1 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
3 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4 Reformat alternatives and attributes so it is easier to identify
5 which constraints/attributes go with which instruction.
6 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7 (mov<mode>_softfloat32, FMOVE64): Likewise.
8 (mov<mode>_hardfloat64, FMOVE64): Likewise.
9 (mov<mode>_softfloat64, FMOVE64): Likewise.
11 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
13 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15 (Basic PowerPC Built-in Functions): The new name of the
16 subsection previously known as "PowerPC Built-in Functions".
17 (Basic PowerPC Built-in Functions Available on all Configurations):
19 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
20 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
21 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
22 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
24 2018-05-11 Martin Jambor <mjambor@suse.cz>
27 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
30 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
33 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
35 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
37 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
38 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
39 (ix86_handle_option): Handle -mwaitpkg.
40 * config.gcc: New header.
41 * config/i386/cpuid.h (bit_WAITPKG): New bit.
42 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
43 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
45 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
46 OPTION_MASK_ISA_WAITPKG.
47 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
48 (ix86_option_override_internal): Add PTA_WAITPKG.
49 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
50 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
52 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
53 __builtin_ia32_umwait and __builtin_ia32_tpause.
54 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
55 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
56 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
57 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
59 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
60 * config/i386/i386.opt: Add -mwaitpkg.
61 * config/i386/waitpkgintrin.h: New file.
62 * config/i386/x86intrin.h: New header.
63 * doc/invoke.texi: Add -mwaitpkg.
65 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
68 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
70 (cortex-m0): Use armv6s-m isa.
71 (cortex-m0plus): Likewise.
72 (cortex-m1): Likewise.
73 (cortex-m0.small-multiply): Likewise.
74 (cortex-m0plus.small-multiply): Likewise.
75 (cortex-m1.small-multiply): Likewise.
77 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
78 Jakub Jelinek <jakub@redhat.com>
80 PR tree-optimization/85692
81 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
82 source permute as well.
84 2018-05-11 Martin Liska <mliska@suse.cz>
87 * doc/extend.texi: Document LLVM style format for no_sanitize
90 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
92 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
93 mode_supports_vsx_dform_quad to mode_supports_dq_form.
94 (mode_supports_vsx_dform_quad): Likewise.
95 (mode_supports_vmx_dform): Move these functions to be next to the
96 other mode_supports functions.
97 (mode_supports_dq_form): Likewise.
98 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
99 mode_supports_dq_form.
100 (reg_offset_addressing_ok_p): Likewise.
101 (offsettable_ok_by_alignment): Likewise.
102 (rs6000_legitimate_offset_address_p): Likewise.
103 (legitimate_lo_sum_address_p): Likewise.
104 (rs6000_legitimize_address): Likewise.
105 (rs6000_legitimize_reload_address): Likewise.
106 (rs6000_secondary_reload_inner): Likewise.
107 (rs6000_preferred_reload_class): Likewise.
108 (rs6000_output_move_128bit): Likewise.
110 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
112 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
113 Generate SImode target register for null target.
114 <case IX86_BUILTIN_XGETBV>: Ditto.
115 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
116 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
118 2018-05-10 Carl Love <cel@us.ibm.com>
120 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
121 dcbtt and dcbtstt if operands[2] is 0.
123 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
126 * config/i386/sse.md (usadv64qi): New expander.
128 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
130 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
131 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
132 -maltivec=be support.
133 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
134 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
135 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
136 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
137 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
138 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
139 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
140 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
141 altivec_vsumsws): Adjust.
142 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
143 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
144 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
146 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
147 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
148 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
149 (altivec_lve<VI_char>x): Delete expand.
150 (*altivec_lve<VI_char>x_internal): Rename to...
151 (altivec_lve<VI_char>x): ... this.
152 (altivec_lvxl_<mode>): Delete expand.
153 (*altivec_lvxl_<mode>_internal): Rename to ...
154 (altivec_lvxl_<mode>): ... this.
155 (altivec_stvxl_<mode>): Delete expand.
156 (*altivec_stvxl_<mode>_internal): Rename to ...
157 (altivec_stvxl_<mode>): ... this.
158 (altivec_stve<VI_char>x): Delete expand.
159 (*altivec_stve<VI_char>x_internal): Rename to ...
160 (altivec_stve<VI_char>x): ... this.
161 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
162 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
163 reduc_plus_scal_<mode>): Adjust.
164 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
166 (rs6000_cpu_cpp_builtins): Adjust.
167 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
168 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
169 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
170 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
171 -maltivec=be support.
172 (rs6000_split_vec_extract_var): Adjust.
173 (rs6000_split_v4si_init): Adjust.
174 (swap_selector_for_mode): Delete.
175 (altivec_expand_lvx_be, altivec_expand_stvx_be,
176 altivec_expand_stvex_be): Delete.
177 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
178 -maltivec=be support.
179 (rs6000_gimple_fold_builtin): Ditto.
180 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
182 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
183 (TARGET_DIRECT_MOVE_64BIT): Adjust.
184 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
185 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
186 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
187 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
188 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
189 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
190 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
191 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
192 anonymous split): Adjust.
193 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
194 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
196 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
198 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
199 when --with-gxx-include-dir is also specified.
200 * configure: Regenerate.
202 2018-05-09 Jim Wilson <jimw@sifive.com>
205 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
206 * config/riscv/t-withmultilib: New.
207 * config/riscv/withmultilib.h: New.
208 * doc/install.texi: Document RISC-V --with-multilib-list support.
210 2018-05-09 Richard Biener <rguenther@suse.de>
212 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
214 (vect_bb_vectorization_profitable_p): Adjust. Compute
215 actual scalar cost using the cost vector and the add_stmt_cost
218 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
220 PR rtl-optimization/85645
221 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
222 in the REG_CFA_REGISTER note for LR, don't leave it empty.
224 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
226 PR rtl-optimization/85645
227 * shrink-wrap.c (spread_components): Return a boolean saying if
228 anything was changed.
229 (try_shrink_wrapping_separate): Iterate spread_components until
230 nothing changes anymore.
232 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
234 PR rtl-optimization/85645
235 * regrename.c (build_def_use): Also kill the chains that include the
236 destination of a REG_CFA_REGISTER note.
238 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
240 PR rtl-optimization/85645
241 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
242 insn that has a REG_CFA_REGISTER note.
244 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
246 * cfgexpand.c (expand_clobber): New function.
247 (expand_gimple_stmt_1): Use it.
248 * tree-vect-stmts.c (vect_clobber_variable): New function,
250 (vectorizable_simd_clone_call): ...here.
251 (vectorizable_store): Emit a clobber either side of an
252 IFN_STORE_LANES sequence.
253 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
255 2018-05-09 Tom de Vries <tom@codesourcery.com>
258 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
259 (define_insn "trap_if_false"): Add exit after trap.
261 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
263 PR rtl-optimization/85638
264 * bb-reorder.c: Include common/common-target.h.
265 (create_forwarder_block): New function extracted from...
266 (fix_up_crossing_landing_pad): ...here. Rename into...
267 (dw2_fix_up_crossing_landing_pad): ...this.
268 (sjlj_fix_up_crossing_landing_pad): New function.
269 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
270 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
271 from both partitions and exit the loop after one iteration.
273 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
276 * doc/extend.texi (PowerPC Built-in Functions): Rename this
278 (Basic PowerPC Built-in Functions): The new name of the
279 subsection previously known as "PowerPC Built-in Functions".
280 (Basic PowerPC Built-in Functions Available on all Configurations):
282 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
284 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
286 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
288 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
291 2018-05-08 Jim Wilson <jimw@sifive.com>
293 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
294 (LD_EMUL_SUFFIX): New.
297 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
299 * doc/extend.texi (PowerPC Built-in Functions): Rename this
301 (Basic PowerPC Built-in Functions): The new name of the
302 subsection previously known as "PowerPC Built-in Functions".
303 (Basic PowerPC Built-in Functions Available on all Configurations):
305 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
307 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
309 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
311 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
314 2018-05-08 Jakub Jelinek <jakub@redhat.com>
317 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
318 after cmpelim optimization.
320 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
322 * config.gcc: Support "goldmont".
323 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
324 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
326 * config/i386/i386.c (m_GOLDMONT): Define.
327 (processor_target_table): Add "goldmont".
328 (PTA_GOLDMONT): Define.
329 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
330 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
331 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
332 (fold_builtin_cpu): Add "goldmont".
333 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
334 (ix86_option_override_internal): Add "goldmont".
335 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
336 (processor_type): Add PROCESSOR_GOLDMONT.
337 * config/i386/i386.md: Add CPU "glm".
338 * config/i386/glm.md: New file.
339 * config/i386/x86-tune.def: Add m_GOLDMONT.
340 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
342 2018-05-08 Jakub Jelinek <jakub@redhat.com>
345 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
347 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
348 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
349 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
352 * config/i386/i386.c (ix86_fold_builtin): Handle
353 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
356 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
357 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
359 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
362 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
363 (check_arch): Likewise.
364 (check_fpu): Return the result rather than printing it.
365 (end arch): Fix operator precedence.
367 (END): Print the result from check_fpu.
369 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
370 Alan Hayward <alan.hayward@arm.com>
371 David Sherwood <david.sherwood@arm.com>
373 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
374 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
375 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
376 (*fcmuo<mode>_and): New patterns.
378 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
380 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
381 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
382 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
383 (cmp_op, sve_imm_con): New code attributes.
384 (SVE_COND_INT_CMP, imm_con): Delete.
385 (cmp_op): Remove above unspecs from int attribute.
386 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
388 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
389 comparison-specific unspecs.
390 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
391 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
392 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
393 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
394 (*vec_fcm<cmp_op><mode>): Rename to...
395 (*fcm<cmp_op><mode>): ...this and adjust likewise.
396 (*vec_fcmuo<mode>): Rename to...
397 (*fcmuo<mode>): ...this and adjust likewise.
398 (*pred_fcm<cmp_op><mode>): New pattern.
399 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
400 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
402 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
404 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
405 (aarch64_emit_sve_predicated_cond): New function.
406 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
407 (aarch64_emit_unspec_cond_or): Replace with...
408 (aarch64_emit_sve_or_conds): ...this new function. Use
409 aarch64_emit_sve_ptrue_op for the individual comparisons and
410 aarch64_emit_binop to OR them together.
411 (aarch64_emit_inverted_unspec_cond): Replace with...
412 (aarch64_emit_sve_inverted_cond): ...this new function. Use
413 aarch64_emit_sve_ptrue_op for the comparison and
414 aarch64_emit_unop to invert the result.
415 (aarch64_expand_sve_vec_cmp_float): Update after the above
416 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
418 2018-05-07 Nathan Sidwell <nathan@acm.org>
420 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
421 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
422 (Backwards Compatibility): Likewise.
424 2018-05-07 Luis Machado <luis.machado@linaro.org>
428 2018-05-07 Luis Machado <luis.machado@linaro.org>
430 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
431 <prefetch_dynamic_strides>: New const bool field.
432 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
433 prefetch_dynamic_strides.
434 (exynosm1_prefetch_tune): Likewise.
435 (thunderxt88_prefetch_tune): Likewise.
436 (thunderx_prefetch_tune): Likewise.
437 (thunderx2t99_prefetch_tune): Likewise.
438 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
439 (aarch64_override_options_internal): Update to set
440 PARAM_PREFETCH_DYNAMIC_STRIDES.
441 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
442 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
443 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
444 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
445 prefetch-dynamic-strides setting.
447 2018-05-07 Luis Machado <luis.machado@linaro.org>
449 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
450 <minimum_stride>: New const int field.
451 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
452 minimum_stride field.
453 (exynosm1_prefetch_tune): Likewise.
454 (thunderxt88_prefetch_tune): Likewise.
455 (thunderx_prefetch_tune): Likewise.
456 (thunderx2t99_prefetch_tune): Likewise.
457 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
458 (aarch64_override_options_internal): Update to set
459 PARAM_PREFETCH_MINIMUM_STRIDE.
460 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
461 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
462 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
463 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
464 stride is constant and is below the minimum stride threshold.
466 2018-05-07 Luis Machado <luis.machado@linaro.org>
468 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
471 2018-05-07 Luis Machado <luis.machado@linaro.org>
473 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
474 <prefetch_dynamic_strides>: New const bool field.
475 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
476 prefetch_dynamic_strides.
477 (exynosm1_prefetch_tune): Likewise.
478 (thunderxt88_prefetch_tune): Likewise.
479 (thunderx_prefetch_tune): Likewise.
480 (thunderx2t99_prefetch_tune): Likewise.
481 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
482 (aarch64_override_options_internal): Update to set
483 PARAM_PREFETCH_DYNAMIC_STRIDES.
484 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
485 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
486 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
487 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
488 prefetch-dynamic-strides setting.
490 2018-05-07 Luis Machado <luis.machado@linaro.org>
492 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
493 <minimum_stride>: New const int field.
494 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
495 minimum_stride field.
496 (exynosm1_prefetch_tune): Likewise.
497 (thunderxt88_prefetch_tune): Likewise.
498 (thunderx_prefetch_tune): Likewise.
499 (thunderx2t99_prefetch_tune): Likewise.
500 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
501 (aarch64_override_options_internal): Update to set
502 PARAM_PREFETCH_MINIMUM_STRIDE.
503 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
504 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
505 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
506 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
507 stride is constant and is below the minimum stride threshold.
509 2018-05-06 Jakub Jelinek <jakub@redhat.com>
512 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
513 the type is addressable. Don't force op into register if it has
516 2018-05-05 Roland McGrath <mcgrathr@google.com>
519 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
520 any section for which we don't know a specific type it should have,
521 regardless of name. Previously this was done only for the exact
522 names ".init_array", ".fini_array", and ".preinit_array".
523 (default_elf_asm_named_section): Add comment about
524 relationship with default_section_type_flags and SECTION_NOTYPE.
525 (get_section): Don't consider it a type conflict if one side has
526 SECTION_NOTYPE and the other doesn't, as long as neither has the
527 SECTION_BSS et al used in the default_section_type_flags logic.
529 2018-05-05 Tom de Vries <tom@codesourcery.com>
532 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
533 (workaround_barsyncs): New function.
534 (nvptx_reorg): Use workaround_barsyncs.
535 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
536 (define_expand "nvptx_membar_cta"): New define_expand.
537 (define_insn "*nvptx_membar_cta"): New insn.
539 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
541 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
542 To improve optimization opportunities.
543 * builtin-types.def: The new needed builtin types for the above.
545 2018-05-04 Richard Biener <rguenther@suse.de>
547 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
548 * gimple-ssa-store-merging.c
549 (imm_store_chain_info::output_merged_store): Remove redundant create,
550 release split_store vector contents on failure.
551 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
552 scalar stmt vector on cache hit.
554 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
556 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
558 * config.gcc (powerpc-xilinx-eabi*): Remove.
559 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
561 (fusion_addis_mem_combo_load): Ditto.
562 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
564 (rs6000_cpu_cpp_builtins): Ditto.
565 * config/rs6000/rs6000-linux.c
566 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
567 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
568 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
570 (rs6000_setup_reg_addr_masks): Ditto.
571 (rs6000_init_hard_regno_mode_ok): Ditto.
572 (rs6000_option_override_internal): Ditto.
573 (legitimate_lo_sum_address_p): Ditto.
574 (rs6000_legitimize_address): Ditto.
575 (rs6000_legitimize_reload_address): Ditto.
576 (rs6000_legitimate_address_p): Ditto.
577 (abi_v4_pass_in_fpr): Ditto.
578 (setup_incoming_varargs): Ditto.
579 (rs6000_gimplify_va_arg): Ditto.
580 (rs6000_split_multireg_move): Ditto.
581 (rs6000_savres_strategy): Ditto.
582 (rs6000_emit_prologue_components): Ditto.
583 (rs6000_emit_epilogue_components): Ditto.
584 (rs6000_emit_prologue): Ditto.
585 (rs6000_emit_epilogue): Ditto.
586 (rs6000_elf_file_end): Ditto.
587 (rs6000_function_value): Ditto.
588 (rs6000_libcall_value): Ditto.
589 * config/rs6000/rs6000.h: Ditto.
590 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
591 (TARGET_MINMAX): ... this. New.
592 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
593 * config/rs6000/rs6000.md: Remove Xilinx FP support.
594 (*movsi_internal1_single): Delete.
595 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
596 mfpu=, mxilinx-fpu): Delete.
597 * config/rs6000/singlefp.h: Delete.
598 * config/rs6000/sysv4.h: Remove Xilinx FP support.
599 * config/rs6000/t-rs6000: Ditto.
600 * config/rs6000/t-xilinx: Delete.
601 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
602 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
603 (VStype_simple): Delete.
604 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
605 * config/rs6000/xfpu.h: Delete.
606 * config/rs6000/xfpu.md: Delete.
607 * config/rs6000/xilinx.h: Delete.
608 * config/rs6000/xilinx.opt: Delete.
609 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
610 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
612 2018-05-04 Tom de Vries <tom@codesourcery.com>
615 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
618 2018-05-04 Richard Biener <rguenther@suse.de>
621 * tree-complex.c (update_complex_assignment): We are always in SSA form.
622 (expand_complex_div_wide): Likewise.
623 (expand_complex_operations_1): Likewise.
624 (expand_complex_libcall): Preserve EH info of the original stmt.
625 (tree_lower_complex): Handle removed blocks.
626 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
627 on complex multiplication and division libcall builtins.
629 2018-05-04 Richard Biener <rguenther@suse.de>
632 * fold-const.c (negate_expr_p): Restrict negation of operand
633 zero of a division to when we know that can happen without
635 (fold_negate_expr_1): Likewise.
637 2018-05-04 Jakub Jelinek <jakub@redhat.com>
640 * real.h (real_nextafter): Declare.
641 * real.c (real_nextafter): New function.
642 * fold-const-call.c (fold_const_nextafter): New function.
643 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
645 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
646 even when arg1_mode is different from arg0_mode.
648 2018-05-03 Nathan Sidwell <nathan@acm.org>
650 * doc/extend.texi (Deprecated Features): Remove
652 (Backwards Compatibility): Likewise.
653 * doc/invoke.texi (C++ Language Options): Likewise.
654 (C++ Dialect Options): Likewise.
656 2018-05-03 Jakub Jelinek <jakub@redhat.com>
659 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
660 _mm512_mask_mullox_epi64): New intrinsics.
662 2018-05-03 Tom de Vries <tom@codesourcery.com>
665 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
666 dump files): Add offload-tree.
668 2018-05-03 Richard Biener <rguenther@suse.de>
670 PR tree-optimization/85615
671 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
672 to loops not nested in BBs loop father to avoid creating multi-entry
675 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
677 PR tree-optimization/70291
678 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
679 arguments. Change return type to tree. Emit libcall as a new
680 statement rather than replacing existing one when inplace_p is true.
