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10900 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10903 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10904 instead of MULT rtx. Update all corresponding splitters.
10905 (*saddl_se): Ditto.
10906 (*ssub<modesuffix>): Ditto.
10907 (*ssubl_se): Ditto.
10908 (*cmp_sadd_di): Update split patterns.
10909 (*cmp_sadd_si): Ditto.
10910 (*cmp_sadd_sidi): Ditto.
10911 (*cmp_ssub_di): Ditto.
10912 (*cmp_ssub_si): Ditto.
10913 (*cmp_ssub_sidi): Ditto.
10914 * config/alpha/predicates.md (const23_operand): New predicate.
10915 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10916 Look for ASHIFT, not MULT inner operand.
10917 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10919 2018-01-04 Martin Liska <mliska@suse.cz>
10921 PR gcov-profile/83669
10922 * gcov.c (output_intermediate_file): Add version to intermediate
10924 * doc/gcov.texi: Document new field 'version' in intermediate
10925 file format. Fix location of '-k' option of gcov command.
10927 2018-01-04 Martin Liska <mliska@suse.cz>
10930 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10932 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10934 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10936 2018-01-03 Martin Sebor <msebor@redhat.com>
10938 PR tree-optimization/83655
10939 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10940 checking calls with invalid arguments.
10942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10944 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10945 (vectorizable_mask_load_store): Delete.
10946 (vectorizable_call): Return false for masked loads and stores.
10947 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10948 instead of gimple_assign_rhs1.
10949 (vectorizable_load): Handle IFN_MASK_LOAD.
10950 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10954 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10956 (vectorizable_mask_load_store): ...here.
10957 (vectorizable_load): ...and here.
10959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10961 * tree-vect-stmts.c (vect_build_all_ones_mask)
10962 (vect_build_zero_merge_argument): New functions, split out from...
10963 (vectorizable_load): ...here.
10965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10967 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10969 (vectorizable_mask_load_store): ...here.
10970 (vectorizable_store): ...and here.
10972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10974 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10976 (vectorizable_mask_load_store): ...here.
10978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10980 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10981 (vect_model_store_cost): Take a vec_load_store_type instead of a
10983 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10984 (vect_model_store_cost): Take a vec_load_store_type instead of a
10986 (vectorizable_mask_load_store): Update accordingly.
10987 (vectorizable_store): Likewise.
10988 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10992 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10993 IFN_MASK_LOAD calls here rather than...
10994 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10997 Alan Hayward <alan.hayward@arm.com>
10998 David Sherwood <david.sherwood@arm.com>
11000 * expmed.c (extract_bit_field_1): For vector extracts,
11001 fall back to extract_bit_field_as_subreg if vec_extract
11004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11005 Alan Hayward <alan.hayward@arm.com>
11006 David Sherwood <david.sherwood@arm.com>
11008 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11009 they are variable or constant sized.
11010 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11011 slots for constant-sized data.
11013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11014 Alan Hayward <alan.hayward@arm.com>
11015 David Sherwood <david.sherwood@arm.com>
11017 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11018 handling COND_EXPRs with boolean comparisons, try to find a better
11019 basis for the mask type than the boolean itself.
11021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11023 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11024 is calculated and how it can be overridden.
11025 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11026 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11028 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11032 Alan Hayward <alan.hayward@arm.com>
11033 David Sherwood <david.sherwood@arm.com>
11035 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11036 Remove the mode argument.
11037 (aarch64_simd_valid_immediate): Remove the mode and inverse
11039 * config/aarch64/iterators.md (bitsize): New iterator.
11040 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11041 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11042 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11043 aarch64_simd_valid_immediate.
11044 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11045 (aarch64_reg_or_bic_imm): Likewise.
11046 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11047 with an insn_type enum and msl with a modifier_type enum.
11048 Replace element_width with a scalar_mode. Change the shift
11049 to unsigned int. Add constructors for scalar_float_mode and
11050 scalar_int_mode elements.
11051 (aarch64_vect_float_const_representable_p): Delete.
11052 (aarch64_can_const_movi_rtx_p)
11053 (aarch64_simd_scalar_immediate_valid_for_move)
11054 (aarch64_simd_make_constant): Update call to
11055 aarch64_simd_valid_immediate.
11056 (aarch64_advsimd_valid_immediate_hs): New function.
11057 (aarch64_advsimd_valid_immediate): Likewise.
11058 (aarch64_simd_valid_immediate): Remove mode and inverse
11059 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11060 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11061 and aarch64_float_const_representable_p on the result.
11062 (aarch64_output_simd_mov_immediate): Remove mode argument.
11063 Update call to aarch64_simd_valid_immediate and use of
11064 simd_immediate_info.
11065 (aarch64_output_scalar_simd_mov_immediate): Update call
11068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11069 Alan Hayward <alan.hayward@arm.com>
11070 David Sherwood <david.sherwood@arm.com>
11072 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11073 (mode_nunits): Likewise CONST_MODE_NUNITS.
11074 * machmode.def (ADJUST_NUNITS): Document.
11075 * genmodes.c (mode_data::need_nunits_adj): New field.
11076 (blank_mode): Update accordingly.
11077 (adj_nunits): New variable.
11078 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11080 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11081 listed in adj_nunits.
11082 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11083 listed in adj_nunits. Don't emit case statements for such modes.
11084 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11085 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11086 nothing if adj_nunits is nonnull.
11087 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11088 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11089 (emit_mode_fbit): Update use of print_maybe_const_decl.
11090 (emit_move_size): Likewise. Treat the array as non-const
11092 (emit_mode_adjustments): Handle adj_nunits.
11094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11096 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11097 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11098 (VECTOR_MODES): Use it.
11099 (make_vector_modes): Take the prefix as an argument.
11101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11102 Alan Hayward <alan.hayward@arm.com>
11103 David Sherwood <david.sherwood@arm.com>
11105 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11106 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11107 for MODE_VECTOR_BOOL.
11108 * machmode.def (VECTOR_BOOL_MODE): Document.
11109 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11110 (make_vector_bool_mode): New function.
11111 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11113 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11114 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11116 * stor-layout.c (int_mode_for_mode): Likewise.
11117 * tree.c (build_vector_type_for_mode): Likewise.
11118 * varasm.c (output_constant_pool_2): Likewise.
11119 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11120 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11121 for MODE_VECTOR_BOOL.
11122 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11123 of mode class checks.
11124 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11125 instead of a list of mode class checks.
11126 (expand_vector_scalar_condition): Likewise.