681 (expand_complex_multiplication_components): New function.
682 (expand_complex_multiplication): Expand floating-point complex
683 multiplication using the above.
684 (expand_complex_division): Rename inner_type parameter to type.
685 Update expand_complex_libcall call-site.
686 (expand_complex_operations_1): Update expand_complex_multiplication
687 and expand_complex_division call-sites.
689 2018-05-02 Jakub Jelinek <jakub@redhat.com>
692 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
693 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
694 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
695 the highest significant bit of the shift count mask is clear. In
696 check whether and[sq]i3 is needed verify that all significant bits
697 of the shift count other than the highest are set.
699 2018-05-02 Tom de Vries <tom@codesourcery.com>
702 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
703 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
704 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
705 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
706 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
707 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
708 __builtin_goacc_parlevel_size.
710 2018-05-02 Richard Biener <rguenther@suse.de>
712 PR tree-optimization/85597
713 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
714 do not use split vect_get_vec_defs call but call vect_get_slp_defs
717 2018-05-02 Tom de Vries <tom@codesourcery.com>
720 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
721 dump files): Add ltrans-tree.
723 2018-05-02 Tom de Vries <tom@codesourcery.com>
726 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
727 dump files): Add wpa-ipa.
729 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
731 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
732 powerpc*-*-linux*paired* target.
733 * config/rs6000/750cl.h: Delete.
734 * config/rs6000/paired.h: Delete.
735 * config/rs6000/paired.md: Delete.
736 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
738 * config/rs6000/rs6000-builtin.def: Remove paired float support.
739 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
740 comment. Remove paired float support.
741 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
742 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
744 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
745 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
747 * config/rs6000/rs6000.c: Remove paired float support.
748 (paired_expand_vector_init, paired_expand_vector_move,
749 paired_emit_vector_compare, paired_emit_vector_cond_expr,
750 (paired_expand_lv_builtin, paired_expand_stv_builtin,
751 paired_expand_builtin, paired_expand_predicate_builtin,
752 paired_init_builtins): Delete.
753 * config/rs6000/rs6000.h: Remove paired float support.
754 * config/rs6000/rs6000.md: Remove paired float support.
755 (move_from_CR_ov_bit): Delete.
756 * config/rs6000/rs6000.opt (mpaired): Delete.
757 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
758 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
760 2018-05-02 Richard Biener <rguenther@suse.de>
763 * gimplify.c (gimplify_save_expr): When in SSA form allow
764 SAVE_EXPRs to compute to SSA vars.
766 2018-05-02 Jakub Jelinek <jakub@redhat.com>
769 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
770 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
771 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
772 clobber operands[2], instead use a new pseudo. Formatting fixes.
774 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
776 PR tree-optimization/85586
777 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
778 exit early for statements in the same group if the accesses are
781 2018-05-02 Tom de Vries <tom@codesourcery.com>
784 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
787 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
789 PR tree-optimization/85143
790 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
792 2018-05-01 Tom de Vries <tom@codesourcery.com>
795 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
796 not found" error message.
798 2018-05-01 Tom de Vries <tom@codesourcery.com>
801 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
802 * vec.c (test_ordered_remove_if): New function.
803 (vec_c_tests): Call test_ordered_remove_if.
804 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
805 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
806 * tree-vect-patterns.c (vect_pattern_recog_1): Use
807 VEC_ORDERED_REMOVE_IF.
809 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
811 PR tree-optimization/82665
812 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
813 pointer subtraction where arguments come from a memchr call.
815 2018-05-01 Jakub Jelinek <jakub@redhat.com>
817 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
818 --push-state --as-needed and --pop-state instead of --as-needed and
819 --no-as-needed if ld supports it.
820 * configure: Regenerated.
823 * doc/install.texi2html: Replace _002d with - and _002a with * in
824 generated html files using sed.
826 2018-04-30 David Malcolm <dmalcolm@redhat.com>
829 * gcc-rich-location.c (blank_line_before_p): New function.
830 (use_new_line): New function.
831 (gcc_rich_location::add_fixit_insert_formatted): New function.
832 * gcc-rich-location.h
833 (gcc_rich_location::add_fixit_insert_formatted): New function.
835 2018-04-30 David Malcolm <dmalcolm@redhat.com>
837 * selftest.c (assert_streq): Rename "expected" and "actual" to
838 "val1" and "val2". Extend NULL-handling to cover both inputs
839 symmetrically, while still requiring both to be non-NULL for a pass.
840 * selftest.h (assert_streq): Rename "expected" and "actual" to
842 (ASSERT_EQ): Likewise.
843 (ASSERT_EQ_AT): Likewise.
844 (ASSERT_KNOWN_EQ): Likewise.
845 (ASSERT_KNOWN_EQ_AT): Likewise.
846 (ASSERT_NE): Likewise.
847 (ASSERT_MAYBE_NE): Likewise.
848 (ASSERT_MAYBE_NE_AT): Likewise.
849 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
850 the assertion to pass.
851 (ASSERT_STREQ_AT): Likewise.
853 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
855 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
856 interaction with -pie.
858 2018-04-30 David Malcolm <dmalcolm@redhat.com>
860 * selftest.h: Fix alphabetization of per-source-file selftest
863 2018-04-30 Jason Merrill <jason@redhat.com>
865 PR c++/61982 - dead stores to destroyed objects.
866 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
869 2018-04-30 Jason Merrill <jason@redhat.com>
871 * tree.c (build_clobber): New.
872 * tree.h: Declare it.
873 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
875 2018-04-30 David Malcolm <dmalcolm@redhat.com>
877 * diagnostic-show-locus.c (layout::layout): Update for
878 location_get_source_line returning a char_span.
879 (struct char_span): Move to input.h.
880 (struct correction): Update for fields in char_span becoming
882 (struct source_line): Update for location_get_source_line
883 returning a char_span.
884 (layout::print_line): Likewise.
885 * edit-context.c (edited_file::print_content): Likewise.
886 (edited_file::print_diff_hunk): Likewise.
887 (edited_file::print_run_of_changed_lines): Likewise.
888 (edited_file::get_num_lines): Likewise.
889 (edited_line::edited_line): Likewise.
890 * final.c (asm_show_source): Likewise.
891 * input.c (location_get_source_line): Convert return type
892 from const char * to char_span, losing the final "line_len"
894 (dump_location_info): Update for the above.
895 (get_substring_ranges_for_loc): Likewise. Use a char_span
896 when handling the literal within the line.
897 (test_reading_source_line): Update for location_get_source_line
898 returning a char_span.
899 * input.h (class char_span): Move here from
900 diagnostic-show-locus.c, converting from a struct to a class.
901 Make data members private.
902 (char_span::operator bool): New.
903 (char_span::length): New.
904 (char_span::get_buffer): New.
905 (char_span::operator[]): New.
906 (char_span::subspan): Make const.
907 (char_span::xstrdup): New.
908 (location_get_source_line): Convert return type from const char *
909 to char_span, losing the final "line_size" param.
911 2018-04-30 Jan Hubicka <jh@suse.cz>
913 * lto-wrapper.c (ltrans_priorities): New static var.
915 (run_gcc): Read priorities and if doing parallel build order
916 the Makefile by them.
918 2018-04-30 David Malcolm <dmalcolm@redhat.com>
920 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
922 2018-04-30 Richard Biener <rguenther@suse.de>
924 * tree-cfg.c (verify_address): Remove base argument, add
925 flag whether to check TREE_ADDRESSABLE and do that.
926 (verify_expr): Remove.
927 (verify_types_in_gimple_reference): Add pieces from verify_expr.
928 (verify_gimple_assign_single): Likewise.
929 (verify_gimple_switch): Likewise.
930 (verify_expr_location_1): Dereference tp once. Add (disabled)
931 piece from verify_expr.
932 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
934 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
936 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
938 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
940 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
941 (small_data_pattern): Likewise.
942 (arc_rewrite_small_data): Likewise.
943 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
944 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
945 (get_symbol_alignment): New function.
946 (legitimate_small_data_address_p): Likewise.
947 (legitimate_scaled_address): Update, call
948 legitimate_small_data_address_p.
949 (output_sdata): New static variable.
950 (arc_print_operand): Update how we handle small data operands.
951 (arc_print_operand_address): Likewise.
952 (arc_legitimate_address_p): Update, use
953 legitimate_small_data_address_p.
954 (arc_rewrite_small_data_p): Remove.
955 (arc_rewrite_small_data_1): Likewise.
956 (arc_rewrite_small_data): Likewise.
957 (small_data_pattern): Likewise.
958 (compact_sda_memory_operand): Update to use
959 legitimate_small_data_address_p and get_symbol_alignment.
960 (prepare_move_operands): Don't rewite sdata pattern.
961 (prepare_extend_operands): Remove.
962 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
964 (zero_extendqisi2): Likewise.
965 (zero_extendhisi2): Likewise.
966 (extendqihi2): Likewise.
967 (extendqisi2): Likewise.
968 (extendhisi2): Likewise.
972 * config/arc/constraints.md (Usd): Change it to memory constraint.
974 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
976 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
977 as source of std instructions.
978 * config/arc/arc.md (movsi_insn): Update pattern predicate to
979 allow 6-bit constants as source for store instructions.
980 (movdi_insn): Update instruction pattern to allow 6-bit constants
981 as source for store instructions.
983 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
985 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
987 2018-04-30 Nathan Sidwell <nathan@acm.org>
988 Sandra Loosemore <sandra@codesourcery.com>
990 * dumpfile.c (dump_open): Allow '-' for stdout.
991 * doc/invoke.texi (Developer Options): Document dump filename
992 determination early. Document stdin/stdout selection.
994 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
996 Microblaze Target: PIC data text relative
998 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
999 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1001 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1002 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1003 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1004 New addressing mode for data-text relative position indepenedent code.
1005 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1006 'ADDRESS_SYMBOLIC_TXT_REL'.
1007 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1008 (microblaze_legitimate_pic_operand): Exclude function calls from
1009 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1010 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1012 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1013 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1014 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1015 for 'address + offset'.
1016 (microblaze_expand_prologue): Add new function prologue call for
1018 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1019 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1020 table in case of TARGET_PIC_DATA_TEXT_REL.
1021 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1022 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1023 Add new macros 'UNSPEC_TEXT',
1024 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1025 + exclude function calls from 'UNSPEC_PLT' in case of data text
1027 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1028 new target hook for generating address diff vector tables in case of
1030 * doc/tm.texi : Regenerate.
1031 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1032 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1033 of addr diff vector generation.
1034 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1035 target hook definition.
1036 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1037 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1038 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1039 option for data text relative.
1041 2018-04-30 Richard Biener <rguenther@suse.de>
1043 * tree-chrec.h (evolution_function_is_constant_p): Remove
1045 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1047 2018-04-30 Richard Biener <rguenther@suse.de>
1050 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1052 2018-04-30 Richard Biener <rguenther@suse.de>
1054 PR tree-optimization/28364
1055 PR tree-optimization/85275
1056 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1057 copying first exit test.
1059 2018-04-28 Mark Wielaard <mark@klomp.org>
1061 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1063 (dwarf_AT): Handle DW_AT_addr_base.
1064 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1066 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1069 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1070 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1071 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1072 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1074 2018-04-28 Richard Biener <rguenther@suse.de>
1076 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1077 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1078 to reflect use. Only add interesting stmts.
1080 2018-04-27 Martin Jambor <mjambor@suse.cz>
1083 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1084 the jump function allows for passing through aggregate values.
1086 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1088 * input.h (in_system_header_at): Convert from macro to inline
1090 (from_macro_expansion_at): Likewise.
1091 (from_macro_definition_at): Likewise.
1093 2018-04-27 Jeff Law <law@redhat.com>
1095 * config.gcc: Mark tile* targets as deprecated/obsolete.
1097 2018-04-27 Richard Biener <rguenther@suse.de>
1099 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1102 2018-04-27 Richard Biener <rguenther@suse.de>
1104 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1106 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1108 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1109 with Yd constraint. Set "preferred_for_speed" attribute from
1110 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1112 (*movdi_internal): Ditto.
1113 (movti_interunit splitters): Remove
1114 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1115 (movdi_interunit splitters): Ditto.
1116 * config/i386/constraints.md (Ye): Remove.
1117 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1119 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1122 * config/aarch64/constraints.md (Usg): Limit to 31.
1125 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1127 PR tree-optimization/85529
1128 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1129 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1130 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1131 zero extension or masking of the MSB bit.
1132 (optimize_range_tests): Add FIRST_BB argument, pass it through
1133 to optimize_range_tests_var_bound.
1134 (maybe_optimize_range_tests, reassociate_bb): Adjust
1135 optimize_range_tests callers.
1137 2018-04-26 Richard Biener <rguenther@suse.de>
1138 Jakub Jelinek <jakub@redhat.com>
1140 * cgraph.h (symbol_table): Just declare debug method here.
1141 * symtab.c (symbol_table::debug): Define.
1143 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1145 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1147 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1149 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1150 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1151 (*movdi_internal): Substitute Yi and Yj constraint with x
1152 and Ym and Yn constraint with y constraint. Update "isa"
1153 attribute and set "preferred_for_speed" attribute from
1154 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1155 (*movsi_internal): Ditto.
1156 (*movdf_internal): Ditto.
1157 (*movsf_internal): Ditto.
1158 (*zero_extendsidi2): Ditto.
1159 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1160 (sse2_loadld): Ditto.
1161 (*vec_extract<ssevecmodelower>_0): Ditto.
1162 (*vec_extractv4si_0_zext_sse4): Ditto.
1163 (vec_concatv2di): Ditto.
1164 (*vec_dup<mode>): Ditto.
1165 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1166 * config/i386/constraints.md (Yi): Remove.
1171 2018-04-26 Nathan Sidwell <nathan@acm.org>
1173 * dumpfile.c (dump_open): New.
1174 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1175 (dump_finish): Detect stdio/stderr by value not name.
1177 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1179 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1181 2018-04-26 Tom de Vries <tom@codesourcery.com>
1184 * config/nvptx/nvptx.c (verify_neutering_jumps)
1185 (verify_neutering_labels): New function
1186 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1188 2018-04-26 Tom de Vries <tom@codesourcery.com>
1191 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1192 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1195 2018-04-26 Richard Biener <rguenther@suse.de>
1196 Tom de Vries <tom@codesourcery.com>
1199 * lto-streamer-out.c (output_function): Fixup loops if required to match
1200 discovery done in the reader.
1202 2018-04-26 Richard Biener <rguenther@suse.de>
1204 PR tree-optimization/85116
1205 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1206 have a loop exit from the single latch predecessor. Remove
1207 case of header with just condition.
1208 (ch_base::copy_headers): Exclude infinite loops from any
1210 (pass_ch::execute): Record exits.
1212 2018-04-26 Richard Biener <rguenther@suse.de>
1214 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1215 prologue cost vector and pass it to vect_get_load_cost.
1216 (vect_get_peeling_costs_all_drs): Likewise.
1217 (vect_peeling_hash_get_lowest_cost): Likewise.
1218 (vect_enhance_data_refs_alignment): Likewise.
1220 2018-04-26 Richard Biener <rguenther@suse.de>
1223 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1224 checking of integer<->pointer conversions.
1225 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1226 sign-/zero-extending pointer types.
1227 (expand_omp_for_static_chunk): Likewise.
1229 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1230 Jean Lee <xiaoyur347@gmail.com>
1232 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1233 (TARGET_ASAN_SHADOW_OFFSET): Define.
1234 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1235 true for -fsanitize=address.
1237 2018-04-25 Mark Wielaard <mark@klomp.org>
1239 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1242 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1244 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1245 than "alu", remove explicit "memory" and "imm_disp" attributes.
1246 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1249 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1250 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1251 gen_lowpart_no_emit.
1253 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1256 * config/i386/i386.c (ix86_expand_builtin): Change memory
1257 operand to XI, extend p0 to Pmode.
1258 * config/i386/i386.md: Change unspec volatile and operand
1259 1 mode to XI, change operand 0 mode to P.
1261 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1263 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1264 GET_MODE_MASK before any checking.
1265 (nds32_can_use_bset_p): Likewise.
1266 (nds32_can_use_btgl_p): Likewise.
1268 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1270 * config/nds32/nds32-doubleword.md: New define_split pattern for
1271 illegal register number.
1273 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1275 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1277 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1279 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1281 2018-04-25 Richard Biener <rguenther@suse.de>
1283 * lto-streamer.h (LTO_major_version): Bump to 8.
1285 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1287 * BASE-VER: Set to 9.0.0.
1289 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1291 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1292 in __abskf2 and __powikf2.
1294 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1297 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1298 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1299 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1300 Use the above on operand 2. Reindent.
1301 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1303 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1306 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1308 * config/i386/i386.opt (mcet): Removed.
1309 * doc/install.texi: Remove -mcet documentation.
1310 * doc/invoke.texi: Likewise.
1312 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1315 * doc/install.texi: Remove -mcet from bootstrap-cet.
1317 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1320 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1321 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1325 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1326 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1327 containing a CONST_VECTOR.
1329 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1331 * doc/install.texi: Update newlib dependency for nvptx.
1333 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1336 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1337 instead of INTVAL when shifting x left.
1339 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1341 PR tree-optimization/85478
1342 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1343 vect_grouped_store_supported for single element vectors.
1345 2018-04-24 Richard Biener <rguenther@suse.de>
1348 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1349 load cost increase to the case of non-constant step.
1351 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1354 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1355 destination if any_malformed_asm.
1357 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1360 * expr.c (store_field): In the bitfield case, if the value comes from
1361 a function call and is returned in registers by means of a PARALLEL,
1362 do not change the mode of the temporary unless BLKmode and VOIDmode.
1364 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1366 PR rtl-optimization/85423
1367 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1368 dependencies to debug insns when the previous insn is non-debug.
1370 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1372 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1373 enums into a single definition.
1374 (fls): Fix predicates and printing.
1377 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1379 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1380 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1381 and short u6 immediate.
1382 (check_if_valid_sleep_operand): Remove.
1383 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1385 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1387 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1388 flag_always_save_lp condition.
1389 * config/nds32/nds32.opt (malways-save-lp): New option.
1391 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1393 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1394 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1395 * config/nds32/nds32.h
1396 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1397 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1399 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1401 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1402 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1404 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1405 Chung-Ju Wu <jasonwucj@gmail.com>
1407 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1409 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1410 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1412 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1414 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1416 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1418 * config/nds32/nds32-protos.h (nds32_data_alignment,
1419 nds32_local_alignment): Declare.