11127 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11130 Alan Hayward <alan.hayward@arm.com>
11131 David Sherwood <david.sherwood@arm.com>
11133 * machmode.h (mode_size): Change from unsigned short to
11135 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11136 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11137 or if measurement_type is not polynomial.
11138 (fixed_size_mode::includes_p): Check for constant-sized modes.
11139 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11140 return a poly_uint16 rather than an unsigned short.
11141 (emit_mode_size): Change the type of mode_size from unsigned short
11142 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11143 (emit_mode_adjustments): Cope with polynomial vector sizes.
11144 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11146 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11148 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11149 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11150 * caller-save.c (setup_save_areas): Likewise.
11151 (replace_reg_with_saved_mem): Likewise.
11152 * calls.c (emit_library_call_value_1): Likewise.
11153 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11154 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11155 (gen_lowpart_for_combine): Likewise.
11156 * convert.c (convert_to_integer_1): Likewise.
11157 * cse.c (equiv_constant, cse_insn): Likewise.
11158 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11159 (cselib_subst_to_values): Likewise.
11160 * dce.c (word_dce_process_block): Likewise.
11161 * df-problems.c (df_word_lr_mark_ref): Likewise.
11162 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11163 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11164 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11165 (rtl_for_decl_location): Likewise.
11166 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11167 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11168 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11169 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11170 (expand_expr_real_1): Likewise.
11171 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11172 (pad_below): Likewise.
11173 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11174 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11175 * ira.c (get_subreg_tracking_sizes): Likewise.
11176 * ira-build.c (ira_create_allocno_objects): Likewise.
11177 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11178 (ira_sort_regnos_for_alter_reg): Likewise.
11179 * ira-costs.c (record_operand_costs): Likewise.
11180 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11181 (resolve_simple_move): Likewise.
11182 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11183 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11184 (lra_constraints): Likewise.
11185 (CONST_POOL_OK_P): Reject variable-sized modes.
11186 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11187 (add_pseudo_to_slot, lra_spill): Likewise.
11188 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11189 * optabs-query.c (get_best_extraction_insn): Likewise.
11190 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11191 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11192 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11193 * recog.c (offsettable_address_addr_space_p): Likewise.
11194 * regcprop.c (maybe_mode_change): Likewise.
11195 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11196 * regrename.c (build_def_use): Likewise.
11197 * regstat.c (dump_reg_info): Likewise.
11198 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11199 (find_reloads, find_reloads_subreg_address): Likewise.
11200 * reload1.c (eliminate_regs_1): Likewise.
11201 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11202 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11203 (simplify_binary_operation_1, simplify_subreg): Likewise.
11204 * targhooks.c (default_function_arg_padding): Likewise.
11205 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11206 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11207 (verify_gimple_assign_ternary): Likewise.
11208 * tree-inline.c (estimate_move_cost): Likewise.
11209 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11210 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11211 (get_address_cost_ainc): Likewise.
11212 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11213 (vect_supportable_dr_alignment): Likewise.
11214 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11215 (vectorizable_reduction): Likewise.
11216 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11217 (vectorizable_operation, vectorizable_load): Likewise.
11218 * tree.c (build_same_sized_truth_vector_type): Likewise.
11219 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11220 * var-tracking.c (emit_note_insn_var_location): Likewise.
11221 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11222 (ADDR_VEC_ALIGN): Likewise.
11224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11225 Alan Hayward <alan.hayward@arm.com>
11226 David Sherwood <david.sherwood@arm.com>
11228 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11230 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11231 or if measurement_type is polynomial.
11232 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11233 * combine.c (make_extraction): Likewise.
11234 * dse.c (find_shift_sequence): Likewise.
11235 * dwarf2out.c (mem_loc_descriptor): Likewise.
11236 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11237 (extract_bit_field, extract_low_bits): Likewise.
11238 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11239 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11240 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11241 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11242 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11243 * reload.c (find_reloads): Likewise.
11244 * reload1.c (alter_reg): Likewise.
11245 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11246 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11247 * tree-if-conv.c (predicate_mem_writes): Likewise.
11248 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11249 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11250 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11251 * valtrack.c (dead_debug_insert_temp): Likewise.
11252 * varasm.c (mergeable_constant_section): Likewise.
11253 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11256 Alan Hayward <alan.hayward@arm.com>
11257 David Sherwood <david.sherwood@arm.com>
11259 * expr.c (expand_assignment): Cope with polynomial mode sizes
11260 when assigning to a CONCAT.
11262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11263 Alan Hayward <alan.hayward@arm.com>
11264 David Sherwood <david.sherwood@arm.com>
11266 * machmode.h (mode_precision): Change from unsigned short to
11268 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11270 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11271 or if measurement_type is not polynomial.
11272 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11273 in which the mode is already known to be a scalar_int_mode.
11274 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11275 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11277 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11278 for GET_MODE_PRECISION.
11279 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11280 for GET_MODE_PRECISION.
11281 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11283 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11284 (expand_field_assignment, make_extraction): Likewise.
11285 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11286 (get_last_value): Likewise.
11287 * convert.c (convert_to_integer_1): Likewise.
11288 * cse.c (cse_insn): Likewise.
11289 * expr.c (expand_expr_real_1): Likewise.
11290 * lra-constraints.c (simplify_operand_subreg): Likewise.
11291 * optabs-query.c (can_atomic_load_p): Likewise.
11292 * optabs.c (expand_atomic_load): Likewise.
11293 (expand_atomic_store): Likewise.
11294 * ree.c (combine_reaching_defs): Likewise.
11295 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11296 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11297 * tree.h (type_has_mode_precision_p): Likewise.
11298 * ubsan.c (instrument_si_overflow): Likewise.
11300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11301 Alan Hayward <alan.hayward@arm.com>
11302 David Sherwood <david.sherwood@arm.com>
11304 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11305 polynomial numbers of units.
11306 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11307 (valid_vector_subparts_p): New function.
11308 (build_vector_type): Remove temporary shim and take the number
11309 of units as a poly_uint64 rather than an int.
11310 (build_opaque_vector_type): Take the number of units as a
11311 poly_uint64 rather than an int.
11312 * tree.c (build_vector_from_ctor): Handle polynomial
11313 TYPE_VECTOR_SUBPARTS.
11314 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11315 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11316 (build_vector_from_val): If the number of units is variable,
11317 use build_vec_duplicate_cst for constant operands and
11318 VEC_DUPLICATE_EXPR otherwise.
11319 (make_vector_type): Remove temporary is_constant ().
11320 (build_vector_type, build_opaque_vector_type): Take the number of
11321 units as a poly_uint64 rather than an int.
11322 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11324 * cfgexpand.c (expand_debug_expr): Likewise.