1420 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1421 nds32_local_alignment): New functions.
1422 (TARGET_CONSTANT_ALIGNMENT): Define.
1423 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1425 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1427 * config/nds32/nds32.c
1428 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1429 (TARGET_MODES_TIEABLE_P): Likewise.
1431 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1433 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1436 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1437 Chung-Ju Wu <jasonwucj@gmail.com>
1439 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1440 for unaligned access.
1441 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1443 * config/nds32/nds32-intrinsic.md: Likewise.
1444 * config/nds32/nds32_intrinsic.h: Likewise.
1445 * config/nds32/nds32.h (nds32_builtins): Likewise.
1446 * config/nds32/nds32.opt (munaligned-access): New option.
1447 * config/nds32/nds32.c (nds32_asm_file_start): Display
1448 flag_unaligned_access status.
1450 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1452 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1453 -mno-relax is present.
1454 * config/riscv/linux.h (LINK_SPEC): Ditto.
1456 2018-04-20 Martin Sebor <msebor@redhat.com>
1459 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1461 (gimple_fold_builtin_stxcpy_chk): Same.
1462 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1464 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1467 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1468 __powikf2 when long double is IEEE 128-bit.
1470 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1472 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1473 step to make sure stack always aligned.
1475 2018-04-20 Carl Love <cel@us.ibm.com>
1478 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1479 size check for arg0.
1481 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1482 Tom de Vries <tom@codesourcery.com>
1485 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1486 Emit insns for calls too.
1487 (nvptx_find_par): Always look for worker-level predecessor insn.
1488 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1490 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1491 (nvptx_process_pars): Propagate frames for calls.
1493 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1496 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1498 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1499 (ix86_handle_option): Don't handle OPT_mibt.
1500 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1502 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1504 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1505 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1506 (ix86_target_macros): Define __CET__ with flag_cf_protection
1507 for -fcf-protection.
1508 * config/i386/i386.c (isa2_opts): Remove -mibt.
1509 * config/i386/i386.h (TARGET_IBT): Removed.
1510 (TARGET_IBT_P): Likewise.
1511 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1512 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1513 * config/i386/i386.opt (mcet): Update help message.
1516 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1517 -mcet as an alias for -mshstk.
1519 2018-04-20 Richard Biener <rguenther@suse.de>
1522 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1523 complexity by forcing a single use of the multiply operand.
1525 2018-04-20 Martin Jambor <mjambor@suse.cz>
1528 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1529 recursion dependency to only apply to non-clones.
1531 2018-04-20 Martin Jambor <mjambor@suse.cz>
1534 * ipa-cp.c (create_specialized_node): Check that clones of
1535 self-recursive edges exist during IPA-CP.
1537 2018-04-19 Toon Moene <toon@moene.org>
1539 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1542 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1544 PR tree-optimization/85467
1545 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1546 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1547 VECTOR_CST element to type.
1549 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1552 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1553 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1554 (builtin_longjmp): Likewise.
1555 (save_stack_nonlocal): New pattern.
1556 (restore_stack_nonlocal): Likewise.
1558 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1561 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1562 Replace ASM_OUTPUT_LABEL with fprintf.
1564 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1567 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1568 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1569 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1570 define __IBT__ and __SHSTK__ for -fcf-protection.
1571 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1573 (ix86_trampoline_init): Likewise.
1574 (x86_output_mi_thunk): Likewise.
1575 (ix86_notrack_prefixed_insn_p): Likewise.
1576 (ix86_option_override_internal): Don't disallow -fcf-protection.
1577 * config/i386/i386.md (rdssp<mode>): Also enable for
1579 (incssp<mode>): Likewise.
1580 (nop_endbr): Likewise.
1581 * config/i386/i386.opt (mcet): Change help message to built-in
1585 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1586 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1587 enable CET built-in functions.
1589 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1591 * common/config/i386/i386-common.c
1592 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1593 OPTION_MASK_ISA_MOVDIRI_UNSET,
1594 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1595 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1596 * config.gcc (movdirintrin.h): New header.
1597 * config/i386/cpuid.h (bit_MOVDIRI,
1598 bit_MOVDIR64B): New bits.
1599 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1601 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1602 (VOID, PVOID, PCVOID)): New function types.
1603 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1604 __builtin_ia32_directstoreu_u64,
1605 __builtin_ia32_movdir64b): New builtins.
1606 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1607 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1609 (ix86_valid_target_attribute_inner_p): Ditto.
1610 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1611 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1612 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1613 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1614 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1615 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1616 (movdiri<mode>, movdir64b_<mode>): New.
1617 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1618 * config/i386/immintrin.h: Include movdirintrin.h.
1619 * config/i386/movdirintrin.h: New file.
1620 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1622 2018-04-19 Richard Biener <rguenther@suse.de>
1625 * cfg.c (clear_bb_flags): When loop state says we have
1626 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1628 2018-04-19 Richard Biener <rguenther@suse.de>
1630 PR tree-optimization/84737
1631 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1632 copying restrict info.
1633 (vect_setup_realignment): Use it.
1634 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1635 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1636 the first DR to all generated stores.
1637 (vectorizable_load): Likewise for loads.
1639 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1641 PR tree-optimization/85446
1642 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1643 the integral and pointer types to have the same precision.
1645 * doc/install.texi: Document --disable-cet being the default and
1648 2018-04-18 Martin Liska <mliska@suse.cz>
1650 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1653 2018-04-18 Martin Liska <mliska@suse.cz>
1656 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1659 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1660 arguments if they are comparable.
1662 2018-04-18 Martin Liska <mliska@suse.cz>
1665 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1668 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1671 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1674 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1675 ENDBR after calling __morestack.
1677 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1680 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1681 by using gcc_base_ver to generate a gcc_driver_version, and use
1682 it when generating GCC_DRIVER_NAME.
1683 * configure: Regenerate.
1685 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1688 * config.gcc: Obsolete powerpc*-*-*spe*.
1690 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1693 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1694 (stabstr_D): Change type of unum from unsigned int to
1695 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1698 2018-04-17 Jim Wilson <jimw@sifive.com>
1701 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1702 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1703 Set arg_pointer_offset after using pretend_args_size.
1705 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1707 PR rtl-optimization/85431
1708 * dse.c (record_store): Ignore zero width stores.
1711 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1712 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1713 __builtin_stack_restore rather than after it.
1714 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1715 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1716 argument instead of virtual_dynamic_stack_rtx.
1718 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1720 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1722 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1723 Add note to error message to explain internal mapping of overloaded
1724 built-in function name to non-overloaded built-in function name.
1725 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1728 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1731 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1732 where the inputs overlap with the output.
1734 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1736 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1737 (=v, v) alternative and explicit "memory" attribute.
1738 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1739 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1741 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1742 "sselog1" type instead of "sselog".
1743 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1744 "sselog". Remove explicit "memory" attribute.
1745 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1746 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1748 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1749 "sselog1" type instead of "sselog". Remove explicit "memory"
1751 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1752 use "sselog1" type instead of "sselog". Remove explicit "memory"
1754 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1755 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1757 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1758 "sselog1" type instead of "sselog". Remove explicit "memory"
1760 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1761 use "sselog1" type instead of "sselog". Remove explicit "memory"
1765 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1768 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1771 2018-04-17 Martin Jambor <mjambor@suse.cz>
1774 * ipa-cp.c (create_specialized_node): Call
1775 expand_all_artificial_thunks if necessary.
1777 2018-04-17 Martin Liska <mliska@suse.cz>
1780 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1781 in message, remote space in between '_G' and '('.
1783 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1786 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1787 avx512f_vmcmp<mode>3<round_saeonly_name>,
1788 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1789 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1790 avx512f_rndscale<mode><round_saeonly_name>,
1791 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1792 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1793 Use %<iptr>2 instead of %2 for -masm=intel.
1794 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1795 avx512f_vcvttss2usi<round_saeonly_name>,
1796 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1798 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1799 avx512f_vcvttsd2usi<round_saeonly_name>,
1800 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1801 Use %q1 instead of %1 for -masm=intel.
1802 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1803 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1804 of %3 for -masm=intel.
1805 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1807 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1809 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1811 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1812 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1814 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1816 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1817 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1818 %g1 and one with %0 and %1.
1819 (avx512er_vmrcp28<mode><round_saeonly_name>,
1820 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1822 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1823 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1824 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1825 of %0 and %{%4%} for -masm=intel.
1826 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1827 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1828 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1829 order of %0 and %{%5%}%{z%} for -masm=intel.
1831 2018-04-17 Jan Hubicka <jh@suse.cz>
1834 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1836 2018-04-17 Martin Liska <mliska@suse.cz>
1839 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1840 for target_clone error message. Make default implementation
1841 clone to be a local declaration.
1842 (separate_attrs): Add new argument and check for an empty
1844 (expand_target_clones): Handle it.
1845 (ipa_target_clone): Make redirection just for target_clones
1848 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1849 Tom de Vries <tom@codesourcery.com>
1852 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1853 tiled basic blocks without omp continue statements.
1855 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1858 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1859 vec_extract expression as having side effects to make sure it gets
1862 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1865 * config/i386/i386.c (get_builtin_code_for_version): Check
1868 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1871 * gcc/config.gcc: Support "skylake".
1872 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1874 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1875 (processor_target_table): Add "skylake".
1876 (ix86_option_override_internal): Add "skylake".
1877 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1878 PROCESSOR_CANNONLAKE.
1879 (get_builtin_code_for_version): Fix priority for
1880 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1881 PROCESSOR_SKYLAKE-AVX512.
1882 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1883 (processor_type): Add PROCESSOR_SKYLAKE.
1885 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1886 Jason Merrill <jason@redhat.com>
1889 * convert.c (convert_to_integer_1): Use direct recursion for
1890 enumeral types and types with a precision less than the number
1891 of bits in their mode.
1893 2018-04-16 Julia Koval <julia.koval@intel.com>
1896 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1897 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1899 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1902 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1903 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1904 and -mno-direct-move.
1906 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1909 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1910 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1911 Ensure negative shifts result in {0}.
1913 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1915 PR rtl-optimization/79916
1916 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1917 regs (if any) to define how to gnerate SD moves when LRA is in
1920 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1922 PR rtl-optimization/85393
1923 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1924 * except.c (expand_dw2_landing_pad_for_region): Make static.
1925 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1926 a label and unconditional jump to old_bb, rather than
1927 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1930 PR rtl-optimization/85376
1931 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1932 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1933 instead of a specific value.
1935 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1936 Bin Cheng <bin.cheng@arm.com>
1938 PR tree-optimization/82965
1939 PR tree-optimization/83991
1940 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1941 by_profile_only parameter.
1942 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1943 information if the loop was predicted to iterate too many times.
1944 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1946 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1949 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1952 2018-04-13 Martin Liska <mliska@suse.cz>
1953 Jakub Jelinek <jakub@redhat.com>
1956 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1957 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1958 * builtins.c (expand_builtin_memory_copy_args): Use
1959 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1960 handle dest_addr == pc_rtx.
1962 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1965 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1966 asked to not generate direct moves.
1967 (fix_trunc<mode>si2_stfiwx): Similar.
1968 (fix_trunc<mode>si2_internal): Similar.
1970 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1973 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1974 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1975 lookup if dest in some wider mode is known to be const0_rtx and
1976 if so, record permanent equivalence for it to be ZERO_EXTEND of
1977 the narrower mode destination.
1979 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1981 * lto-streamer-out.c (output_function): Revert 259346.
1982 * omp-expand.c (expand_oacc_for): Likewise.
1984 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1986 PR rtl-optimization/85354
1987 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1988 * sel-sched.c (sel_global_init): ... here.
1990 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1993 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1994 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1995 mode for PE-COFF targets.
1996 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1997 (i386_pe_asm_lto_end): Likewise.
1998 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1999 (TARGET_ASM_LTO_END): Likewise.
2000 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2001 (i386_pe_asm_lto_start): New function.
2002 (i386_pe_asm_lto_end): Likewise.
2004 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2005 Richard Biener <rguenther@suse.de>
2008 * lto-streamer-out.c (output_function): Fix CFG loop state before
2010 * omp-expand.c (expand_oacc_for): Handle calls to internal
2011 functions like regular functions.
2013 2018-04-12 Richard Biener <rguenther@suse.de>
2016 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2017 for the early LTO debug to properly generate references to it
2018 during DIE emission. Do not re-use that for the skeleton for
2020 (dwarf2out_early_finish): Likewise.
2022 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2025 * config/i386/sse.md
2026 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2027 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2028 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2029 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2030 and output is a reg, avoid creating invalid lowpart subreg, but
2031 instead split into a 512-bit move. Don't split if not AVX512VL,
2032 input is xmm16+ reg and output is a mem.
2033 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2034 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2035 xmm16+ reg and output is a mem.
2037 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2039 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2040 also for flag_dwarf2_cfi_asm.
2042 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2044 PR rtl-optimization/85342
2045 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2046 a bool scalar var inside of the loop instead. Don't try to update
2047 recog_data.operand after failed apply_change_group.
2049 2018-04-12 Tom de Vries <tom@codesourcery.com>
2052 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2053 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2054 array with flexible array member as array without given dimension.
2055 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2056 argument for undefined param to true.
2058 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2061 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2062 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2063 from PowerPC section.
2064 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2065 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2066 help text that is too long.
2067 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2068 help text that is too long.
2069 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2070 help text that is too long.
2072 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2074 * config/alpha/alpha.md (stack_probe_internal): Rename
2075 from "probe_stack". Update all callers.
2077 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2079 PR rtl-optimization/84566
2080 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2081 sched_macro_fuse_insns.
2083 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2086 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2087 (compute_block_dependences): ... from here.
2089 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2091 PR tree-optimization/85331
2092 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2093 from int to HOST_WIDE_INT.
2095 2018-04-11 Martin Jambor <mjambor@suse.cz>
2098 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2099 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2100 not the same as the source val.
2101 (cgraph_edge_brings_value_p): New parameter.
2102 (gather_edges_for_value): Pass destination value to
2103 cgraph_edge_brings_value_p.
2104 (perhaps_add_new_callers): Likewise.
2105 (get_info_about_necessary_edges): Likewise and exclude values brought
2106 only by self-recursive edges.
2107 (create_specialized_node): Redirect only clones of self-calling edges.
2108 (+self_recursive_pass_through_p): New function.
2109 (find_more_scalar_values_for_callers_subset): Use it.
2110 (find_aggregate_values_for_callers_subset): Likewise.
2111 (known_aggs_to_agg_replacement_list): Removed.
2112 (decide_whether_version_node): Re-calculate known constants for all
2113 remaining context clones.
2115 2018-04-11 Richard Biener <rguenther@suse.de>
2118 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2119 from early DWARF output.
2120 (dwarf2out_early_finish): Output line info unconditionally into
2121 early DWARF and add reference to it.
2123 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2126 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2127 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2128 other than V2DFmode using iptr mode attribute.
2129 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2131 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2133 PR rtl-optimization/84659
2134 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2136 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2139 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2141 (output_loc_list): Pass address of a dummy size variable even in the
2142 locview handling loop.
2143 (index_location_lists): Add comment on why skip_loc_list_entry can't
2146 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2149 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2152 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2155 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2156 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2157 and -mstring-compare-inline-limit.
2159 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2162 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2163 for stack clash protection in a register whenever we need it to be in
2166 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2168 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2169 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2171 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2174 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2176 (mlong-double-): Ditto.
2177 * config/rs6000/sysv4.opt (msdata=): Ditto.
2178 (mtls-size=): Ditto.
2180 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2182 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2183 erroneous entries for
2184 "vector int vec_ldl (int, long int *)", and
2185 "vector unsigned int vec_ldl (int, unsigned long int *)".
2186 Add comments and entries for
2187 "vector bool char vec_ldl (int, bool char *)",
2188 "vector bool short vec_ldl (int, bool short *)",
2189 "vector bool int vec_ldl (int, bool int *)",
2190 "vector bool long long vec_ldl (int, bool long long *)",
2191 "vector pixel vec_ldl (int, pixel *)",
2192 "vector long long vec_ldl (int, long long *)",
2193 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2194 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2195 type tree bool_long_long_type_node and correct definition of
2196 bool_V2DI_type_node to make reference to this new type tree.
2197 (rs6000_mangle_type): Replace erroneous reference to
2198 bool_long_type_node with bool_long_long_type_node.
2199 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2200 comments to emphasize sign distinctions for char and int types and
2201 replace RS6000_BTI_bool_long constant with
2202 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2203 use of RS6000_BTI_pixel.
2204 (bool_long_type_node): Remove this macro definition.
2205 (bool_long_long_type_node): New macro definition
2207 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2209 PR rtl-optimization/85300
2210 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2211 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2212 simplify_unary_operation fails.
2214 2018-04-10 Martin Liska <mliska@suse.cz>
2216 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2217 cgraph_edge and ipa_ref.
2219 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2223 * config/i386/sse.md
2224 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2225 computation of the VEC_MERGE selector from mask.
2226 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2227 Fix decoding of the VEC_MERGE selector into mask.
2229 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2231 PR tree-optimization/85286
2232 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2234 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2236 * final.c (final_1): Set insn_last_address as well as
2237 insn_current_address.
2239 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2242 * explow.c (emit_stack_probe): Call validize_mem on memory location
2243 before passing it to gen_probe_stack. Create address operand and
2244 legitimize it for the probe_stack_address case.
2246 2018-04-09 Jan Hubicka <jh@suse.cz>
2249 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2250 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2251 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2253 (free_lang_data): Rebuild type inheritance graph.
2255 2018-04-09 Martin Sebor <msebor@redhat.com>
2257 * invoke.texi (-finline-small-functions): Mention other optimization
2259 (-findirect-inlining, -fpartial-inlining): Same.
2260 (-finline-functions-called-once): Same.
2261 (-freorder-blocks-and-partition): Same.
2263 2018-04-09 Jan Hubicka <jh@suse.cz>
2266 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2267 jumps; choose last target that matches the criteria (i.e.
2268 no partition changes for non-crossing jumps).
2269 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2270 support for redirecting crossing jumps to non-crossing.
2272 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2274 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2275 also for naked functions.
2277 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2279 * config/arc/arc.md (add_shift): New pattern.
2280 (add_shift2): Likewise.
2281 (sub_shift): Likewise.
2282 (sub_shift_cmp0_noout): Likewise.
2283 (compare_si_ashiftsi): Likewise.
2284 (xbfu_cmp0_noout): New combine pattern.
2285 (xbfu_cmp0"): Likewise.
2286 (movsi_set_cc_insn): Place the predicable variant first.
2287 (commutative_binary_cmp0_noout): Remove clobber.
2288 (commutative_binary_cmp0): New pattern.