11325 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11326 (store_constructor, expand_expr_real_1): Likewise.
11327 (const_scalar_mask_from_tree): Likewise.
11328 * fold-const-call.c (fold_const_reduction): Likewise.
11329 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11330 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11331 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11332 (fold_relational_const): Likewise.
11333 (native_interpret_vector): Likewise. Change the size from an
11334 int to an unsigned int.
11335 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11336 TYPE_VECTOR_SUBPARTS.
11337 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11338 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11339 duplicating a non-constant operand into a variable-length vector.
11340 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11341 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11342 * ipa-icf.c (sem_variable::equals): Likewise.
11343 * match.pd: Likewise.
11344 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11345 * print-tree.c (print_node): Likewise.
11346 * stor-layout.c (layout_type): Likewise.
11347 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11348 * tree-cfg.c (verify_gimple_comparison): Likewise.
11349 (verify_gimple_assign_binary): Likewise.
11350 (verify_gimple_assign_ternary): Likewise.
11351 (verify_gimple_assign_single): Likewise.
11352 * tree-pretty-print.c (dump_generic_node): Likewise.
11353 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11354 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11355 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11356 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11357 (vect_shift_permute_load_chain): Likewise.
11358 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11359 (expand_vector_condition, optimize_vector_constructor): Likewise.
11360 (lower_vec_perm, get_compute_type): Likewise.
11361 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11362 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11363 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11364 (vect_recog_mask_conversion_pattern): Likewise.
11365 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11366 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11367 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11368 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11369 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11370 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11371 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11372 (supportable_widening_operation): Likewise.
11373 (supportable_narrowing_operation): Likewise.
11374 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11376 * varasm.c (output_constant): Likewise.
11378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11379 Alan Hayward <alan.hayward@arm.com>
11380 David Sherwood <david.sherwood@arm.com>
11382 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11383 so that both the length == 3 and length != 3 cases set up their
11384 own permute vectors. Add comments explaining why we know the
11385 number of elements is constant.
11386 (vect_permute_load_chain): Likewise.
11388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11389 Alan Hayward <alan.hayward@arm.com>
11390 David Sherwood <david.sherwood@arm.com>
11392 * machmode.h (mode_nunits): Change from unsigned char to
11394 (ONLY_FIXED_SIZE_MODES): New macro.
11395 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11396 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11397 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11399 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11400 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11401 or if measurement_type is not polynomial.
11402 * genmodes.c (ZERO_COEFFS): New macro.
11403 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11405 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11406 Use ZERO_COEFFS when emitting initializers.
11407 * data-streamer.h (bp_pack_poly_value): New function.
11408 (bp_unpack_poly_value): Likewise.
11409 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11410 for GET_MODE_NUNITS.
11411 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11412 for GET_MODE_NUNITS.
11413 * tree.c (make_vector_type): Remove temporary shim and make
11414 the real function take the number of units as a poly_uint64
11415 rather than an int.
11416 (build_vector_type_for_mode): Handle polynomial nunits.
11417 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11418 * emit-rtl.c (const_vec_series_p_1): Likewise.
11419 (gen_rtx_CONST_VECTOR): Likewise.
11420 * fold-const.c (test_vec_duplicate_folding): Likewise.
11421 * genrecog.c (validate_pattern): Likewise.
11422 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11423 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11424 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11425 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11426 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11427 * rtlanal.c (subreg_get_info): Likewise.
11428 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11429 (vect_grouped_load_supported): Likewise.
11430 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11431 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11432 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11433 (simplify_const_unary_operation, simplify_binary_operation_1)
11434 (simplify_const_binary_operation, simplify_ternary_operation)
11435 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11436 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11437 instead of CONST_VECTOR_NUNITS.
11438 * varasm.c (output_constant_pool_2): Likewise.
11439 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11440 explicit-encoded elements in the XVEC for variable-length vectors.
11442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11444 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11447 Alan Hayward <alan.hayward@arm.com>
11448 David Sherwood <david.sherwood@arm.com>
11450 * coretypes.h (fixed_size_mode): Declare.
11451 (fixed_size_mode_pod): New typedef.
11452 * builtins.h (target_builtins::x_apply_args_mode)
11453 (target_builtins::x_apply_result_mode): Change type to
11454 fixed_size_mode_pod.
11455 * builtins.c (apply_args_size, apply_result_size, result_vector)
11456 (expand_builtin_apply_args_1, expand_builtin_apply)
11457 (expand_builtin_return): Update accordingly.
11459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11461 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11462 * cselib.c (cselib_hash_rtx): Likewise.
11463 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11464 CONST_VECTOR encoding.
11466 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11467 Jeff Law <law@redhat.com>
11470 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11471 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11472 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11473 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11476 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11477 explicitly probe *sp in a noreturn function if there were any callee
11478 register saves or frame pointer is needed.
11480 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11483 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11484 BLKmode for ternary, binary or unary expressions.
11487 * var-tracking.c (delete_vta_debug_insn): New inline function.
11488 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11489 insns from get_insns () to NULL instead of each bb separately.
11490 Use delete_vta_debug_insn. No longer static.
11491 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11492 delete_vta_debug_insns callers.
11493 * rtl.h (delete_vta_debug_insns): Declare.
11494 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11495 instead of variable_tracking_main.
11497 2018-01-03 Martin Sebor <msebor@redhat.com>
11499 PR tree-optimization/83603
11500 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11501 arguments past the endof the argument list in functions declared
11502 without a prototype.
11503 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11504 Avoid checking when arguments are null.
11506 2018-01-03 Martin Sebor <msebor@redhat.com>
11509 * doc/extend.texi (attribute const): Fix a typo.
11510 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11511 issuing -Wsuggest-attribute for void functions.
11513 2018-01-03 Martin Sebor <msebor@redhat.com>
11515 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11516 offset_int::from instead of wide_int::to_shwi.
11517 (maybe_diag_overlap): Remove assertion.
11518 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11519 * gimple-ssa-sprintf.c (format_directive): Same.
11520 (parse_directive): Same.
11521 (sprintf_dom_walker::compute_format_length): Same.
11522 (try_substitute_return_value): Same.
11524 2018-01-03 Jeff Law <law@redhat.com>
11526 PR middle-end/83654
11527 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11528 non-constant residual for zero at runtime and avoid probing in
11529 that case. Reorganize code for trailing problem to mirror handling
11532 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11534 PR tree-optimization/83501
11535 * tree-ssa-strlen.c (get_string_cst): New.
11536 (handle_char_store): Call get_string_cst.
11538 2018-01-03 Martin Liska <mliska@suse.cz>
11540 PR tree-optimization/83593
11541 * tree-ssa-strlen.c: Include tree-cfg.h.