2289 (noncommutative_binary_cmp0): Likewise.
2290 (noncommutative_binary_cmp0_noout): Likewise.
2291 (noncommutative_binary_comparison_result_used): Removed.
2292 (rsub_cmp0): New pattern.
2293 (rsub_cmp0_noout): Likewise.
2294 (extzvsi): Changed, keep only meaningful variants.
2295 (SQH, SEZ): New iterators.
2296 (SQH_postfix): New mode attribute.
2297 (SEZ_prefix): New code attribute.
2298 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2299 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2300 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2302 (noncommutative_operator): Check the availability of barrel
2305 2018-04-09 Richard Biener <rguenther@suse.de>
2307 PR tree-optimization/85284
2308 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2309 Only use the niter constraining form of simple_iv when the exit
2312 2018-04-09 Tom de Vries <tom@codesourcery.com>
2315 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2316 (define_expand "*memory_barrier"): New define_expand.
2317 (define_insn "memory_barrier"): New insn.
2319 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2321 PR rtl-optimization/80463
2322 PR rtl-optimization/83972
2323 PR rtl-optimization/83480
2325 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2326 correct producer for the insn.
2327 (tidy_control_flow): Fixup seqnos in case of debug insns.
2329 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2331 PR rtl-optimization/83913
2333 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2334 different sched-times when merging exprs.
2336 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2338 PR rtl-optimization/83962
2340 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2341 tidy_fallthru_edge and tidy_control_flow.
2343 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2345 PR rtl-optimization/83530
2347 * sel-sched.c (force_next_insn): New global variable.
2348 (remove_insn_for_debug): When force_next_insn is true, also leave only
2349 next insn in the ready list.
2350 (sel_sched_region): When the region wasn't scheduled, make another pass
2351 over it with force_next_insn set to 1.
2353 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2355 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2357 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2358 for interrupt control.
2359 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2360 functions for interrupt control.
2361 * config/nds32/nds32-intrinsic.md: Likewise.
2362 * config/nds32/nds32_intrinsic.h: Likewise.
2363 * config/nds32/nds32.h (nds32_builtins): Likewise.
2365 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2367 * config/nds32/nds32.c (nds32_init_machine_status,
2368 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2369 strict_aligned_p field.
2370 (nds32_expand_to_rtl_hook): New function.
2371 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2372 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2374 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2375 Chung-Ju Wu <jasonwucj@gmail.com>
2377 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2378 * config/nds32/nds32-n7.md: New file.
2379 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2380 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2382 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2383 * config/nds32/nds32.md (pipeline_model): Add n7.
2384 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2385 * config/nds32/pipelines.md: Include n7 settings.
2387 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2388 Chung-Ju Wu <jasonwucj@gmail.com>
2390 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2391 * config/nds32/nds32-e8.md: New file.
2392 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2393 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2395 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2396 * config/nds32/nds32.md (pipeline_model): Add e8.
2397 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2398 * config/nds32/pipelines.md: Include e8 settings.
2400 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2401 Chung-Ju Wu <jasonwucj@gmail.com>
2403 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2404 * config/nds32/nds32-n8.md: New file.
2405 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2406 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2408 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2409 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2410 * config/nds32/nds32.md (pipeline_model): Add n8.
2411 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2412 * config/nds32/pipelines.md: Include n8 settings.
2414 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2415 Chung-Ju Wu <jasonwucj@gmail.com>
2417 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2418 * config/nds32/nds32-n9-2r1w.md: New file.
2419 * config/nds32/nds32-n9-3r2w.md: New file.
2420 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2421 nds32_register_ports): New or modify for cpu n9.
2422 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2424 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2425 * config/nds32/nds32-utils.c: New file.
2426 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2427 TARGET_MUL_SLOW): Define.
2428 * config/nds32/nds32.md (pipeline_model): New attribute.
2429 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2430 New options that support cpu n9.
2431 * config/nds32/pipelines.md: Include n9 settings.
2432 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2434 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2436 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2437 information if necessary.
2438 (output_cond_branch_compare_zero): Likewise.
2439 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2440 (nds32_target_alignment): Refine for alignment.
2441 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2442 (FUNCTION_BOUNDARY): Modify.
2443 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2445 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2447 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2449 * config/nds32/constants.md (unspec_volatile_element): Add values for
2450 TLB operation and data prefetch.
2451 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2452 functions for TLB operation and data prefetch.
2453 * config/nds32/nds32-intrinsic.md: Likewise.
2454 * config/nds32/nds32_intrinsic.h: Likewise.
2455 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2456 (nds32_print_operand): Likewise.
2457 * config/nds32/nds32.h (nds32_builtins): Likewise.
2459 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2460 Andrew Pinski <pinsika@gcc.gnu.org>
2463 * match.pd: Use constant_boolean_node of correct type instead of
2464 boolean_true_node or boolean_false_node for simplifying
2465 pointer comparisons to zero.
2467 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2469 PR tree-optimization/80021
2470 * tree.c (verify_type_variant): Make error call in verify_variant_match
2471 translatable and remove final full stop.
2473 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2475 * config/nds32/constants.md (unspec_volatile_element): Add
2476 UNSPEC_VOLATILE_EH_RETURN.
2477 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2478 nds32_output_stack_pop): Support dwarf exception handling process.
2479 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2480 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2481 exception handling process.
2482 (nds32_compute_stack_frame): Likewise.
2483 (nds32_return_addr_rtx): Likewise.
2484 (nds32_initial_elimination_offset): Likewise.
2485 (nds32_expand_prologue): Likewise.
2486 (nds32_expand_epilogue): Likewise.
2487 (nds32_dynamic_chain_address): New function.
2488 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2490 (DYNAMIC_CHAIN_ADDRESS): Define.
2491 (EH_RETURN_DATA_REGNO): Define.
2492 (EH_RETURN_STACKADJ_RTX): Define.
2493 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2494 patterns for dwarf exception handling.
2496 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2498 * config/nds32/nds32.h: Clean up obsolete macros.
2500 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2502 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2503 Add enum values for particular instructions.
2504 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2505 particular intrinsic functions.
2506 * config/nds32/nds32-intrinsic.md: Likewise.
2507 * config/nds32/nds32_intrinsic.h: Likewise.
2508 * config/nds32/nds32.h (nds32_builtins): Likewise.
2509 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2510 (btst, ave): New patterns for particular instructions.
2512 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2514 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2515 Add enum values for atomic load/store and memory sync.
2516 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2518 * config/nds32/nds32-intrinsic.md: Likewise.
2519 * config/nds32/nds32_intrinsic.h: Likewise.
2520 * config/nds32/nds32.h (nds32_builtins): Likewise.
2522 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2524 PR tree-optimization/85257
2525 * fold-const.c (native_encode_vector): If not all elts could fit
2526 and off is -1, return 0 rather than offset.
2527 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2528 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2529 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2530 adjust buffer in native_interpret_expr call.
2532 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2534 * config/nds32/constants.md (unspec_volatile_element): Add cache
2535 control enum values.
2536 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2537 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2538 * config/nds32/nds32.c (nds32_cctl_names): New.
2539 (nds32_print_operand): Handle cache control register names.
2540 * config/nds32/nds32.h (nds32_builtins): New enum values.
2541 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2543 * config/nds32/nds32.md (type): Add mmu.
2544 * config/nds32/pipelines.md (simple_insn): Add mmu.
2546 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2548 * config/nds32/nds32.md (type): Remove call.
2549 * config/nds32/pipelines.md (simple_insn): Likewise.
2551 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2553 * config/nds32/constants.md (unspec_volatile_element): Add
2554 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2555 UNSPEC_VOLATILE_FMFCFG.
2556 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2557 description for fmfcfg and fmfcsr.
2558 (bdesc_1arg): Add fmtcsr.
2559 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2560 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2561 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2562 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2563 unspec_fmfcfg): New patterns.
2564 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2565 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2566 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2567 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2568 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2569 __nds32__fmfcfg): Define.
2571 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2573 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2574 intrinsic register names.
2575 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2576 intrinsic register enum values and macros.
2578 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2580 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2581 for load/store addressing form.
2582 (nds32_print_operand_address): Likewise.
2584 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2587 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2588 based on LABEL_REF. Remove useless assertion.
2589 (pic_address_needs_scratch): Fix formatting.
2590 (sparc_legitimize_pic_address): Minor tweaks.
2591 (sparc_delegitimize_address): Adjust assertion accordingly.
2592 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2593 into symbolic_operand.
2594 (movsi_high_pic_label_ref): Likewise.
2595 (movsi_lo_sum_pic_label_ref): Likewise.
2596 (movdi_pic_label_ref): Likewise.
2597 (movdi_high_pic_label_ref): Likewise.
2598 (movdi_lo_sum_pic_label_ref): Likewise.
2600 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2602 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2603 custom LIB_SPEC setup.
2605 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2606 Kito Cheng <kito.cheng@gmail.com>
2608 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2609 * config/riscv/freebsd.h: New.
2611 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2613 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2614 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2617 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2618 Kito Cheng <kito.cheng@gmail.com>
2620 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2621 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2622 * config/nds32/nds32-protos.h (nds32_output_call,
2623 nds32_output_return): Declare.
2624 * config/nds32/nds32.md: Refine all the call and return patterns.
2626 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2629 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2630 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2632 PR rtl-optimization/84872
2633 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2634 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2637 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2639 * expr.c (copy_blkmode_to_reg): Revert 254862.
2640 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2642 2018-04-06 Richard Biener <rguenther@suse.de>
2645 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2646 after seeing a component reference with an adjacent field. Treat
2647 refs to arrays at struct end of external decls similar to
2648 refs to unconstrained commons.
2650 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2653 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2654 look through SAVE_EXPRs with non-side-effects argument. Adjust
2656 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2659 2018-04-06 Richard Biener <rguenther@suse.de>
2662 * alias.c (find_base_term): New wrapper around find_base_term
2663 unwinding CSELIB_VAL_PTR changes.
2664 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2667 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2669 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2671 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2672 constant definitions.
2673 ("nop"): lr 0,0 -> nopr r0
2674 ("nop_lr0", "nop_lr1"): New insn definitions.
2676 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2678 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2679 NDS32_V3PUSH_AVAILABLE_P macro.
2681 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2682 Chung-Ju Wu <jasonwucj@gmail.com>
2684 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2685 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2686 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2687 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2688 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2689 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2690 * config/nds32/constraints.md: New constraints and checking for hard
2691 float configuration.
2692 * config/nds32/iterators.md: New mode iterator and attribute for hard
2693 float configuration.
2694 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2696 * config/nds32/nds32-fpu.md: New file.
2697 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2698 deal with hard float code generation.
2699 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2701 (abi_type, float_reg_number): New enum type.
2702 * config/nds32/nds32-predicates.c: New predicates for hard float.
2703 * config/nds32/nds32-protos.h: Declare functions for hard float.
2704 * config/nds32/nds32.c: Implementation for hard float configuration.
2705 * config/nds32/nds32.h: Definitions for hard float configuration.
2706 * config/nds32/nds32.md: Include hard float machine description and
2707 modify patterns for hard float configuration.
2708 * config/nds32/nds32.opt: New options for hard float configuration.
2709 * config/nds32/predicates.md: New predicates for hard float
2712 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2714 * common/config/nds32/nds32-common.c
2715 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2717 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2720 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2721 CONSTRUCTOR_ELT (ctor, ...)->value.
2723 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2726 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2728 2018-04-05 Tom de Vries <tom@codesourcery.com>
2731 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2734 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2735 Kito Cheng <kito.cheng@gmail.com>
2737 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2738 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2739 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2740 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2742 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2743 Kito Cheng <kito.cheng@gmail.com>
2745 * config/nds32/constraints.md (Ufe): New memory constraint.
2746 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2747 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2748 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2750 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2751 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2753 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2755 * config/nds32/nds32.md: Use optimize_size in the condition for
2756 alu-shift instructions.
2758 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2760 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2762 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2764 * config/nds32/nds32.md (negsi2): Refine pattern.
2766 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2767 Chung-Ju Wu <jasonwucj@gmail.com>
2769 * config/nds32/iterators.md (shift_rotate): New code iterator.
2770 (shift): New code attribute.
2771 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2772 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2773 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2774 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2775 bit-wise operations.
2776 (andsi3, *andsi3): Ditto.
2777 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2778 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2779 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2780 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2781 nds32_ior_operand, nds32_xor_operand): New predicates.
2783 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2785 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2786 (addsi3, subsi3): ... this.
2788 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2790 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2792 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2794 * config/nds32/nds32.md: Adjust indention.
2796 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2798 * config/nds32/nds32.md (feature): New attribute.
2800 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2802 * config/nds32/nds32.md (subtype): New attribute.
2804 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2807 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2808 expansion to perform a bitwise AND of the argument followed by a
2809 boolean negation of the result.
2811 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2813 PR rtl-optimization/84878
2814 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2815 the basic block. Assert the use reference is not artificial and that
2816 it has an associated insn.
2818 2018-04-04 Michael Matz <matz@suse.de>
2820 * builtins.c (compute_objsize): Pass correct operand
2821 to array_at_struct_end_p.
2823 2018-04-04 Richard Biener <rguenther@suse.de>
2826 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2827 from contexts for DINFO_LEVEL_TERSE and below.
2829 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2831 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2832 resiter_operand condition.
2833 * config/nds32/nds32.md (*move<mode>): Ditto.
2835 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2836 Monk Chiang <sh.chiang04@gmail.com>
2838 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2840 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2842 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2844 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2845 Kito Cheng <kito.cheng@gmail.com>
2847 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2848 nds32_cond_code_str, output_cond_branch,
2849 output_cond_branch_compare_zero, nds32_expand_cbranch,
2850 nds32_expand_cstore, nds32_expand_movcc,
2851 nds32_output_cbranchsi4_equality_zero,
2852 nds32_output_cbranchsi4_equality_reg,
2853 nds32_output_cbranchsi4_equality_reg_or_const_int,
2854 nds32_output_cbranchsi4_greater_less_zero: New functions.
2855 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2856 nds32_expand_cstore, nds32_expand_movcc,
2857 nds32_output_cbranchsi4_equality_zero,
2858 nds32_output_cbranchsi4_equality_reg,
2859 nds32_output_cbranchsi4_equality_reg_or_const_int,
2860 nds32_output_cbranchsi4_greater_less_zero): Declare.
2861 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2862 nds32_rimm11s_operand): New predicates.
2863 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2864 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2867 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2869 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2870 * config/nds32/nds32.md: Ditto.
2871 * config/nds32/pipelines.md: Ditto.
2873 2018-04-04 Richard Biener <rguenther@suse.de>
2875 PR tree-optimization/85168
2876 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2877 propagating abnormals.
2879 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2881 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2883 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2884 Kito Cheng <kito.cheng@gmail.com>
2886 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2887 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2888 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2889 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2890 * config/nds32/nds32.md (sibcall_internal): New.
2891 (sibcall_register): Remove.
2892 (sibcall_immediate): Remove.
2893 (sibcall_value_internal): New.
2894 (sibcall_value_register): Remove.
2895 (sibcall_value_immediate): Remove.
2896 * config/nds32/predicates.md (nds32_general_register_operand): New.
2897 (nds32_call_address_operand): New.
2899 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2901 PR rtl-optimization/85167
2902 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2903 bb_defs if *split_p, instead preinitialize it to NULL.
2905 PR tree-optimization/85156
2906 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2907 evaluating the argument multiple times.
2909 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2911 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2913 (_mm_cvtpd_ps): Likewise.
2914 (_mm_cvttpd_epi32): Likewise.
2915 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2916 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2917 vector, pixel, and bool following altivec.h include.
2919 2018-04-03 Martin Sebor <msebor@redhat.com>
2921 * doc/extend.texi (Common Function Attributes): Clarify.
2922 (const attribute): Likewise.
2923 (pure attribute): Likewise.
2925 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2928 * config/i386/i386.c (ix86_expand_vector_set): Use
2929 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2931 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2933 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2934 instructions when changing rounding bits to preserve precision bits
2935 in the x87 control word.
2937 2018-04-03 Martin Liska <mliska@suse.cz>
2939 PR tree-optimization/82491
2940 * rtl.h (strip_offset_and_add): Replace += suboffset with
2941 poly_uint64 () + suboffset.
2943 2018-03-29 Martin Liska <mliska@suse.cz>
2944 Martin Jambor <mjambor@suse.cz>
2947 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2948 param_type is not an integral or pointer type.
2950 2018-04-03 Richard Biener <rguenther@suse.de>
2952 * sese.h (recompute_all_dominators): Remove.
2954 2018-04-02 Martin Sebor <msebor@redhat.com>
2956 * doc/invoke.texi (-Wrestrict): Fix typos.
2958 2018-04-02 Jim Wilson <jimw@sifive.com>
2960 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2961 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2962 (<optab>di3, <optab>si3_extend): Likewise.
2963 (<optab>si3_mask, <optab>si3_mask_1): New.
2964 (<optab>di3_mask, <optab>di3_mask_1): New.
2965 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2966 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2967 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2969 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2971 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2974 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2976 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2977 (nds32_canonicalize_comparison): New function.
2979 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2980 Kito Cheng <kito.cheng@gmail.com>
2981 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2983 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2984 * config/nds32/constants.md (unspec_volatile_element): Add
2985 UNSPEC_VOLATILE_RELAX_GROUP.
2986 * config/nds32/nds32-relax-opt.c: New file.
2987 * config/nds32/nds32-predicates.c
2988 (nds32_symbol_load_store_p): New function.
2989 * config/nds32/nds32-protos.h
2990 (nds32_symbol_load_store_p): Declare function.
2991 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2992 * config/nds32/nds32.c
2993 (nds32_register_pass): New function to register pass.
2994 (nds32_register_passes): New function to register passes.
2995 * config/nds32/nds32.md (relax_group): New pattern.
2996 * config/nds32/nds32.opt (mrelax-hint): New option.
2997 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2999 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3001 * config/nds32/t-nds32: Modify files dependency.
3003 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3005 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3006 (PROFILE_HOOK): Define its implementation.
3008 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3010 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3011 type and 32-bit size.
3013 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3016 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3017 (V_128_256): New mode iterator.
3018 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3019 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3020 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3022 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3023 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3025 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3028 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3029 NaN inputs correctly.
3031 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3034 * config/rs6000/vsx.md (??r): New mode attribute.
3035 (*vsx_mov<mode>_64bit): Use it.
3036 (*vsx_mov<mode>_32bit): Likewise.
3038 2018-03-30 Martin Sebor <msebor@redhat.com>
3040 PR tree-optimization/84818
3041 * builtins.c (check_access): Use warning_n.
3043 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3046 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3048 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3051 2018-03-30 Julia Koval <julia.koval@intel.com>
3054 * x86-tune.def (movx, partial_reg_dependency): Enable for
3057 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3060 * lra-constraints.c (process_alt_operands): Move setting
3061 this_alternative_matches below.
3063 2018-03-29 Martin Liska <mliska@suse.cz>
3066 * doc/invoke.texi: Document how LTO works with debug info.
3067 Describe auto-load support of binutils. Mention 'x86-64'
3068 as valid option value of -march option.
3070 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3072 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3075 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3076 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3077 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3080 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3083 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3084 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3085 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3086 for RS6000_BTM_POWERPC64.
3087 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3088 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3089 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3094 2018-03-28 Carl Love <cel@us.ibm.com>
3097 2017-09-27 Carl Love <cel@us.ibm.com>
3099 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3100 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3101 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3104 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3106 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3107 instead of __vector bool.
3108 (_mm_max_pu8): Likewise.
3109 (_mm_min_pi16): Likewise.
3111 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3114 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3115 (DIVWEUO): Likewise.
3117 (DIVDEUO): Likewise.
3118 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3119 DIVWEUO and DIVDEUO.
3120 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3121 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3122 (div_extend): Likewise.
3123 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3125 (__builtin_divweuo): Likewise.
3126 (__builtin_divdeo): Likewise.
3127 (__builtin_divdeuo): Likewise.
3129 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3132 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3133 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3135 PR tree-optimization/82004
3136 * gimple-match-head.c (optimize_pow_to_exp): New function.
3137 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3138 Don't fold to exp if optimize_pow_to_exp is false.
3140 2018-03-28 Martin Liska <mliska@suse.cz>
3143 * calls.c (initialize_argument_information): Fix trailing space.
3144 * common.opt: Fix typo and provide better explanation for
3145 -fsanitize-coverage option.
3146 * config/i386/i386.opt: Fix typo.
3148 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3149 Martin Liska <mliska@suse.cz>
3152 * gimplify.c (asan_poison_variable): Don't do the check for
3153 gimplify_omp_ctxp here.
3154 (gimplify_decl_expr): Do it here.
3155 (gimplify_target_expr): Likewise.
3157 2018-03-28 Martin Liska <mliska@suse.cz>
3160 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3161 chkp_type_bounds_count if MPX is not enabled.
3163 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3165 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3167 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3170 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3171 function to create the function decl for complex long double
3172 multiply and divide for -mabi=ieeelongdouble.
3173 (init_float128_ieee): Call it.
3175 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3178 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3179 -fcf-protection=branch -mibt.
3180 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3182 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3185 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3187 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3190 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3191 extern array declarations.
3193 2018-03-27 Richard Biener <rguenther@suse.de>
3196 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3197 explicit single_use checks.
3199 2018-03-27 Richard Biener <rguenther@suse.de>
3201 PR tree-optimization/85082
3202 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3205 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3207 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3208 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3209 Turn on fasynchronous-unwind-tables and funwind-tables.
3211 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3214 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3215 (*bmi_blsr_<mode>_ccz): Ditto.
3217 2018-03-26 Tom de Vries <tom@codesourcery.com>
3219 PR tree-optimization/85063
3220 * omp-general.c (offloading_function_p): New function. Factor out
3222 * omp-offload.c (pass_omp_target_link::gate): ... here.
3223 * omp-general.h (offloading_function_p): Declare.
3224 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3225 with attribute omp declare target for offloading functions.
3227 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3229 PR tree-optimization/84005
3230 * tree-data-ref.h (get_base_for_alignment): Declare.
3231 * tree-data-ref.c (get_base_for_alignment_1): New function.
3232 (get_base_for_alignment): Likewise.
3233 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3234 get_base_for_alignment to find a suitable base object, instead
3235 of always using drb->base_address.
3237 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3240 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3241 known size by default.
3243 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3246 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3247 and non BLKmode operands.
3249 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3252 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3253 Clean up attributes.
3255 2018-03-23 Richard Biener <rguenther@suse.de>
3258 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3259 we are going to emit early debug for LTO.
3261 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3264 * function.c (match_asm_constraints_1): Don't optimize if input
3265 doesn't satisfy general_operand predicate for output's mode.
3268 * alias.c (write_dependence_p): Don't require for x_canonicalized
3269 non-VOIDmode if x has VOIDmode.
3272 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3273 just don't try to optimize it rather than assert it never happens.
3275 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3277 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3278 macro expansions for definition of ST_INTERNAL_<mode> and
3279 LD_INTERNAL_<mode> builtins.
3280 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3282 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3284 (altivec_expand_st_builtin): Likewise.
3285 (altivec_expand_builtin): Remove calls to deleted functions.
3286 (rs6000_address_for_altivec): Delete this function.
3287 * config/rs6000/vector.md: Remove expands for
3288 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3290 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3293 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3294 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3295 re-computing once computed.
3296 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3297 (arm_init_machine_status): Initialize
3298 machine->static_chain_stack_bytes.
3300 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3303 * doc/extend.texi: Add four new prototypes for vec_ld.
3304 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3305 definitions for more logical presentation.
3306 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3307 entries for V1TI variants of __builtin_altivec_ld builtin.
3308 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3309 handling of V1TI variant of LVX icode pattern.
3310 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3311 (rs6000_gimple_fold_builtin): Likewise.
3312 (altivec_init_builtins): Add code to define
3313 __builtin_altivec_lvx_v1ti function.
3315 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3318 * function.c (match_asm_constraints_1): Don't do the optimization
3319 if input isn't a REG, SUBREG, MEM or constant.
3321 2018-03-22 Tom de Vries <tom@codesourcery.com>
3323 PR tree-optimization/84956
3324 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3325 bb_has_abnormal_pred.
3327 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3330 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3331 DECL_INITIAL (decl) to decl at the end.
3332 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3335 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3337 * doc/extend.texi (__builtin_tgmath): Document when complex
3338 integer types are treated as _Complex _Float64.
3340 2018-03-21 Tom de Vries <tom@codesourcery.com>
3342 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3344 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3346 PR tree-optimization/84960
3347 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3348 if it is ENTRY block, move them into single succ of ENTRY in that case.
3350 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3352 PR tree-optimization/84811
3353 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3354 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3355 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3356 (known_subrange_p): Update accordingly. Cast each value involved
3357 in the size comparison, rather than casting the result of the
3360 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3362 PR tree-optimization/84982
3363 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3364 by flipping the least significant bit rather than all bits from
3365 bitpos to bitpos + bitsize - 1.
3367 2018-03-21 Nathan Sidwell <nathan@acm.org>
3369 * doc/extend.texi (Deprecated Features): Remove mention of
3370 long-deleted deprecations.
3372 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3375 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3376 * configure: Regenerate.
3378 2018-03-21 Tom de Vries <tom@codesourcery.com>
3380 PR tree-optimization/83126
3381 * tree-parloops.c (num_phis): New function.
3382 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3384 2018-03-21 Nathan Sidwell <nathan@acm.org>
3386 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3387 mention anon-struct/union members and trailing attributes.
3389 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3391 PR tree-optimization/84969
3392 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3393 builtin memset partitions if they set different rhs values.
3395 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3397 PR rtl-optimization/84989
3398 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3399 VEC_DUPLICATE with scalar result mode.
3401 2018-03-21 Martin Liska <mliska@suse.cz>
3404 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3405 not intended return statement.
3407 2018-03-21 Martin Liska <mliska@suse.cz>
3410 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3411 (chkp_find_bound_slots_1): Limit number of iterations.
3413 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3416 * Minor grammar fixes for x86 options.
3418 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3421 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3422 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3424 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3427 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3428 (altivec_lvx_<mode>_1op): Likewise.
3429 (altivec_stvx_<mode>_2op): Likewise.
3430 (altivec_stvx_<mode>_1op): Likewise.
3431 (altivec_lvx_<VM2:mode>): New define_expand.
3432 (altivec_stvx_<VM2:mode>): Likewise.
3433 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3434 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3435 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3436 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3437 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3438 (rs6000_gen_lvx): Likewise.
3439 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3440 (altivec_expand_stv_builtin): Likewise.
3441 (altivec_expand_builtin): Likewise.
3442 * config/rs6000/vector.md: Likewise.
3444 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3447 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3450 2018-03-20 Richard Biener <rguenther@suse.de>
3453 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3454 sign-conversions as zero, fall back to standard scalar_stmt
3457 2018-03-20 Martin Liska <mliska@suse.cz>
3460 * predict.c (rebuild_frequencies): Handle case when we have
3461 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3463 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3466 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3467 flag_section_anchors.
3468 * varasm.c (use_blocks_for_decl_p): Remove hack for
3469 dw2_force_const_mem.
3472 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3474 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3475 be created, use lowpart_subreg of operands[0] rather than operands[0]
3477 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3478 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3479 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3480 and n constraint instead of aarch64_shift_imm_di and Usd.
3481 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3482 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3484 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3487 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3488 to favor GPR over NEON registers.
3489 (<shift>di3_neon): Likewise.
3491 2018-03-20 Tom de Vries <tom@codesourcery.com>
3494 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3495 (nvptx_process_pars): Emit bar.sync asap and alap.
3497 2018-03-20 Tom de Vries <tom@codesourcery.com>
3500 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3501 seen_label if seen_label is already set.
3503 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3506 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3507 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3508 Use 1U instead of 1. Formatting fixes.
3511 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3512 instead of TREE_TYPE (s1) for the return value.
3514 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3516 PR tree-optimization/84946
3517 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3518 bitsize + bitsize in poly_uint64 rather than poly_int64.
3521 * dwarf2asm.c: Include fold-const.c.
3522 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3523 of decl rather than decl itself.
3525 PR rtl-optimization/84643
3526 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3528 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3531 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3532 calling assemble_variable.
3534 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3537 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3538 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3540 2018-03-19 Jim Wilson <jimw@sifive.com>
3543 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3544 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3545 (riscv_first_stack_step): Likewise.
3546 (riscv_option_override): Use STACK_BOUNDARY instead of
3548 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3550 (BIGGEST_ALIGNMENT): Set to 128.
3551 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3552 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3555 2018-03-19 Richard Biener <rguenther@suse.de>
3557 PR tree-optimization/84933
3558 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3559 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3561 2018-03-19 Richard Biener <rguenther@suse.de>
3563 PR tree-optimization/84859
3564 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3565 (cond_if_else_store_replacement): Perform sinking operation on
3566 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3567 Generalize what a BB with a single eligible store is.
3569 2018-03-19 Richard Biener <rguenther@suse.de>
3571 PR tree-optimization/84929
3572 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3573 chrec_is_positive against non-chrec arg.
3575 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3578 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3580 2018-03-18 Martin Liska <mliska@suse.cz>
3582 PR rtl-optimization/84635
3583 * regrename.c (build_def_use): Use matches_mode only when
3586 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3588 PR tree-optimization/84913
3589 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3590 vectorize chains of COND_EXPRs.
3592 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3594 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3596 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3598 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3600 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3602 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3604 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3605 Kito Cheng <kito.cheng@gmail.com>
3607 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3608 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3609 (nds32_adjust_reg_alloc_order): New function.
3610 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3612 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3614 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3615 nds32_print_operand, nds32_print_operand_address): Use
3616 HOST_WIDE_INT_PRINT_DEC instead.
3618 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3620 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3622 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3625 * config/i386/i386.c (initial_ix86_tune_features,
3626 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3628 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3629 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3630 rather than 1u << ix86_tune. Formatting fix.
3631 (ix86_option_override_internal): Change ix86_arch_mask from
3632 unsigned int to unsigned HOST_WIDE_INT, initialize to
3633 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3634 (ix86_function_specific_restore): Likewise.
3636 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3639 * postreload.c (reload_combine_recognize_pattern): Perform
3640 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3641 truncate_int_for_mode the result for the destination's mode.
3644 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3645 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3647 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3650 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3653 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3654 regno_allocno_class_array and sorted_pseudos.
3655 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3656 insns where regno is used.
3658 2018-03-16 Martin Liska <mliska@suse.cz>
3661 * multiple_target.c (create_dispatcher_calls): Redirect
3662 reference in the symbol table.
3664 2018-03-16 Martin Liska <mliska@suse.cz>
3667 * multiple_target.c (create_dispatcher_calls): Redirect also
3670 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3674 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3675 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3676 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3678 2018-03-16 Julia Koval <julia.koval@intel.com>
3680 * doc/invoke.texi (Skylake Server): Add CLWB.
3681 Cannonlake): Remove CLWB.
3683 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3685 PR tree-optimization/84841
3686 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3688 (FLOAT_ONE_CONST_TYPE): Define.
3689 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3690 (sort_by_operand_rank): Put entries with higher constant_type last
3691 rather than first to match comments.
3693 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3695 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3698 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3701 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3702 check and use address of target always.
3704 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3707 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3708 (indirect_thunk_bnd_needed): Likewise.
3709 (indirect_thunks_used): Likewise.
3710 (indirect_thunks_bnd_used): Likewise.
3711 (indirect_return_needed): New.
3712 (indirect_return_bnd_needed): Likewise.
3713 (output_indirect_thunk_function): Add a bool argument for
3715 (output_indirect_thunk_function): Don't generate alias for
3716 function return thunk.
3717 (ix86_code_end): Call output_indirect_thunk_function to generate
3718 function return thunks.
3719 (ix86_output_function_return): Set indirect_return_bnd_needed
3720 and indirect_return_needed instead of indirect_thunk_bnd_needed
3721 and indirect_thunk_needed.
3723 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3725 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3726 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3727 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3729 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3730 Paul Hua <paul.hua.gm@gmail.com>
3733 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3735 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3737 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3738 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3741 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3744 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3745 instead of GET_MODE_SIZE when comparing Units.
3747 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3750 * varasm.c (hash_section): Return an unchangeble hash value
3751 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3752 Return !aarch64_can_use_per_function_literal_pools_p ().
3754 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3757 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3758 rather than address of cmode as last argument to prepare_cmp_insn.
3760 2018-03-15 Julia Koval <julia.koval@intel.com>
3762 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3763 F_AVX512VNNI, F_AVX512BITALG): New.
3765 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3768 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3769 insn for floating-point loads and stores.
3771 2018-03-14 Carl Love <cel@us.ibm.com>
3773 * config/rs6000/rs6000-c.c: Add macro definitions for
3774 ALTIVEC_BUILTIN_VEC_PERMXOR.
3775 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3776 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3777 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3778 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3780 * config/doc/extend.texi: Add prototypes for vec_permxor.
3782 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3785 * diagnostic-show-locus.c (class layout_point): Convert m_line
3786 from int to linenum_type.
3787 (line_span::comparator): Use linenum "compare" function when
3788 comparing line numbers.
3789 (test_line_span): New function.
3790 (layout_range::contains_point): Convert param "row" from int to
3792 (layout_range::intersects_line_p): Likewise.
3793 (layout::will_show_line_p): Likewise.
3794 (layout::print_source_line): Likewise.
3795 (layout::should_print_annotation_line_p): Likewise.
3796 (layout::print_annotation_line): Likewise.
3797 (layout::print_leading_fixits): Likewise.
3798 (layout::annotation_line_showed_range_p): Likewise.
3799 (struct line_corrections): Likewise for field m_row.
3800 (line_corrections::line_corrections): Likewise for param "row".
3801 (layout::print_trailing_fixits): Likewise.
3802 (layout::get_state_at_point): Likewise.
3803 (layout::get_x_bound_for_row): Likewise.
3804 (layout::print_line): Likewise.
3805 (diagnostic_show_locus): Likewise for locals "last_line" and
3807 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3808 * input.c (selftest::test_linenum_comparisons): New function.
3809 (selftest::input_c_tests): Call it.
3810 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3811 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3812 * selftest.h (ASSERT_GT): New macro.
3813 (ASSERT_GT_AT): New macro.
3814 (ASSERT_LT): New macro.
3815 (ASSERT_LT_AT): New macro.
3817 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3819 PR rtl-optimization/84780
3820 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3822 2018-03-14 Martin Liska <mliska@suse.cz>
3824 * tree.c (record_node_allocation_statistics): Use
3825 get_stats_node_kind.
3826 (get_stats_node_kind): New function extracted from
3827 record_node_allocation_statistics.
3828 (free_node): Use get_stats_node_kind.
3830 2018-03-14 Richard Biener <rguenther@suse.de>
3832 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3833 that the value-set of ANTIC_IN doesn't grow.
3836 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3838 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3839 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3841 2018-03-14 Julia Koval <julia.koval@intel.com>
3843 * config.gcc (icelake-client, icelake-server): New.
3845 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3846 (initial_ix86_arch_features): Ditto.
3847 (PTA_SKYLAKE): Add SGX.
3848 (PTA_ICELAKE): Remove.
3849 (PTA_ICELAKE_CLIENT): New.
3850 (PTA_ICELAKE_SERVER): New.
3851 (ix86_option_override_internal): Split up icelake on icelake client and
3853 (get_builtin_code_for_version): Ditto.
3854 (fold_builtin_cpu): Ditto.
3855 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3856 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3857 * config/i386/i386.h (processor_type): Ditto.
3858 * doc/invoke.texi: Ditto.
3860 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3863 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3864 INTEGER_CST offset, add it together with bitpos / 8 and
3865 sign extend based on POINTER_SIZE.
3869 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3872 * config/i386/constraints.md (Yc): New register constraint.
3873 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3874 Use Yc constraint for alternative 2 of operand 0. Remove
3875 preferred_for_speed attribute.
3877 2018-03-14 Richard Biener <rguenther@suse.de>
3879 PR tree-optimization/84830
3880 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3881 with the old one to avoid oscillations.
3883 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3886 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3888 (assign_by_spills): Return a flag of reload assignment failure.
3889 Do not process the reload assignment failures. Do not spill other
3890 reload pseudos if they has the same reg class. Update n if
3892 (lra_assign): Add a return arg. Set up from the result of
3893 assign_by_spills call.
3894 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3895 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3896 usage_insns if it is not NULL.
3897 (spill_hard_reg_in_range): New function.
3898 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3899 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3900 function prototypes.
3901 (lra_assign): Change prototype.
3902 * lra.c (lra): Add code to deal with fails by splitting hard reg
3905 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3907 * config/riscv/riscv.opt (mrelax): New option.
3908 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3909 "norelax" when riscv_mrelax is disabled.
3910 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3912 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3915 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3916 reassociation for int modes.
3918 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3920 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3921 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3923 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3924 * config/aarch64/aarch64-sve.md
3925 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3926 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3927 (*extend<mode><Vwide>2): Rename to...
3928 (aarch64_sve_extend<mode><Vwide>2): ...this.
3929 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3930 renaming the old pattern to...