11542 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11543 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11544 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11546 (strlen_dom_walker::before_dom_children): Call
11547 gimple_purge_dead_eh_edges. Dump tranformation with details
11549 (strlen_dom_walker::before_dom_children): Update call by adding
11550 new argument cleanup_eh.
11551 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11553 2018-01-03 Martin Liska <mliska@suse.cz>
11556 * cif-code.def (VARIADIC_THUNK): New enum value.
11557 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11560 2018-01-03 Jan Beulich <jbeulich@suse.com>
11562 * sse.md (mov<mode>_internal): Tighten condition for when to use
11563 vmovdqu<ssescalarsize> for TI and OI modes.
11565 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11567 Update copyright years.
11569 2018-01-03 Martin Liska <mliska@suse.cz>
11572 * ipa-visibility.c (function_and_variable_visibility): Skip
11573 functions with noipa attribure.
11575 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11577 * gcc.c (process_command): Update copyright notice dates.
11578 * gcov-dump.c (print_version): Ditto.
11579 * gcov.c (print_version): Ditto.
11580 * gcov-tool.c (print_version): Ditto.
11581 * gengtype.c (create_file): Ditto.
11582 * doc/cpp.texi: Bump @copying's copyright year.
11583 * doc/cppinternals.texi: Ditto.
11584 * doc/gcc.texi: Ditto.
11585 * doc/gccint.texi: Ditto.
11586 * doc/gcov.texi: Ditto.
11587 * doc/install.texi: Ditto.
11588 * doc/invoke.texi: Ditto.
11590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11592 * vector-builder.h (vector_builder::m_full_nelts): Change from
11593 unsigned int to poly_uint64.
11594 (vector_builder::full_nelts): Update prototype accordingly.
11595 (vector_builder::new_vector): Likewise.
11596 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11597 (vector_builder::operator ==): Likewise.
11598 (vector_builder::finalize): Likewise.
11599 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11600 Take the number of elements as a poly_uint64 rather than an
11602 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11603 from unsigned int to poly_uint64.
11604 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11605 (vec_perm_indices::new_vector): Likewise.
11606 (vec_perm_indices::length): Likewise.
11607 (vec_perm_indices::nelts_per_input): Likewise.
11608 (vec_perm_indices::input_nelts): Likewise.
11609 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11610 number of elements per input as a poly_uint64 rather than an
11611 unsigned int. Use the original encoding for variable-length
11612 vectors, rather than clamping each individual element.
11613 For the second and subsequent elements in each pattern,
11614 clamp the step and base before clamping their sum.
11615 (vec_perm_indices::series_p): Handle polynomial element counts.
11616 (vec_perm_indices::all_in_range_p): Likewise.
11617 (vec_perm_indices_to_tree): Likewise.
11618 (vec_perm_indices_to_rtx): Likewise.
11619 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11620 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11621 (tree_vector_builder::new_binary_operation): Handle polynomial
11622 element counts. Return false if we need to know the number
11623 of elements at compile time.
11624 * fold-const.c (fold_vec_perm): Punt if the number of elements
11625 isn't known at compile time.
11627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11629 * vec-perm-indices.h (vec_perm_builder): Change element type
11630 from HOST_WIDE_INT to poly_int64.
11631 (vec_perm_indices::element_type): Update accordingly.
11632 (vec_perm_indices::clamp): Handle polynomial element_types.
11633 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11634 (vec_perm_indices::all_in_range_p): Likewise.
11635 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11637 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11638 polynomial vec_perm_indices element types.
11639 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11640 * fold-const.c (fold_vec_perm): Likewise.
11641 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11642 * tree-vect-generic.c (lower_vec_perm): Likewise.
11643 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11644 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11645 element type to HOST_WIDE_INT.
11647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11648 Alan Hayward <alan.hayward@arm.com>
11649 David Sherwood <david.sherwood@arm.com>
11651 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11652 rather than an int. Use plus_constant.
11653 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11654 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11657 Alan Hayward <alan.hayward@arm.com>
11658 David Sherwood <david.sherwood@arm.com>
11660 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11661 a HOST_WIDE_INT to a poly_int64.
11663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11664 Alan Hayward <alan.hayward@arm.com>
11665 David Sherwood <david.sherwood@arm.com>
11667 * calls.c (load_register_parameters): Cope with polynomial
11668 mode sizes. Require a constant size for BLKmode parameters
11669 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11670 forces a parameter to be padded at the lsb end in order to
11671 fill a complete number of words, require the parameter size
11672 to be ordered wrt UNITS_PER_WORD.
11674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11675 Alan Hayward <alan.hayward@arm.com>
11676 David Sherwood <david.sherwood@arm.com>
11678 * reload1.c (spill_stack_slot_width): Change element type
11679 from unsigned int to poly_uint64_pod.
11680 (alter_reg): Treat mode sizes as polynomial.
11682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11683 Alan Hayward <alan.hayward@arm.com>
11684 David Sherwood <david.sherwood@arm.com>
11686 * reload.c (complex_word_subreg_p): New function.
11687 (reload_inner_reg_of_subreg, push_reload): Use it.
11689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11690 Alan Hayward <alan.hayward@arm.com>
11691 David Sherwood <david.sherwood@arm.com>
11693 * lra-constraints.c (process_alt_operands): Reject matched
11694 operands whose sizes aren't ordered.
11695 (match_reload): Refer to this check here.
11697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11698 Alan Hayward <alan.hayward@arm.com>
11699 David Sherwood <david.sherwood@arm.com>
11701 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11702 that the mode size is in the set {1, 2, 4, 8, 16}.
11704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11705 Alan Hayward <alan.hayward@arm.com>
11706 David Sherwood <david.sherwood@arm.com>
11708 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11709 Use plus_constant instead of gen_rtx_PLUS.
11711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11712 Alan Hayward <alan.hayward@arm.com>
11713 David Sherwood <david.sherwood@arm.com>
11715 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11716 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11717 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11718 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11719 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11720 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11721 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11722 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11723 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11724 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11726 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11727 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11728 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11729 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11730 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11731 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11732 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11733 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11734 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11735 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11737 * expr.c (emit_move_resolve_push): Treat the input and result
11738 of PUSH_ROUNDING as a poly_int64.
11739 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11740 (emit_push_insn): Likewise.
11741 * lra-eliminations.c (mark_not_eliminable): Likewise.
11742 * recog.c (push_operand): Likewise.
11743 * reload1.c (elimination_effects): Likewise.
11744 * rtlanal.c (nonzero_bits1): Likewise.
11745 * calls.c (store_one_arg): Likewise. Require the padding to be
11746 known at compile time.
11748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11749 Alan Hayward <alan.hayward@arm.com>
11750 David Sherwood <david.sherwood@arm.com>
11752 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11753 Use plus_constant instead of gen_rtx_PLUS.
11755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11756 Alan Hayward <alan.hayward@arm.com>
11757 David Sherwood <david.sherwood@arm.com>
11759 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11760 rather than an int.
11762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11763 Alan Hayward <alan.hayward@arm.com>
11764 David Sherwood <david.sherwood@arm.com>
11766 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11767 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11768 via stack temporaries. Treat the mode size as polynomial too.
11770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11771 Alan Hayward <alan.hayward@arm.com>
11772 David Sherwood <david.sherwood@arm.com>
11774 * expr.c (expand_expr_real_2): When handling conversions involving
11775 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11776 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11777 as a poly_uint64 too.
11779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11780 Alan Hayward <alan.hayward@arm.com>
11781 David Sherwood <david.sherwood@arm.com>
11783 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11786 Alan Hayward <alan.hayward@arm.com>
11787 David Sherwood <david.sherwood@arm.com>
11789 * combine.c (can_change_dest_mode): Handle polynomial
11790 REGMODE_NATURAL_SIZE.
11791 * expmed.c (store_bit_field_1): Likewise.
11792 * expr.c (store_constructor): Likewise.
11793 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11794 and polynomial REGMODE_NATURAL_SIZE.
11795 (gen_lowpart_common): Likewise.
11796 * reginfo.c (record_subregs_of_mode): Likewise.
11797 * rtlanal.c (read_modify_subreg_p): Likewise.
11799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11800 Alan Hayward <alan.hayward@arm.com>
11801 David Sherwood <david.sherwood@arm.com>
11803 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11804 numbers of elements.
11806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11807 Alan Hayward <alan.hayward@arm.com>
11808 David Sherwood <david.sherwood@arm.com>
11810 * match.pd: Cope with polynomial numbers of vector elements.
11812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11813 Alan Hayward <alan.hayward@arm.com>
11814 David Sherwood <david.sherwood@arm.com>
11816 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11817 in a POINTER_PLUS_EXPR.
11819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11820 Alan Hayward <alan.hayward@arm.com>
11821 David Sherwood <david.sherwood@arm.com>
11823 * omp-simd-clone.c (simd_clone_subparts): New function.
11824 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11825 (ipa_simd_modify_function_body): Likewise.
11827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11828 Alan Hayward <alan.hayward@arm.com>
11829 David Sherwood <david.sherwood@arm.com>
11831 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11832 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11833 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11834 (expand_vector_condition, vector_element): Likewise.
11835 (subparts_gt): New function.
11836 (get_compute_type): Use subparts_gt.
11837 (count_type_subparts): Delete.
11838 (expand_vector_operations_1): Use subparts_gt instead of
11839 count_type_subparts.
11841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11842 Alan Hayward <alan.hayward@arm.com>
11843 David Sherwood <david.sherwood@arm.com>
11845 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11846 (vect_compile_time_alias): ...this new function. Do the calculation
11847 on poly_ints rather than trees.
11848 (vect_prune_runtime_alias_test_list): Update call accordingly.
11850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11851 Alan Hayward <alan.hayward@arm.com>
11852 David Sherwood <david.sherwood@arm.com>
11854 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11856 (vect_schedule_slp_instance): Likewise.
11858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11859 Alan Hayward <alan.hayward@arm.com>
11860 David Sherwood <david.sherwood@arm.com>
11862 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11863 constant and extern definitions for variable-length vectors.
11864 (vect_get_constant_vectors): Note that the number of units
11865 is known to be constant.
11867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11868 Alan Hayward <alan.hayward@arm.com>
11869 David Sherwood <david.sherwood@arm.com>
11871 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11872 of units as polynomial. Choose between WIDE and NARROW based
11875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11876 Alan Hayward <alan.hayward@arm.com>
11877 David Sherwood <david.sherwood@arm.com>
11879 * tree-vect-stmts.c (simd_clone_subparts): New function.
11880 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11883 Alan Hayward <alan.hayward@arm.com>
11884 David Sherwood <david.sherwood@arm.com>
11886 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11887 vectors as polynomial. Use build_index_vector for
11888 IFN_GOMP_SIMD_LANE.
11890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11891 Alan Hayward <alan.hayward@arm.com>
11892 David Sherwood <david.sherwood@arm.com>
11894 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11895 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11896 for variable-length vectors.
11897 (vectorizable_mask_load_store): Treat the number of units as
11898 polynomial, asserting that it is constant if the condition has
11899 already been enforced.
11900 (vectorizable_store, vectorizable_load): Likewise.
11902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11903 Alan Hayward <alan.hayward@arm.com>
11904 David Sherwood <david.sherwood@arm.com>
11906 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11907 of units as polynomial. Punt if we can't tell at compile time
11908 which vector contains the final result.
11910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11911 Alan Hayward <alan.hayward@arm.com>
11912 David Sherwood <david.sherwood@arm.com>
11914 * tree-vect-loop.c (vectorizable_induction): Treat the number
11915 of units as polynomial. Punt on SLP inductions. Use an integer
11916 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11917 cast of such a series for variable-length floating-point
11920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11921 Alan Hayward <alan.hayward@arm.com>
11922 David Sherwood <david.sherwood@arm.com>
11924 * tree.h (build_index_vector): Declare.
11925 * tree.c (build_index_vector): New function.
11926 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11927 of units as polynomial, forcibly converting it to a constant if
11928 vectorizable_reduction has already enforced the condition.
11929 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11930 to create a {1,2,3,...} vector.
11931 (vectorizable_reduction): Treat the number of units as polynomial.
11932 Choose vectype_in based on the largest scalar element size rather
11933 than the smallest number of units. Enforce the restrictions
11936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11937 Alan Hayward <alan.hayward@arm.com>
11938 David Sherwood <david.sherwood@arm.com>
11940 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11941 number of units as polynomial.
11943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11944 Alan Hayward <alan.hayward@arm.com>
11945 David Sherwood <david.sherwood@arm.com>
11947 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11948 * target.def (autovectorize_vector_sizes): Return the vector sizes
11949 by pointer, using vector_sizes rather than a bitmask.
11950 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11951 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11952 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11954 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11955 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11956 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11957 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11958 * omp-general.c (omp_max_vf): Likewise.
11959 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11960 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11961 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11962 * tree-vect-slp.c (vect_slp_bb): Likewise.
11963 * doc/tm.texi: Regenerate.