3931 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3933 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3934 define_expand, renaming the old pattern to...
3935 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3936 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3937 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3938 account when deciding which SVE instruction the optab should use.
3939 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3941 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3943 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3944 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3945 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3946 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3947 (tlsdesc_small_<mode>): Turn a define_expand and use
3948 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3949 (tlsdesc_small_advsimd_<mode>): ...this.
3950 (tlsdesc_small_sve_<mode>): New pattern.
3952 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3954 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3955 (UNSPEC_UMUL_HIGHPART): New constants.
3956 (MUL_HIGHPART): New int iteraor.
3957 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3958 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3960 (*<su>mul<mode>3_highpart): New define_insn.
3962 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3965 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3968 2018-03-13 Martin Liska <mliska@suse.cz>
3971 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3973 (sem_item_optimizer::~sem_item_optimizer): Release it.
3974 (sem_item_optimizer::merge_classes): Register variable aliases.
3975 (sem_item_optimizer::fixup_pt_set): New function.
3976 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3977 * ipa-icf.h: Declare new variables and functions.
3979 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3982 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3983 integer_pow2p@2 and test integer_pow2p in condition.
3984 (A < 0 ? C : 0): Similarly for @1.
3987 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3988 characters starting at p contain '\0' character, don't look beyond
3992 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3993 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3996 * reg-stack.c (change_stack): Change update_end var from int to
3997 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3998 also call set_block_for_insn on the newly added insns and rescan.
4001 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4002 on the last operand.
4005 * tree.c (stabilize_reference_1): Return save_expr (e) for
4006 STATEMENT_LIST even if it doesn't have side-effects.
4008 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4010 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4012 2018-03-12 Renlin Li <renlin.li@arm.com>
4014 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4015 aarch64_output_scalar_simd_mov_immediate.
4017 2018-03-12 Martin Sebor <msebor@redhat.com>
4019 PR tree-optimization/83456
4020 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4021 for perfectly overlapping calls to memcpy.
4022 (gimple_fold_builtin_memory_chk): Same.
4023 (gimple_fold_builtin_strcpy): Handle no-warning.
4024 (gimple_fold_builtin_stxcpy_chk): Same.
4025 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4027 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4029 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4030 parameter. Use it for SFmode.
4031 (rs6000_function_arg_advance_1): Adjust.
4032 (rs6000_function_arg): Adjust.
4033 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4035 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4037 PR rtl-optimization/84169
4038 PR rtl-optimization/84780
4039 * combine.c (can_combine_p): Check for a 2-insn combination whether
4040 the destination register is used between the two insns, too.
4042 2018-03-12 Richard Biener <rguenther@suse.de>
4044 PR tree-optimization/84803
4045 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4046 for refs DR analysis didn't process.
4048 2018-03-12 Richard Biener <rguenther@suse.de>
4050 PR tree-optimization/84777
4051 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4052 force-vectorize loops ignore whether we are optimizing for size.
4054 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4056 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4057 (TARGET_MD_ASM_ADJUST): Define.
4059 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4060 Kito Cheng <kito.cheng@gmail.com>
4061 Chung-Ju Wu <jasonwucj@gmail.com>
4063 * config/nds32/nds32.c (nds32_compute_stack_frame,
4064 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4065 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4066 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4067 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4068 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4069 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4070 * config/nds32/nds32.md (prologue, epilogue): Use macro
4071 NDS32_V3PUSH_AVAILABLE_P to do checking.
4073 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4076 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4077 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4078 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4079 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4080 addition of most attributes on !orig_type_die or the attribute not
4081 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4083 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4084 Chung-Ju Wu <jasonwucj@gmail.com>
4086 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4088 * config/nds32/nds32.opt (mvh): New option.
4090 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4091 Chung-Ju Wu <jasonwucj@gmail.com>
4093 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4095 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4096 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4099 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4100 Chung-Ju Wu <jasonwucj@gmail.com>
4102 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4104 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4105 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4107 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4108 Kito Cheng <kito.cheng@gmail.com>
4109 Chung-Ju Wu <jasonwucj@gmail.com>
4111 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4112 UNSPEC_FFMISM and UNSPEC_FLMISM.
4113 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4114 for ffb, ffmism and flmism.
4115 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4116 (unspec_ffmism): Ditto.
4117 (unspec_flmism): Ditto.
4118 (nds32_expand_builtin_impl): Check if string extension is available.
4119 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4120 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4122 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4125 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4128 * lra-assigns.c (assign_by_spills): Return a flag of reload
4129 assignment failure. Do not process the reload assignment
4130 failures. Do not spill other reload pseudos if they has the same
4132 (lra_assign): Add a return arg. Set up from the result of
4133 assign_by_spills call.
4134 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4135 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4136 usage_insns if it is not NULL.
4137 (spill_hard_reg_in_range): New function.
4138 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4139 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4140 function prototypes.
4141 (lra_assign): Change prototype.
4142 * lra.c (lra): Add code to deal with fails by splitting hard reg
4145 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4148 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4150 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4153 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4154 (dw_val_node): Add val_symbolic_view.
4155 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4156 (symview_upper_bound): New.
4157 (new_line_info_table): Initialize symviews_since_reset.
4158 (dwarf2out_source_line): Count symviews_since_reset and set
4159 symview_upper_bound.
4160 (dw_val_equal_p): Handle symview.
4161 (add_AT_symview): New.
4162 (print_dw_val): Handle symview.
4163 (attr_checksum, attr_checksum_ordered): Likewise.
4164 (same_dw_val_p, size_of_die): Likewise.
4165 (value_format, output_die): Likewise.
4166 (add_high_low_attributes): Use add_AT_symview for entry_view.
4167 (dwarf2out_finish): Reset symview_upper_bound, clear
4170 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4173 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4174 Add strict argument and use it.
4175 (rs6000_split_multireg_move): Update for new strict argument.
4176 (mem_operand_gpr): Disallow all non-offsettable addresses.
4177 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4179 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4182 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4183 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4184 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4187 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4188 decl, use remap_type if we want to use the type.
4190 2018-03-09 Martin Sebor <msebor@redhat.com>
4192 PR tree-optimization/84526
4193 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4195 (builtin_access::generic_overlap): Be prepared to handle non-array
4198 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4200 PR rtl-optimization/84682
4201 * lra-constraints.c (process_address_1): Check is_address flag
4202 for address constraints.
4203 (process_alt_operands): Likewise.
4204 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4205 preprocess_constraints.
4206 * recog.h (preprocess_constraints): Add oploc parameter.
4208 * recog.c (preprocess_constraints): Test address_operand for
4209 CT_ADDRESS constraints.
4211 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4214 * lra-assigns.c (assign_by_spills): Return a flag of reload
4215 assignment failure. Do not process the reload assignment
4216 failures. Do not spill other reload pseudos if they has the same
4218 (lra_assign): Add a return arg. Set up from the result of
4219 assign_by_spills call.
4220 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4221 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4222 usage_insns if it is not NULL.
4223 (spill_hard_reg_in_range): New function.
4224 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4225 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4226 function prototypes.
4227 (lra_assign): Change prototype.
4228 * lra.c (lra): Add code to deal with fails by splitting hard reg
4231 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4234 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4235 Accept complain bool parameter. Only emit errors if it is true.
4236 (arm_parse_cpu_option_name): Likewise.
4237 (arm_target_thumb_only): Adjust callers of the above.
4238 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4239 prototype to take a default true bool parameter.
4240 (arm_parse_arch_option_name): Likewise.
4242 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4243 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4247 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4248 * configure: Regenerate.
4249 * configure.ac ("linker --version-script option"): New.
4250 ("linker soname option"): New.
4252 2018-03-09 Richard Biener <rguenther@suse.de>
4254 PR tree-optimization/84775
4255 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4256 immediate uses of predicate stmts and mark them modified.
4259 PR tree-optimization/84178
4260 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4262 (version_loop_for_if_conversion): Delay update_ssa call.
4263 (tree_if_conversion): Delay update_ssa until after predicate
4266 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4269 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4270 when the function accesses prior frames.
4272 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4275 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4276 gen_llsym, otherwise call maybe_gen_llsym.
4279 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4280 has ',' character inside of it.
4282 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4285 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4286 as clobbering CC_REGNUM.
4288 2018-03-08 Richard Biener <rguenther@suse.de>
4291 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4292 (follow_copies_to_constant): Do not follow SSA names registered
4295 2018-03-08 Richard Biener <rguenther@suse.de>
4297 PR tree-optimization/84178
4298 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4300 (version_loop_for_if_conversion): Delay update_ssa call.
4301 (tree_if_conversion): Delay update_ssa until after predicate
4304 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4306 PR tree-optimization/84178
4307 * tree-if-conv.c (release_bb_predicate): Remove the
4308 the assertion that the stmts have NULL use_ops.
4309 Discard the statements, asserting that they haven't
4310 yet been added to a BB.
4312 2018-03-08 Richard Biener <rguenther@suse.de>
4314 PR tree-optimization/84746
4315 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4316 (phi_translate): Pass in destination ANTIC_OUT set.
4317 (phi_translate_1): Likewise. For a simplified result lookup
4318 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4319 (phi_translate_set): Adjust.
4320 (do_pre_regular_insertion): Likewise.
4321 (do_pre_partial_partial_insertion): Likewise.
4323 2018-03-08 Martin Liska <mliska@suse.cz>
4325 PR gcov-profile/84735
4326 * doc/gcov.texi: Document usage of profile files.
4327 * gcov-io.h: Document changes in the format.
4329 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4333 * dwarf2out.c (struct dw_line_info_table): Update comments for
4335 (FORCE_RESET_NEXT_VIEW): New.
4336 (FORCE_RESETTING_VIEW_P): New.
4337 (RESETTING_VIEW_P): Check for -1 too.
4338 (ZERO_VIEW_P): Likewise.
4339 (new_line_info_table): Force-reset next view.
4340 (dwarf2out_begin_function): Likewise.
4341 (dwarf2out_source_line): Simplify zero_view_p initialization.
4342 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4343 view directly. Omit view when omitting .loc at line 0.
4345 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4347 PR tree-optimization/84740
4348 * tree-switch-conversion.c (process_switch): Call build_constructors
4349 only if info.phi_count is non-zero.
4351 PR tree-optimization/84739
4352 * tree-tailcall.c (find_tail_calls): Check call arguments against
4353 DECL_ARGUMENTS (current_function_decl) rather than
4354 DECL_ARGUMENTS (func) when checking for tail recursion.
4356 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4358 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4359 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4360 Volker Reichelt's entry and add entries for people that perform
4361 GCC fuzzy testing and report numerous bugs.
4363 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4366 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4367 readonly data in sdata, if that is disabled.
4368 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4369 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4370 -mreadonly-in-sdata option.
4372 2018-03-07 Martin Sebor <msebor@redhat.com>
4374 PR tree-optimization/84468
4375 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4376 basic block when looking for nul assignment.
4378 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4381 * except.h (output_function_exception_table): Adjust prototype.
4382 * except.c (output_function_exception_table): Remove FNNAME parameter
4383 and add SECTION parameter. Ouput one part of the table at a time.
4384 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4385 the first part of the exception table and emit unwind directives.
4386 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4387 (i386_pe_seh_cold_init): Likewise.
4388 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4389 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4390 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4391 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4392 * config/i386/winnt.c: Include except.h.
4393 (struct seh_frame_state): Add reg_offset, after_prologue and
4394 in_cold_section fields.
4395 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4396 (i386_pe_seh_cold_init): New function.
4397 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4398 to seh->in_cold_section.
4399 (seh_emit_push): Record the offset of the push.
4400 (seh_emit_save): Record the offet of the save.
4401 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4402 Test seh->after_prologue to disregard the epilogue.
4403 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4404 (i386_pe_end_cold_function): New function.
4406 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4409 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4410 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4413 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4415 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4416 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4419 * multiple_target.c: Include tree-inline.h and intl.h.
4420 (expand_target_clones): Diagnose and fail if node->definition and
4421 !tree_versionable_function_p (node->decl).
4423 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4425 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4427 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4428 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4429 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4431 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4434 * combine.c (try_combine): Use reg_or_subregno instead of handling
4435 just paradoxical SUBREGs and REGs.
4437 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4439 * config/arc/arc.c (arc_finalize_pic): Remove function.
4440 (arc_must_save_register): We use single base PIC register, remove
4441 checks to save/restore the PIC register.
4442 (arc_expand_prologue): Likewise.
4443 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4445 (arc_verify_short): Likewise.
4446 (arc_attr_type): Likewise.
4447 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4448 (walk_stores): Likewise.
4449 (arc_address_cost): Make it static.
4450 (arc_verify_short): Likewise.
4451 (branch_dest): Likewise.
4452 (arc_attr_type): Likewise.
4453 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4454 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4455 (arc_final_prescan_insn): Remove inserting the nops due to
4456 hardware hazards. It is done in reorg step.
4457 (insn_length_variant_t): Remove.
4458 (insn_length_parameters_t): Likewise.
4459 (arc_insn_length_parameters): Likewise.
4460 (arc_get_insn_variants): Likewise.
4461 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4463 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4466 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4469 PR tree-optimization/84687
4470 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4472 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4474 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4476 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4477 Rename to ppc_speculation_barrier.
4478 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4479 __builtin_ppc_speculation_barrier.
4481 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4484 * combine.c (combine_simplify_rtx): Don't try to simplify if
4485 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4488 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4490 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4491 to 32 bytes when compiling for POWER9.
4493 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4496 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4497 regparm >= 3 with no arg reg available also for calls with
4498 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4501 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4503 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4505 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4508 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4510 2018-03-05 Richard Biener <rguenther@suse.de>
4512 PR tree-optimization/84486
4513 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4514 When inserting a __builtin_assume_aligned call set the LHS
4515 SSA name alignment info accordingly.
4517 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4519 PR tree-optimization/84114
4520 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4521 Avoid reassociation of FLOAT_MODE addition.
4523 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4525 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4526 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4527 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4528 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4529 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4530 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4531 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4533 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4534 __builtin_ia32_wbinvd): New builtins.
4535 (SPECIAL_ARGS2): New.
4536 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4537 (SPECIAL_ARGS2): New.
4538 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4539 (ix86_valid_target_attribute_inner_p): Ditto.
4540 (ix86_init_mmx_sse_builtins): Add special_args2.
4541 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4542 TARGET_WBNOINVD_P): New.
4543 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4544 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4545 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4546 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4547 * config/i386/pconfigintrin.h: New file.
4548 * config/i386/wbnoinvdintrin.h: Ditto.
4549 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4550 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4552 2018-03-05 Richard Biener <rguenther@suse.de>
4554 PR tree-optimization/84670
4555 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4557 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4558 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4559 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4560 doesn't grow if all successors (recursively) were visited at least
4563 2018-03-05 Richard Biener <rguenther@suse.de>
4565 PR tree-optimization/84650
4566 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4567 if executed in the loop pipeline.
4569 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4571 * doc/configfiles.texi (Configuration Files): Move info about
4572 conditionalizing $target-protos.h to...
4573 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4574 differs from $target-protos.h.
4576 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4577 Chung-Ju Wu <jasonwucj@gmail.com>
4579 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4580 * config/nds32/nds32-multiple.md (setmemsi): Define.
4581 * config/nds32/nds32-memory-manipulation.c
4582 (nds32_gen_dup_4_byte_to_word_value): New.
4583 (emit_setmem_word_loop): New.
4584 (emit_setmem_byte_loop): New.
4585 (nds32_expand_setmem_loop): New.
4586 (nds32_expand_setmem_loop_v3m): New.
4587 (nds32_expand_setmem_unroll): New.
4588 (nds32_expand_setmem): New.
4590 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4591 Chung-Ju Wu <jasonwucj@gmail.com>
4593 * config/nds32/nds32-memory-manipulation.c
4594 (nds32_emit_load_store): New.
4595 (nds32_emit_post_inc_load_store): New.
4596 (nds32_emit_mem_move): New.
4597 (nds32_emit_mem_move_block): New.
4598 (nds32_expand_movmemsi_loop_unknown_size): New.
4599 (nds32_expand_movmemsi_loop_known_size): New.
4600 (nds32_expand_movmemsi_loop): New.
4601 (nds32_expand_movmemsi_unroll): New.
4602 (nds32_expand_movmemqi): Rename ...
4603 (nds32_expand_movmemsi): ... to this.
4604 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4605 (movmemsi): ... to this.
4606 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4607 (nds32_expand_movmemsi): ... to this.
4609 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4610 Monk Chiang <sh.chiang04@gmail.com>
4611 Chung-Ju Wu <jasonwucj@gmail.com>
4613 * config/nds32/nds32-protos.h
4614 (nds32_expand_load_multiple): New arguments.
4615 (nds32_expand_store_multiple): Ditto.
4616 (nds32_valid_multiple_load_store): Rename ...
4617 (nds32_valid_multiple_load_store_p): ... to this.
4618 * config/nds32/nds32-memory-manipulation.c
4619 (nds32_expand_load_multiple): Refine implementation.
4620 (nds32_expand_store_multiple): Ditto.
4621 * config/nds32/nds32-multiple.md
4622 (load_multiple): Update nds32_expand_load_multiple interface.
4623 (store_multiple): Update nds32_expand_store_multiple interface.
4624 * config/nds32/nds32-predicates.c
4625 (nds32_valid_multiple_load_store): Rename ...
4626 (nds32_valid_multiple_load_store_p): ... to this and refine
4628 * config/nds32/predicates.md
4629 (nds32_load_multiple_and_update_address_operation): New predicate.
4630 (nds32_store_multiple_and_update_address_operation): New predicate.
4632 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4633 Chung-Ju Wu <jasonwucj@gmail.com>
4635 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4636 (combo): New attribute.
4637 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4639 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4641 * config/nds32/nds32.opt: Change -mcmodel= default value.
4643 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4644 Monk Chiang <sh.chiang04@gmail.com>
4645 Chung-Ju Wu <jasonwucj@gmail.com>
4647 * config/nds32/constants.md (unspec_element): New enum.
4648 * config/nds32/constraints.md (Umw): New constraint.
4649 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4650 * config/nds32/nds32-intrinsic.md: Likewise.
4651 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4652 (nds32_valid_smw_lwm_base_p): New.
4653 (nds32_output_smw_single_word): New.
4654 (nds32_output_lmw_single_word): New.
4655 (nds32_expand_unaligned_load): New.
4656 (nds32_expand_unaligned_store): New.
4657 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4658 (nds32_output_smw_single_word): Declare.
4659 (nds32_output_lmw_single_word): Declare.
4660 (nds32_expand_unaligned_load): Declare.
4661 (nds32_expand_unaligned_store): Declare.
4662 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4663 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4664 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4665 NDS32_BUILTIN_UASTORE_DW.