11964 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11966 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11967 the vector size as a poly_uint64 rather than an unsigned int.
11968 (current_vector_size): Change from an unsigned int to a poly_uint64.
11969 (get_vectype_for_scalar_type): Update accordingly.
11970 * tree.h (build_truth_vector_type): Take the size and number of
11971 units as a poly_uint64 rather than an unsigned int.
11972 (build_vector_type): Add a temporary overload that takes
11973 the number of units as a poly_uint64 rather than an unsigned int.
11974 * tree.c (make_vector_type): Likewise.
11975 (build_truth_vector_type): Take the number of units as a poly_uint64
11976 rather than an unsigned int.
11978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11979 Alan Hayward <alan.hayward@arm.com>
11980 David Sherwood <david.sherwood@arm.com>
11982 * target.def (get_mask_mode): Take the number of units and length
11983 as poly_uint64s rather than unsigned ints.
11984 * targhooks.h (default_get_mask_mode): Update accordingly.
11985 * targhooks.c (default_get_mask_mode): Likewise.
11986 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11987 * doc/tm.texi: Regenerate.
11989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11990 Alan Hayward <alan.hayward@arm.com>
11991 David Sherwood <david.sherwood@arm.com>
11993 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11994 * omp-general.c (omp_max_vf): Likewise.
11995 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11996 (expand_omp_simd): Handle polynomial safelen.
11997 * omp-low.c (omplow_simd_context): Add a default constructor.
11998 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11999 (lower_rec_simd_input_clauses): Update accordingly.
12000 (lower_rec_input_clauses): Likewise.
12002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12003 Alan Hayward <alan.hayward@arm.com>
12004 David Sherwood <david.sherwood@arm.com>
12006 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12007 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12008 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12009 (vect_analyze_slp_cost): Likewise.
12010 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12011 (vect_model_load_cost): Likewise.
12013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12014 Alan Hayward <alan.hayward@arm.com>
12015 David Sherwood <david.sherwood@arm.com>
12017 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12018 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12019 from an unsigned int * to a poly_uint64_pod *.
12020 (calculate_unrolling_factor): New function.
12021 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12024 Alan Hayward <alan.hayward@arm.com>
12025 David Sherwood <david.sherwood@arm.com>
12027 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12028 from an unsigned int to a poly_uint64.
12029 (_loop_vec_info::slp_unrolling_factor): Likewise.
12030 (_loop_vec_info::vectorization_factor): Change from an int
12032 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12033 (vect_get_num_vectors): New function.
12034 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12035 (vect_get_num_copies): Use vect_get_num_vectors.
12036 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12037 to an unsigned int *.
12038 (vect_analyze_data_refs): Change min_vf from an int * to a
12040 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12041 than an unsigned HOST_WIDE_INT.
12042 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12043 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12044 to an unsigned int *.
12045 (vect_analyze_data_ref_dependences): Likewise.
12046 (vect_compute_data_ref_alignment): Handle polynomial vf.
12047 (vect_enhance_data_refs_alignment): Likewise.
12048 (vect_prune_runtime_alias_test_list): Likewise.
12049 (vect_shift_permute_load_chain): Likewise.
12050 (vect_supportable_dr_alignment): Likewise.
12051 (dependence_distance_ge_vf): Take the vectorization factor as a
12052 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12053 (vect_analyze_data_refs): Change min_vf from an int * to a
12055 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12056 vfm1 as a poly_uint64 rather than an int. Make the same change
12057 for the returned bound_scalar.
12058 (vect_gen_vector_loop_niters): Handle polynomial vf.
12059 (vect_do_peeling): Likewise. Update call to
12060 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12061 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12063 * tree-vect-loop.c (vect_determine_vectorization_factor)
12064 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12065 (vect_get_known_peeling_cost): Likewise.
12066 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12067 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12068 (vect_transform_loop): Likewise. Use the lowest possible VF when
12069 updating the upper bounds of the loop.
12070 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12071 rather than an int.
12072 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12073 polynomial unroll factors.
12074 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12075 (vect_make_slp_decision): Likewise.
12076 (vect_supported_load_permutation_p): Likewise, and polynomial
12078 (vect_analyze_slp_cost): Handle polynomial vf.
12079 (vect_slp_analyze_node_operations): Likewise.
12080 (vect_slp_analyze_bb_1): Likewise.
12081 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12082 than an unsigned HOST_WIDE_INT.
12083 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12084 (vectorizable_load): Handle polynomial vf.
12085 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12087 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12090 Alan Hayward <alan.hayward@arm.com>
12091 David Sherwood <david.sherwood@arm.com>
12093 * match.pd: Handle bit operations involving three constants
12094 and try to fold one pair.
12096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12098 * tree-vect-loop-manip.c: Include gimple-fold.h.
12099 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12100 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12101 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12102 Add a path that uses a step of VF instead of 1, but disable it
12104 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12105 and niters_no_overflow parameters. Update calls to
12106 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12107 Create a new SSA name if the latter choses to use a ste other
12108 than zero, and return it via niters_vector_mult_vf_var.
12109 * tree-vect-loop.c (vect_transform_loop): Update calls to
12110 vect_do_peeling, vect_gen_vector_loop_niters and
12111 slpeel_make_loop_iterate_ntimes.
12112 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12113 (vect_gen_vector_loop_niters): Update declarations after above changes.
12115 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12117 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12118 128-bit round to integer instructions.
12119 (ceil<mode>2): Likewise.
12120 (btrunc<mode>2): Likewise.
12121 (round<mode>2): Likewise.
12123 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12125 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12126 unaligned VSX load/store on P8/P9.
12127 (expand_block_clear): Allow the use of unaligned VSX
12128 load/store on P8/P9.
12130 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12132 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12134 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12135 swap associated with both a load and a store.
12137 2018-01-02 Andrew Waterman <andrew@sifive.com>
12139 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12140 * config/riscv/riscv.md (clear_cache): Use it.
12142 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12144 * web.c: Remove out-of-date comment.
12146 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12148 * expr.c (fixup_args_size_notes): Check that any existing
12149 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12150 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12151 (emit_single_push_insn): ...here.
12153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12155 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12156 (const_vector_encoded_nelts): New function.
12157 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12158 (const_vector_int_elt, const_vector_elt): Declare.
12159 * emit-rtl.c (const_vector_int_elt_1): New function.
12160 (const_vector_elt): Likewise.
12161 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12162 of CONST_VECTOR_ELT.
12164 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12166 * expr.c: Include rtx-vector-builder.h.
12167 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12168 directly on the tree encoding.
12169 (const_vector_from_tree): Likewise.