4666 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4669 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4670 Kito Cheng <kito.cheng@gmail.com>
4671 Chung-Ju Wu <jasonwucj@gmail.com>
4673 * config/nds32/nds32-intrinsic.c
4674 (nds32_expand_builtin_null_ftype_reg): Delete.
4675 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4676 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4677 (nds32_read_argument): New.
4678 (nds32_legitimize_target): Ditto.
4679 (nds32_legitimize_argument): Ditto.
4680 (nds32_check_constant_argument): Ditto.
4681 (nds32_expand_unop_builtin): Ditto.
4682 (nds32_expand_unopimm_builtin): Ditto.
4683 (nds32_expand_binop_builtin): Ditto.
4684 (nds32_builtin_decl_impl): Ditto.
4685 (builtin_description): Ditto.
4686 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4687 (nds32_init_builtins_impl): Ditto.
4688 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4689 (nds32_builtin_decl): New.
4690 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4691 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4693 2018-03-02 Jeff Law <law@redhat.com>
4695 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4696 (redundant_insn, fill_simple_delay_slots): Likewise.
4697 (fill_slots_from_thread): Likewise.
4698 * resource.c (mark_referenced_resources): Likewise.
4699 (mark_set_resources, find_dead_or_set_registers): Likewise.
4701 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4703 * substring-locations.h (format_warning_va): Formatting fix for
4705 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4707 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4708 * substring-locations.c: Include intl.h.
4709 (format_warning_va): Turned into small wrapper around
4710 format_warning_n_va, renamed to ...
4711 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4712 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4714 (format_warning_at_substring_n): New function.
4715 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4716 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4717 format_warning_at_substring with just a shorter name instead of
4718 const function pointer.
4719 (fmtwarn_n): New function.
4720 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4721 appropriate, get rid of all the fmtstr temporaries, move conditionals
4722 with G_() wrapped string literals directly into fmtwarn arguments,
4723 cast dir.len to (int), formatting fixes.
4725 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4727 * doc/invoke.texi: Remove "Cilk Plus" references.
4729 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4730 Richard Biener <rguenther@suse.de>
4733 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4734 for error or warning attributes if CALL_FROM_THUNK_P is set.
4737 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4740 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4741 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4744 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4745 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4747 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4748 instead of -1U in last predictors element's probability member.
4750 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4753 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4754 arguments if they are comparable.
4756 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4758 PR tree-optimization/84634
4759 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4760 masks and masked_loop_p with a single loop_masks, making sure it's
4761 null for bb vectorization.
4763 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4765 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4766 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4767 loop->force_vectorize to check whether there is no alias.
4769 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4772 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4774 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4776 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4777 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4778 instead of a loop around prev_real_insn.
4779 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4783 * config/i386/i386.c (ix86_print_operand): Use conditional
4784 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4787 2018-03-02 Richard Biener <rguenther@suse.de>
4789 PR tree-optimization/84427
4790 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4791 (bitmap_set_subtract_values): Rewrite to handle multiple
4794 (prune_clobbered_mems): Likewise.
4795 (phi_translate): Take edge instead of pred/phiblock.
4796 (phi_translate_1): Likewise.
4797 (phi_translate_set): Likewise. Insert all translated
4798 exprs for a value into the set, keeping possibly multiple
4799 expressions per value.
4800 (compute_antic_aux): Adjust for phi_translate changes.
4801 When intersecting union the expressions and prune those
4802 not in the final value set, keeping possibly multiple
4803 expressions per value. Do not use value-insertion
4804 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4805 all expressions. Add verification that the value-sets
4806 only shrink during iteration.
4807 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4808 (do_pre_regular_insertion): Likewise.
4809 (do_pre_partial_partial_insertion): Likewise.
4811 2018-03-02 Richard Biener <rguenther@suse.de>
4814 * config/darwin.c (saved_debug_info_level): New static global.
4815 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4816 (darwin_asm_lto_end): Restore debug info generation settings.
4818 2018-03-01 Martin Liska <mliska@suse.cz>
4821 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4824 2018-03-01 Richard Biener <rguenther@suse.de>
4827 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4828 type annotation with LTO when debug was disabled at compile-time.
4830 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4832 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4834 (mips_final_postscan_insn): Likewise.
4836 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4838 PR rtl-optimization/84528
4839 * alias.c (init_alias_target): Add commentary.
4840 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4841 a unique base value if the frame pointer is not eliminated
4842 to the stack pointer.
4844 2018-03-01 Tom de Vries <tom@codesourcery.com>
4846 PR rtl-optimization/83327
4847 * lra-int.h (hard_regs_spilled_into): Declare.
4848 * lra.c (hard_regs_spilled_into): Define.
4849 (init_reg_info): Init hard_regs_spilled_into.
4850 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4851 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4852 (process_bb_lives): Handle hard_regs_spilled_into.
4853 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4854 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4856 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4858 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4859 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4860 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4861 * config/rs6000/aix72.h: New file.
4863 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4865 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4866 instead of warning_at with conditional singular and plural messages
4870 * stor-layout.c (update_alignment_for_field): For
4871 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4872 && !DECL_PACKED (field), do the alignment update, just use
4873 only desired_align instead of MAX (type_align, desired_align)
4875 (place_field): Don't do known_align < desired_align handling
4876 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4877 is non-NULL, instead do it after rli->prev_field handling and
4878 only if not within a bitfield word. For DECL_PACKED (field)
4879 use type_align of BITS_PER_UNIT.
4881 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4883 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4884 superfluous parentheses and trailing spaces.
4886 2018-02-28 Richard Biener <rguenther@suse.de>
4888 PR tree-optimization/84584
4889 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4890 SCoPs with fake exit edge.
4892 2018-02-28 Martin Liska <mliska@suse.cz>
4895 * timevar.c (timer::print): Fix format to properly print 100%
4898 2018-02-28 Richard Biener <rguenther@suse.de>
4901 * genmatch.c (capture_info::walk_match): Do not mark
4902 captured expressions without operands as expr_p given
4903 they act more like predicates and should be subject to
4904 "lost tail" side-effect preserving.
4906 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4908 PR rtl-optimization/81611
4909 * auto-inc-dec.c (attempt_change): Move dead note from
4910 mem_insn if it's the next use of regno
4911 (find_address): Take address use of reg holding
4912 non-incremented value. Add parm to limit search to the named
4914 (merge_in_block): Attempt to use a mem insn that is the next
4915 use of the original regno.
4917 2018-02-27 Martin Sebor <msebor@redhat.com>
4920 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4921 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4923 2018-02-27 Martin Sebor <msebor@redhat.com>
4925 PR translation/84207
4926 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4927 n argument to unsigned HOST_WIDE_INT.
4928 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4929 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4930 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4931 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4933 2018-02-27 Richard Biener <rguenther@suse.de>
4935 PR tree-optimization/84512
4936 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4937 Do not use the estimate returned from record_stmt_cost for
4938 the scalar iteration cost but sum properly using add_stmt_cost.
4940 2018-02-27 Richard Biener <rguenther@suse.de>
4942 PR tree-optimization/84466
4943 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4944 Adjust last change to less strictly validate use operands.
4946 2018-02-27 Martin Liska <mliska@suse.cz>
4948 PR gcov-profile/84548
4949 * gcov.c (process_file): Allow partial overlap and consider it
4950 also as group functions.
4951 (output_lines): Properly calculate range of lines for a group.
4953 2018-02-27 Martin Liska <mliska@suse.cz>
4955 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4956 'ggc' suffixes. Change first column width.
4957 (timer::print): Fix formatting of the column.
4959 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4961 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4962 preserve inline entry blocks for the sake of debug inline
4963 entry point markers alone.
4964 (remove_unused_locals): Suggest in comments a better place to
4965 force the preservation of inline entry blocks that are
4966 otherwise unused, but do not preserve them.
4968 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4970 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4972 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4975 * config/i386/constraints.md (Bs): Replace
4976 ix86_indirect_branch_register with
4977 TARGET_INDIRECT_BRANCH_REGISTER.
4979 * config/i386/i386.md (indirect_jump): Likewise.
4980 (tablejump): Likewise.
4981 (*sibcall_memory): Likewise.
4982 (*sibcall_value_memory): Likewise.
4983 Peepholes of indirect call and jump via memory: Likewise.
4984 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4985 (*sibcall_value_GOT_32): Likewise.
4986 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4987 (GOT_memory_operand): Likewise.
4988 (call_insn_operand): Likewise.
4989 (sibcall_insn_operand): Likewise.
4990 (GOT32_symbol_operand): Likewise.
4991 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4993 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4995 PR rtl-optimization/83496
4996 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4997 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4998 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4999 redundant insn, if any.
5000 (relax_delay_slots): Likewise.
5001 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5003 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5005 PR tree-optimization/83965
5006 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5007 that grouped statements are part of a reduction chain. Return
5008 true if the statement is not marked as a reduction itself but
5010 (vect_recog_dot_prod_pattern): Don't check whether the statement
5011 is part of a group here.
5012 (vect_recog_sad_pattern): Likewise.
5013 (vect_recog_widen_sum_pattern): Likewise.
5015 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5018 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5020 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5023 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5025 (ix86_output_indirect_function_return): New prototype.
5026 (ix86_split_simple_return_pop_internal): Likewise.
5027 * config/i386/i386.c (indirect_return_via_cx): New.
5028 (indirect_return_via_cx_bnd): Likewise.
5029 (indirect_thunk_name): Handle return va CX_REG.
5030 (output_indirect_thunk_function): Create alias for
5031 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5032 (ix86_output_indirect_jmp): Remove the bool argument.
5033 (ix86_output_indirect_function_return): New function.
5034 (ix86_split_simple_return_pop_internal): Likewise.
5035 * config/i386/i386.md (*indirect_jump): Don't pass false
5036 to ix86_output_indirect_jmp.
5037 (*tablejump_1): Likewise.
5038 (simple_return_pop_internal): Change it to define_insn_and_split.
5039 Call ix86_split_simple_return_pop_internal to split it for
5041 (simple_return_indirect_internal): Call
5042 ix86_output_indirect_function_return instead of
5043 ix86_output_indirect_jmp.
5045 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5048 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5049 memset and value initialization afterwards.
5051 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5053 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5055 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5058 * common/config/aarch64/aarch64-common.c
5059 (aarch_option_optimization_table[]): Switch
5060 off fomit-frame-pointer
5062 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5063 Chung-Ju Wu <jasonwucj@gmail.com>
5065 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5067 (store_multiple): Ditto.
5069 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5071 * config.gcc: Add --with-cpu support for nds32 target.
5072 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5073 * config/nds32/nds32.opt: Add -mcpu= option.
5075 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5077 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5078 isel=yes): Warn for these deprecated options.
5080 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5082 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5083 ISA_2_5_MASKS_EMBEDDED.
5085 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5087 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5088 p->max as pointers rather than using iterative_hash_expr.
5090 2018-02-23 Carl Love <cel@us.ibm.com>
5092 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5093 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5095 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5096 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5097 P8V_BUILTIN_VEC_VUNSIGNED2.
5099 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5102 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5103 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5105 * lra-constraints.c (curr_insn_transform): Set up
5106 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5108 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5109 * lra-eliminations.c (spill_pseudos): Ditto.
5110 (process_insn_for_elimination): Ditto.
5111 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5112 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5113 LRA_NON_CLOBBERED_ALT.
5115 2018-02-22 Martin Sebor <msebor@redhat.com>
5117 PR tree-optimization/84480
5118 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5119 to maybe_diag_stxncpy_trunc. Call it.
5120 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5121 from gimple_fold_builtin_strcpy. Print inlining stack.
5122 (handle_builtin_stxncpy): Print inlining stack.
5123 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5125 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5128 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5129 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5130 and -fcheck-pointer-bounds are used together.
5131 (indirect_thunk_prefix): New enum.
5132 (indirect_thunk_need_prefix): New function.
5133 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5134 "_nt" instead of "_bnd" for NOTRACK prefix.
5135 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5136 (output_indirect_thunk_function): Likewise.
5138 (ix86_code_end): Update output_indirect_thunk_function calls.
5139 (ix86_output_indirect_branch_via_reg): Replace
5140 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5141 (ix86_output_indirect_branch_via_push): Likewise.
5142 (ix86_output_function_return): Likewise.
5143 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5144 incompatible with -fcf-protection=branch and
5145 -fcheck-pointer-bounds.
5147 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5150 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5151 Change gcc_assert call to output_operand_lossage.
5153 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5155 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5157 2018-02-22 DJ Delorie <dj@redhat.com>
5158 Sebastian Perta <sebastian.perta@renesas.com>
5159 Oleg Endo <olegendo@gcc.gnu.org>
5161 * config/rx/rx.c (rx_rtx_costs): New function.
5162 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5164 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5166 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5168 2018-02-22 Martin Liska <mliska@suse.cz>
5171 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5172 Add "native" as a possible value.
5174 2018-02-22 Martin Liska <mliska@suse.cz>
5177 * config/i386/i386.c (ix86_option_override_internal):
5178 Add "native" as a possible value for -march and -mtune.
5180 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5183 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5184 to all type variants.
5186 PR tree-optimization/84503
5187 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5188 width as info->bitpos + info->bitsize - start.
5189 (merged_store_group::merge_overlapping): Simplify width computation.
5190 (check_no_overlap): New function.
5191 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5192 start + width and last_order of the group, fail if check_no_overlap
5194 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5195 to group if check_no_overlap fails.
5197 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5199 * config/rs6000/altivec.md: Delete contraint arguments to
5200 define_expand, define_split, and define_peephole2, and in
5201 define_insn_and_split if always unused.
5202 * config/rs6000/darwin.md: Ditto.
5203 * config/rs6000/dfp.md: Ditto.
5204 * config/rs6000/rs6000.md: Ditto.
5205 * config/rs6000/sync.md: Ditto.
5206 * config/rs6000/vector.md: Ditto.
5207 * config/rs6000/vsx.md: Ditto.
5209 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5211 * config/rs6000/altivec.md: Write output control strings as braced
5212 blocks instead of double-quoted strings.
5213 * config/rs6000/darwin.md: Ditto.
5214 * config/rs6000/rs6000.md: Ditto.
5215 * config/rs6000/vector.md: Ditto.
5216 * config/rs6000/vsx.md: Ditto.
5218 2018-02-21 Jason Merrill <jason@redhat.com>
5220 PR c++/84314 - ICE with templates and fastcall attribute.
5221 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5223 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5225 * ipa-cp.c (determine_versionability): Fix comment typos.
5227 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5230 * ipa-cp.c (determine_versionability): Do not version functions caling
5233 2018-02-21 Martin Liska <mliska@suse.cz>
5236 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5237 Add "native" as a possible value.
5238 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5239 the macro when native cpu detection is available.
5241 2018-02-21 Martin Liska <mliska@suse.cz>
5244 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5245 Add "native" as a possible value.
5246 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5247 when native cpu detection is available.
5249 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5250 Martin Sebor <msebor@redhat.com>
5252 PR tree-optimization/84478
5253 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5255 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5256 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5257 support which is conservatively correct, for 2 only stay conservative
5258 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5259 argument to the 2 argument get_range_strlen, adjust 6 arg
5260 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5262 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5263 (gimple_fold_builtin_strlen): Pass true as last argument to
5266 2018-02-20 Martin Sebor <msebor@redhat.com>
5269 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5270 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5271 (builtin_memref::builtin_memref): Factor out parts into
5272 set_base_and_offset and call it.
5274 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5277 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5278 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5279 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5280 search at the associated MODE_INT.
5282 2018-02-20 Jeff Law <law@redhat.com>
5285 PR tree-optimization/81592
5287 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5288 for range data rather than using global data.
5289 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5290 range data rather than using global data.
5291 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5292 pass it to children as needed.
5293 (struct directive::fmtresult): Similarly.
5294 (struct directive::set_width): Similarly.
5295 (struct directive::set_precision): Similarly.
5296 (format_integer, format_directive, parse_directive): Similarly.
5297 (format_none): Accept unnamed vr_values parameter.
5298 (format_percent, format_floating, format_character): Similarly.
5299 (format_string, format_plain): Similarly.
5300 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5301 the EVRP range analyzer for range data rather than using global data.
5302 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5303 gimple-ssa-evrp-analyze.h
5304 (class sprintf_dom_walker): Add after_dom_children member function.
5305 Add evrp_range_analyzer member.
5306 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5307 range analyzer as needed.
5308 (sprintf_dom_walker::after_dom_children): New member function.
5309 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5311 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5312 (evrp_range_analyzer::pop_to_marker): Likewise.
5314 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5316 PR tree-optimization/84419
5317 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5318 with the required type if its current type is compatible but
5321 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5324 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5325 after vectorization.
5327 2018-02-20 Martin Liska <mliska@suse.cz>
5330 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5331 possible values if we don't have a hint.
5333 2018-02-20 Martin Liska <mliska@suse.cz>
5337 * final.c (shorten_branches): Build align_tab array with one
5339 * opts.c (finish_options): Add alignment option limit check.
5340 (MAX_CODE_ALIGN): Likewise.
5341 (MAX_CODE_ALIGN_VALUE): Likewise.
5342 * doc/invoke.texi: Document maximum allowed option value for
5343 all -falign-* options.
5345 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5348 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5349 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5350 * var-tracking.c (emit_note_insn_var_location): Remove all references
5351 to NOTE_INSN_CALL_ARG_LOCATION.
5352 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5353 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5354 Use copy_rtx_if_shared.
5355 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5356 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5357 (dwarf2out_var_location): Remove handling of
5358 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5360 * final.c (final_scan_insn): Remove all references to
5361 NOTE_INSN_CALL_ARG_LOCATION.
5362 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5363 before dumping final insns.
5364 * except.c (emit_note_eh_region_end): Remove all references to
5365 NOTE_INSN_CALL_ARG_LOCATION.
5366 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5367 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5368 * config/arc/arc.c (hwloop_optimize): Likewise.
5369 * config/arm/arm.c (create_fix_barrier): Likewise.
5370 * config/s390/s390.c (s390_chunkify_start): Likewise.
5371 * config/sh/sh.c (find_barrier): Likewise.
5372 * config/i386/i386.c (rest_of_insert_endbranch,
5373 ix86_seh_fixup_eh_fallthru): Likewise.
5374 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5375 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5376 * config/frv/frv.c (frv_function_prologue): Likewise.
5377 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5379 (note_outside_basic_block_p): Remove all references to
5380 NOTE_INSN_CALL_ARG_LOCATION.
5381 * gengtype.c (adjust_field_rtx_def): Likewise.