12170 * optabs.c: Include rtx-vector-builder.h.
12171 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12172 sequence of "u" values.
12173 * vec-perm-indices.c: Include rtx-vector-builder.h.
12174 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12175 directly on the vec_perm_indices encoding.
12177 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12179 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12180 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12181 * rtx-vector-builder.h: New file.
12182 * rtx-vector-builder.c: Likewise.
12183 * rtl.h (rtx_def::u2): Add a const_vector field.
12184 (CONST_VECTOR_NPATTERNS): New macro.
12185 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12186 (CONST_VECTOR_DUPLICATE_P): Likewise.
12187 (CONST_VECTOR_STEPPED_P): Likewise.
12188 (CONST_VECTOR_ENCODED_ELT): Likewise.
12189 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12190 (unwrap_const_vec_duplicate): Likewise.
12191 (const_vec_series_p): Check for a non-duplicated vector encoding.
12192 Say that the function only returns true for integer vectors.
12193 * emit-rtl.c: Include rtx-vector-builder.h.
12194 (gen_const_vec_duplicate_1): Delete.
12195 (gen_const_vector): Call gen_const_vec_duplicate instead of
12196 gen_const_vec_duplicate_1.
12197 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12198 (gen_const_vec_duplicate): Use rtx_vector_builder.
12199 (gen_const_vec_series): Likewise.
12200 (gen_rtx_CONST_VECTOR): Likewise.
12201 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12202 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12203 Build a new vector rather than modifying a CONST_VECTOR in-place.
12204 (handle_special_swappables): Update call accordingly.
12205 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12206 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12207 Build a new vector rather than modifying a CONST_VECTOR in-place.
12208 (handle_special_swappables): Update call accordingly.
12210 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12212 * simplify-rtx.c (simplify_const_binary_operation): Use
12213 CONST_VECTOR_ELT instead of XVECEXP.
12215 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12217 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12218 the selector elements to be different from the data elements
12219 if the selector is a VECTOR_CST.
12220 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12221 ssizetype for the selector.
12223 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12225 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12226 before testing each element individually.
12227 * tree-vect-generic.c (lower_vec_perm): Likewise.
12229 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12231 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12232 * selftest-run-tests.c (selftest::run_tests): Call it.
12233 * vector-builder.h (vector_builder::operator ==): New function.
12234 (vector_builder::operator !=): Likewise.
12235 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12236 (vec_perm_indices::all_from_input_p): New function.
12237 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12238 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12239 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12240 instead of reading the VECTOR_CST directly. Detect whether both
12241 vector inputs are the same before constructing the vec_perm_indices,
12242 and update the number of inputs argument accordingly. Use the
12243 utility functions added above. Only construct sel2 if we need to.
12245 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12247 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12248 the broadcast of the low byte.
12249 (expand_mult_highpart): Use an explicit encoding for the permutes.
12250 * optabs-query.c (can_mult_highpart_p): Likewise.
12251 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12252 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12253 (vectorizable_bswap): Likewise.
12254 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12255 explicit encoding for the power-of-2 permutes.
12256 (vect_permute_store_chain): Likewise.
12257 (vect_grouped_load_supported): Likewise.
12258 (vect_permute_load_chain): Likewise.
12260 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12262 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12263 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12264 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12265 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12266 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12267 (vect_gen_perm_mask_any): Likewise.
12269 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12271 * int-vector-builder.h: New file.
12272 * vec-perm-indices.h: Include int-vector-builder.h.
12273 (vec_perm_indices): Redefine as an int_vector_builder.
12274 (auto_vec_perm_indices): Delete.
12275 (vec_perm_builder): Redefine as a stand-alone class.
12276 (vec_perm_indices::vec_perm_indices): New function.
12277 (vec_perm_indices::clamp): Likewise.
12278 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12279 (vec_perm_indices::new_vector): New function.
12280 (vec_perm_indices::new_expanded_vector): Update for new
12281 vec_perm_indices class.
12282 (vec_perm_indices::rotate_inputs): New function.
12283 (vec_perm_indices::all_in_range_p): Operate directly on the
12284 encoded form, without computing elided elements.
12285 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12286 encoding. Update for new vec_perm_indices class.
12287 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12288 the given vec_perm_builder.
12289 (expand_vec_perm_var): Update vec_perm_builder constructor.
12290 (expand_mult_highpart): Use vec_perm_builder instead of
12291 auto_vec_perm_indices.
12292 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12293 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12294 or double series encoding as appropriate.
12295 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12296 vec_perm_indices instead of auto_vec_perm_indices.
12297 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12298 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12299 (vect_permute_store_chain): Likewise.
12300 (vect_grouped_load_supported): Likewise.
12301 (vect_permute_load_chain): Likewise.
12302 (vect_shift_permute_load_chain): Likewise.
12303 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12304 (vect_transform_slp_perm_load): Likewise.
12305 (vect_schedule_slp_instance): Likewise.
12306 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12307 (vectorizable_mask_load_store): Likewise.
12308 (vectorizable_bswap): Likewise.
12309 (vectorizable_store): Likewise.
12310 (vectorizable_load): Likewise.
12311 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12312 vec_perm_indices instead of auto_vec_perm_indices. Use
12313 tree_to_vec_perm_builder to read the vector from a tree.
12314 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12315 vec_perm_builder instead of a vec_perm_indices.
12316 (have_whole_vector_shift): Use vec_perm_builder and
12317 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12318 truncation to calc_vec_perm_mask_for_shift.
12319 (vect_create_epilog_for_reduction): Likewise.
12320 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12321 from auto_vec_perm_indices to vec_perm_indices.
12322 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12323 instead of changing individual elements.
12324 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12325 the vector in d.perm.
12326 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12327 from auto_vec_perm_indices to vec_perm_indices.
12328 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12329 instead of changing individual elements.
12330 (arm_vectorize_vec_perm_const): Use new_vector to install
12331 the vector in d.perm.
12332 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12333 Update vec_perm_builder constructor.
12334 (rs6000_expand_interleave): Likewise.
12335 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12336 (rs6000_expand_interleave): Likewise.
12338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12340 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12341 to qimode could truncate the indices.
12342 * optabs.c (expand_vec_perm_var): Likewise.
12344 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12346 * Makefile.in (OBJS): Add vec-perm-indices.o.
12347 * vec-perm-indices.h: New file.
12348 * vec-perm-indices.c: Likewise.
12349 * target.h (vec_perm_indices): Replace with a forward class
12351 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12352 * optabs.h: Include vec-perm-indices.h.
12353 (expand_vec_perm): Delete.
12354 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12355 (expand_vec_perm_const): Declare.
12356 * target.def (vec_perm_const_ok): Replace with...
12357 (vec_perm_const): ...this new hook.
12358 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12359 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12360 * doc/tm.texi: Regenerate.
12361 * optabs.def (vec_perm_const): Delete.
12362 * doc/md.texi (vec_perm_const): Likewise.
12363 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12364 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12365 expand_vec_perm for constant permutation vectors. Assert that
12366 the mode of variable permutation vectors is the integer equivalent
12367 of the mode that is being permuted.
12368 * optabs-query.h (selector_fits_mode_p): Declare.
12369 * optabs-query.c: Include vec-perm-indices.h.
12370 (selector_fits_mode_p): New function.
12371 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12372 is defined, instead of checking whether the vec_perm_const_optab
12373 exists. Use targetm.vectorize.vec_perm_const instead of
12374 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12375 fit in the vector mode before using a variable permute.
12376 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12377 vec_perm_indices instead of an rtx.
12378 (expand_vec_perm): Replace with...
12379 (expand_vec_perm_const): ...this new function. Take the selector
12380 as a vec_perm_indices rather than an rtx. Also take the mode of
12381 the selector. Update call to shift_amt_for_vec_perm_mask.
12382 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12383 Use vec_perm_indices::new_expanded_vector to expand the original
12384 selector into bytes. Check whether the indices fit in the vector
12385 mode before using a variable permute.
12386 (expand_vec_perm_var): Make global.
12387 (expand_mult_highpart): Use expand_vec_perm_const.
12388 * fold-const.c: Includes vec-perm-indices.h.
12389 * tree-ssa-forwprop.c: Likewise.
12390 * tree-vect-data-refs.c: Likewise.
12391 * tree-vect-generic.c: Likewise.
12392 * tree-vect-loop.c: Likewise.
12393 * tree-vect-slp.c: Likewise.
12394 * tree-vect-stmts.c: Likewise.
12395 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12397 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12398 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12399 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12400 (aarch64_vectorize_vec_perm_const): ...this new function.
12401 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12402 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12403 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12404 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12405 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12406 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12407 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12409 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12410 check for NEON modes.
12411 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12412 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12413 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12414 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12416 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12417 the old VEC_PERM_CONST conditions.
12418 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12419 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12420 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12421 (ia64_vectorize_vec_perm_const_ok): Merge into...
12422 (ia64_vectorize_vec_perm_const): ...this new function.
12423 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12424 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12425 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12426 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12427 * config/mips/mips.c (mips_expand_vec_perm_const)
12428 (mips_vectorize_vec_perm_const_ok): Merge into...
12429 (mips_vectorize_vec_perm_const): ...this new function.
12430 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12431 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12432 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12433 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12434 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12435 (rs6000_expand_vec_perm_const): Delete.
12436 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12438 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12439 (altivec_expand_vec_perm_const_le): Take each operand individually.
12440 Operate on constant selectors rather than rtxes.
12441 (altivec_expand_vec_perm_const): Likewise. Update call to
12442 altivec_expand_vec_perm_const_le.
12443 (rs6000_expand_vec_perm_const): Delete.
12444 (rs6000_vectorize_vec_perm_const_ok): Delete.
12445 (rs6000_vectorize_vec_perm_const): New function.
12446 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12447 an element count and rtx array.
12448 (rs6000_expand_extract_even): Update call accordingly.
12449 (rs6000_expand_interleave): Likewise.
12450 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12451 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12452 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12453 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12454 (rs6000_expand_vec_perm_const): Delete.
12455 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12456 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12457 (altivec_expand_vec_perm_const_le): Take each operand individually.
12458 Operate on constant selectors rather than rtxes.
12459 (altivec_expand_vec_perm_const): Likewise. Update call to
12460 altivec_expand_vec_perm_const_le.
12461 (rs6000_expand_vec_perm_const): Delete.
12462 (rs6000_vectorize_vec_perm_const_ok): Delete.
12463 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12464 reference to the SPE evmerge intructions.
12465 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12466 an element count and rtx array.
12467 (rs6000_expand_extract_even): Update call accordingly.
12468 (rs6000_expand_interleave): Likewise.
12469 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12470 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12472 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12474 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12476 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12477 vector mode and that that mode matches the mode of the data
12479 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12480 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12481 directly using expand_vec_perm_1 when forcing selectors into
12483 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12485 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12487 * optabs-query.h (can_vec_perm_p): Delete.
12488 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12489 * optabs-query.c (can_vec_perm_p): Split into...
12490 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12491 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12492 particular selector is valid.
12493 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12494 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12495 (vect_grouped_load_supported): Likewise.
12496 (vect_shift_permute_load_chain): Likewise.
12497 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12498 (vect_transform_slp_perm_load): Likewise.
12499 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12500 (vectorizable_bswap): Likewise.
12501 (vect_gen_perm_mask_checked): Likewise.
12502 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12503 implementations of variable permutation vectors into account
12504 when deciding which selector to use.
12505 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12506 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12507 with a false third argument.
12508 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12509 to test whether the constant selector is valid and can_vec_perm_var_p
12510 to test whether a variable selector is valid.
12512 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12514 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12515 * optabs-query.c (can_vec_perm_p): Likewise.
12516 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12517 instead of vec_perm_indices.
12518 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12519 (vect_gen_perm_mask_checked): Likewise,
12520 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12521 (vect_gen_perm_mask_checked): Likewise,
12523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12525 * optabs-query.h (qimode_for_vec_perm): Declare.
12526 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12527 (qimode_for_vec_perm): ...this new function.
12528 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12530 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12532 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12533 does not have a conditional at the top.
12535 2018-01-02 Richard Biener <rguenther@suse.de>
12537 * ipa-inline.c (big_speedup_p): Fix expression.
12539 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12542 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12545 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12549 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12550 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12551 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12552 cond_taken_branch_cost 3->4.
12554 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12556 PR tree-optimization/83581
12557 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12558 TODO_cleanup_cfg if any changes have been made.
12560 PR middle-end/83608
12561 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12562 convert_modes if target mode has the right side, but different mode
12565 PR middle-end/83609
12566 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12567 last argument when extracting from CONCAT. If either from_real or
12568 from_imag is NULL, use expansion through memory. If result is not
12569 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12570 the parts directly to inner mode, if even that fails, use expansion
12573 PR middle-end/83623
12574 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12575 check for bswap in mode rather than HImode and use that in expand_unop
12578 Copyright (C) 2018 Free Software Foundation, Inc.
12580 Copying and distribution of this file, with or without modification,
12581 are permitted in any medium without royalty provided the copyright
12582 notice and this notice are preserved.