5382 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5384 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5385 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5388 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5391 PR tree-optimization/84452
5392 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5393 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5396 2018-02-19 Martin Liska <mliska@suse.cz>
5399 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5401 2018-02-19 Martin Liska <mliska@suse.cz>
5402 Richard Sandiford <richard.sandiford@linaro.org>
5404 PR tree-optimization/82491
5405 * gimple-fold.c (get_base_constructor): Make earlier bail out
5408 2018-02-19 Carl Love <cel@us.ibm.com>
5410 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5411 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5413 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5414 P8V_BUILTIN_VEC_NEG.
5416 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5418 * config/rl78/rl78.md (movdf): New define expand.
5420 2018-02-19 Martin Liska <mliska@suse.cz>
5423 * doc/invoke.texi: Fix typo.
5424 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5426 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5428 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5429 handle rs6000_single_float and rs6000_double_float specially for
5432 2018-02-16 Jeff Law <law@redhat.com>
5434 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5437 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5440 * ipa-inline.c (inline_small_functions): Fix a typo.
5442 2018-02-16 Nathan Sidwell <nathan@acm.org>
5444 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5446 2018-02-16 Carl Love <cel@us.ibm.com>
5448 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5449 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5450 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5451 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5452 expansion to P8V_BUILTIN_VEC_FLOAT2.
5454 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5456 PR rtl-optimization/70023
5457 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5458 src_regno into account.
5460 2018-02-16 Carl Love <cel@us.ibm.com>
5462 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5463 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5464 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5465 * config/rs6000/rs6000.c: Remove case statements for
5466 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5467 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5468 and P9V_BUILTIN_VEC_VINSERT4B.
5469 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5470 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5471 * config/rs6000/vsx.md:
5472 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5475 2018-02-16 Carl Love <cel@us.ibm.com>
5477 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5479 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5481 * config/rs6000/rs6000-c.c: Add the definitions for
5482 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5483 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5484 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5485 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5486 definition for insert4b and define insn *insert3b_internal.
5487 * doc/extend.texi: Add documentation for vec_extract4b.
5489 2018-02-16 Nathan Sidwell <nathan@acm.org>
5491 * doc/extend.texi (Backwards Compatibility): Mention friend
5492 injection. Note for-scope is deprecated.
5493 * doc/invoke.texi (-ffriend-injection): Deprecate.
5495 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5497 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5498 that moved to I2, also allow destinations that are a paradoxical
5499 subreg (instead of a normal reg).
5501 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5504 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5507 2018-02-16 Richard Biener <rguenther@suse.de>
5509 PR tree-optimization/84037
5510 PR tree-optimization/84016
5512 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5513 Adjust vec_construct for the fact we need additional higher latency
5514 128bit inserts for AVX256 and AVX512 vector builds.
5515 (ix86_add_stmt_cost): Scale vector construction cost for
5518 2018-02-16 Richard Biener <rguenther@suse.de>
5520 PR tree-optimization/84417
5521 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5522 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5523 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5525 2018-02-16 Martin Liska <mliska@suse.cz>
5528 * internal-fn.def (ASAN_CHECK): Set proper flags.
5529 (ASAN_MARK): Likewise.
5531 2018-02-16 Julia Koval <julia.koval@intel.com>
5533 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5534 from PTA_CANNONLAKE.
5536 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5539 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5540 Use ++iter rather than iter++ for std::list iterators.
5541 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5542 defer deleting them until all nodes in the forest are processed. Do
5543 free even leaf nodes. Change to_process into auto_vec.
5546 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5547 * vec.h (vec_default_construct): Use memset instead of placement new
5548 if BROKEN_VALUE_INITIALIZATION is defined.
5549 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5550 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5553 PR rtl-optimization/83723
5554 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5555 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5556 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5558 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5560 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5562 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5564 PR rtl-optimization/81443
5565 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5566 from inner REGs to paradoxical SUBREGs.
5568 2018-02-16 Richard Biener <rguenther@suse.de>
5570 PR tree-optimization/84399
5571 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5572 For operands we can analyze at their definition make sure we can
5573 analyze them at each use as well.
5575 2018-02-16 Richard Biener <rguenther@suse.de>
5577 PR tree-optimization/84190
5578 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5579 volatile accesses if the decl isn't volatile.
5581 2018-02-15 Jason Merrill <jason@redhat.com>
5583 PR c++/84314 - ICE with templates and fastcall attribute.
5584 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5585 TYPE_CANONICAL on an existing type.
5587 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5589 PR tree-optimization/84383
5590 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5591 dstoff nor call operand_equal_p if dstbase is NULL.
5593 PR tree-optimization/84334
5594 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5595 also a CONSTANT_CLASS_P, punt.
5597 2018-02-14 Jim Wilson <jimw@sifive.com>
5599 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5600 first SMALL_OPERAND check. New local min_second_step. Move assert
5601 to where locals are set. Add TARGET_RVC support.
5602 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5604 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5606 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5608 2018-02-14 Martin Sebor <msebor@redhat.com>
5610 PR tree-optimization/83698
5611 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5612 arrays constrain the offset range to their bounds.
5613 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5614 (builtin_access::overlap): Avoid setting the size of overlap if it's
5616 (maybe_diag_overlap): Also consider arrays when deciding what values
5617 of offsets to include in diagnostics.
5619 2018-02-14 Martin Sebor <msebor@redhat.com>
5622 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5623 that correspond to the kind of a declaration.
5625 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5628 * config/pa/pa.md: Load address of PIC label using the linkage table
5629 if the label is nonlocal.
5631 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5633 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5634 warning message if user requests -maltivec=be.
5635 * doc/invoke.texi: Document deprecation of -maltivec=be.
5637 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5640 * config/rs6000/rs6000-c.c: Update definitions for
5641 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5642 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5644 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5647 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5648 add _get_ssp intrinsics. Remove argument from
5649 __builtin_ia32_rdssp[d|q].
5650 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5651 * config/i386/i386-builtin.def: Remove argument from
5652 __builtin_ia32_rdssp[d|q].
5653 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5654 ix86_expand_special_args_builtin for _rdssp[d|q].
5655 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5656 Clear register before usage.
5657 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5658 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5660 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5662 PR tree-optimization/84357
5663 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5664 operand 1 of an ARRAY_REF too.
5666 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5669 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5670 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5672 (set_of_reg): New struct.
5673 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5674 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5675 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5677 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5678 Split into bitclr, bitset, bitinvert patterns if appropriate.
5679 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5680 use rx_fuse_in_memory_bitop.
5681 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5682 to named insn, correct maximum insn length.
5684 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5687 * machmode.def: Define a complex mode for PARTIAL_INT.
5688 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5690 * doc/rtl.texi: Document CSPImode.
5691 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5693 (msp430_hard_regno_nregs_with_padding): Likewise.
5695 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5698 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5700 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5702 PR rtl-optimization/84169
5703 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5704 we generated a parallel as new i3 and we split that to new i2 and i3
5705 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5706 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5707 those to i2, not i1. Partially rewrite this scan code.
5709 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5712 * stor-layout.c (place_field): For variable length fields, adjust
5713 offset_align afterwards not just based on the field's alignment,
5714 but also on the size.
5717 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5718 of exps and logs in the use_exp2 case.
5720 2018-02-13 Jeff Law <law@redhat.com>
5722 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5725 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5728 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5732 * common.opt (gas-loc-support, gas-locview-support): New.
5733 (ginline-points, ginternal-reset-location-views): New.
5734 * doc/invoke.texi: Document them. Use @itemx where intended.
5735 (gvariable-location-views): Adjust.
5736 * target.def (reset_location_view): New.
5737 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5738 (TARGET_RESET_LOCATION_VIEW): New.
5739 * doc/tm.texi: Rebuilt.
5740 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5741 (dwarf2out_default_as_locview_support): New.
5742 (output_asm_line_debug_info): Use option variables.
5743 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5744 (output_loc_list): Likewise.
5745 (add_high_low_attributes): Check option variables.
5746 Don't output entry view attribute in strict mode.
5747 (gen_inlined_subroutine_die): Check option variables.
5748 (dwarf2out_inline_entry): Likewise.
5749 (init_sections_and_labels): Likewise.
5750 (dwarf2out_early_finish): Likewise.
5751 (maybe_reset_location_view): New, from...
5752 (dwarf2out_var_location): ... here. Call it.
5753 * debug.h (dwarf2out_default_as_loc_support): Declare.
5754 (dwarf2out_default_as_locview_support): Declare.
5755 * hooks.c (hook_int_rtx_insn_0): New.
5756 * hooks.h (hook_int_rtx_insn_0): Declare.
5757 * toplev.c (process_options): Take -gas-loc-support and
5758 -gas-locview-support from dwarf2out. Enable
5759 -gvariable-location-views by default only with locview
5760 assembler support. Enable -ginternal-reset-location-views by
5761 default only if the target defines the corresponding hook.
5762 Enable -ginline-points by default if location views are
5763 enabled; force it disabled if statement frontiers are
5765 * tree-inline.c (expand_call_inline): Check option variables.
5766 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5768 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5770 PR tree-optimization/84321
5771 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5772 handling. Also check whether the anti-range contains any values
5773 that satisfy the mask; switch to a VR_RANGE if not.
5775 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5778 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5780 2018-02-13 Martin Jambor <mjambor@suse.cz>
5783 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5784 of call statements, also set location of a load to a temporary.
5786 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5788 * config/rl78/rl78.c (add_vector_labels): New function.
5789 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5790 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5791 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5792 which checks that no arguments are passed.
5793 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5794 * doc/extend.texi: Documentation for the new attribute.
5796 2018-02-13 Andreas Schwab <schwab@suse.de>
5798 * config/riscv/linux.h (CPP_SPEC): Define.
5800 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5803 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5804 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5805 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5806 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5807 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5808 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5809 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5810 temporarily for AES and PCLMUL builtins.
5812 PR tree-optimization/84339
5813 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5814 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5818 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5819 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5820 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5822 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5824 * omp-simd-clone.h: New file.
5825 * omp-simd-clone.c: Include omp-simd-clone.h.
5826 (expand_simd_clones): No longer static.
5827 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5828 cgraph.h and omp-simd-clone.h.
5829 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5830 (vect_recog_widen_shift_pattern): Formatting fix.
5831 (vect_pattern_recog_1): Don't check optab for calls.
5834 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5835 operands[2] into a REG before using gen_lowpart on it.
5837 2018-02-12 Jeff Law <law@redhat.com>
5840 * config/sh/sh.c (find_barrier): Consider a sibling call
5843 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5844 successfully back substituting a reg.
5846 2018-02-12 Richard Biener <rguenther@suse.de>
5848 PR tree-optimization/84037
5849 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5850 parameter, move visited init to caller.
5851 (vect_slp_analyze_operations): Separate cost from validity
5852 check, initialize visited once for all instances.
5853 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5855 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5857 (vect_model_promotion_demotion_cost): Likewise.
5858 (vectorizable_bswap): Guard cost modeling with !slp_node
5859 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5861 (vectorizable_call): Likewise.
5862 (vectorizable_conversion): Likewise.
5863 (vectorizable_assignment): Likewise.
5864 (vectorizable_shift): Likewise.
5865 (vectorizable_operation): Likewise.
5866 (vectorizable_store): Likewise.
5867 (vectorizable_load): Likewise.
5868 (vectorizable_condition): Likewise.
5869 (vectorizable_comparison): Likewise.
5871 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5874 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5875 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5876 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5878 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5881 * params.def (inline-min-speedup): Increase from 8 to 15.
5882 (max-inline-insns-auto): Decrease from 40 to 30.
5883 * ipa-split.c (consider_split): Add some buffer for function to
5884 be considered inlining candidate.
5885 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5888 2018-02-12 Richard Biener <rguenther@suse.de>
5890 PR tree-optimization/84037
5891 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5892 matched stmts if we cannot swap the non-matched ones.
5894 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5896 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5897 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5898 _mm_maskz_scalef_round_ss): New intrinsics.
5899 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5900 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5901 __builtin_ia32_scalefss_round): Remove.
5902 (__builtin_ia32_scalefsd_mask_round,
5903 __builtin_ia32_scalefss_mask_round): New intrinsics.
5904 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5905 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5906 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5907 "<round_constraint>")): Changed to ...
5908 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5909 "<round_scalar_constraint>")): ... this.
5910 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5911 %0, %1, %2<round_op3>}"): Changed to ...
5912 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5913 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5914 %2<round_scalar_mask_op3>}"): ... this.
5915 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5917 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5919 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5920 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5921 (_mm_maskz_sqrt_round_ss): New intrinsics.
5922 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5923 (__builtin_ia32_sqrtsd_mask_round)
5924 (__builtin_ia32_sqrtss_mask_round): New builtins.
5925 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5926 (__builtin_ia32_sqrtss_round): Remove.
5927 (__builtin_ia32_sqrtsd_mask_round)
5928 (__builtin_ia32_sqrtss_mask_round): New builtins.
5929 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5930 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5931 ((match_operand:VF_128 1 "vector_operand"
5932 "xBm,<round_constraint>")): Changed to ...
5933 ((match_operand:VF_128 1 "vector_operand"
5934 "xBm,<round_scalar_constraint>")): ... this.
5935 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5936 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5937 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5938 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5939 %<iptr>1<round_scalar_mask_op3>}): ... this.
5940 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5941 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5943 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5946 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5947 Cast vec_cmpeq result to correct type.
5948 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5949 Cast vec_cmpgt result to correct type.
5951 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5953 * final.c (final_scan_insn_1): Renamed from...
5954 (final_scan_insn): ... this. New wrapper, to recover
5955 seen from the outermost call in recursive ones.
5956 * config/sparc/sparc.c (output_return): Drop seen from call.
5957 (output_sibcall): Likewise.
5958 * config/visium/visium.c (output_branch): Likewise.
5960 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5962 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5965 2018-02-10 Alan Modra <amodra@gmail.com>
5968 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5969 Specify LR as an input.
5971 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5974 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5975 remove_member_access_dummy_vars): New functions.
5976 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5977 lower_omp_1, execute_lower_omp): Use them.
5979 PR rtl-optimization/84308
5980 * shrink-wrap.c (spread_components): Release todo vector.
5982 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5984 PR rtl-optimization/57193
5985 * ira-color.c (struct allocno_color_data): Add member
5986 conflict_allocno_hard_prefs.
5987 (update_conflict_allocno_hard_prefs): New.
5988 (bucket_allocno_compare_func): Add a preference based on
5989 conflict_allocno_hard_prefs.
5990 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5991 (color_allocnos): Remove a dead code. Initiate
5992 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5994 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5997 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5998 constraint from =wa to wa. Avoid a subreg on the output operand,
5999 instead use a pseudo and subreg it in a move.
6000 (p9_xxbrd_<mode>): Changed to ...
6001 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6002 (p9_xxbrd_v2df): New expander.
6003 (p9_xxbrw_<mode>): Changed to ...
6004 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6005 (p9_xxbrw_v4sf): New expander.
6007 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6009 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6011 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6014 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6015 multiply in 32-bit mode.
6016 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6017 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6020 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6022 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6023 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6024 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6025 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6027 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6030 * dwarf2out.c (is_trivial_indirect_ref): New function.
6031 (dwarf2out_late_global_decl): Do not generate a location
6032 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6033 and that are not defined in the current unit.
6035 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6037 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6038 instead of a libcall for UNORDERED.
6040 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6043 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6044 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6046 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6049 * config/s390/s390.c (s390_set_current_function): Invoke
6050 s390_indirect_branch_settings also if fndecl didn't change.
6052 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6054 * config/rs6000/rs6000.md (blockage): Set length to zero.
6056 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6058 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6060 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6063 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6064 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6068 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6069 PARALLEL incoming that failed vt_get_decl_and_offset check.
6072 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6073 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6075 (get_variable_section): For decls in named .bss* sections pass true as
6076 second argument to bss_initializer_p.
6078 2018-02-09 Marek Polacek <polacek@redhat.com>
6079 Jakub Jelinek <jakub@redhat.com>
6082 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6083 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6084 Sync some changes from cxx_fold_indirect_ref.
6086 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6088 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6090 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6091 (BLOCK_INLINE_ENTRY_LABEL): New.
6092 (dwarf2out_var_location): Disregard inline entry markers.
6093 (inline_entry_data): New struct.
6094 (inline_entry_data_hasher): New hashtable type.
6095 (inline_entry_data_hasher::hash): New.
6096 (inline_entry_data_hasher::equal): New.
6097 (inline_entry_data_table): New variable.
6098 (add_high_low_attributes): Add DW_AT_entry_pc and
6099 DW_AT_GNU_entry_view attributes if a pending entry is found
6100 in inline_entry_data_table. Add old entry_pc attribute only
6101 if debug nonbinding markers are disabled.
6102 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6103 markers are enabled.
6104 (block_within_block_p, dwarf2out_inline_entry): New.
6105 (dwarf2out_finish): Check that no entries remained in
6106 inline_entry_data_table.
6107 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6108 (final_scan_insn, notice_source_line): Likewise.
6109 (rest_of_clean_state): Skip inline entry markers.
6110 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6112 * gimple.c (gimple_build_debug_inline_entry): New.
6113 * gimple.h (enum gimple_debug_subcode): Add
6114 GIMPLE_DEBUG_INLINE_ENTRY.
6115 (gimple_build_debug_inline_entry): Declare.
6116 (gimple_debug_inline_entry_p): New.
6117 (gimple_debug_nonbind_marker_p): Adjust.
6118 * insn-notes.def (INLINE_ENTRY): New.
6119 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6120 inline entry marker notes.
6121 (print_insn): Likewise.
6122 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6123 (INSN_DEBUG_MARKER_KIND): Likewise.
6124 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6125 * tree-inline.c (expand_call_inline): Build and insert
6126 debug_inline_entry stmt.
6127 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6128 inline entry blocks early, if nonbind markers are enabled.
6129 (dump_scope_block): Dump fragment info.
6130 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6131 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6132 (gimple_build_debug_inline_entry): New.
6133 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6134 Enable/disable inline entry points too.
6135 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6136 (DEBUG_INSN): Describe inline entry markers.
6138 * common.opt (gvariable-location-views): New.
6139 (gvariable-location-views=incompat5): New.
6140 * config.in: Rebuilt.
6141 * configure: Rebuilt.
6142 * configure.ac: Test assembler for view support.
6143 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6144 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6145 * dwarf2out.c (var_loc_view): New typedef.
6146 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6147 (dwarf2out_locviews_in_attribute): New.
6148 (dwarf2out_locviews_in_loclist): New.
6149 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6150 (enum dw_line_info_opcode): Add LI_adv_address.
6151 (struct dw_line_info_table): Add view.
6152 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6153 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6154 (zero_view_p): New variable.
6155 (ZERO_VIEW_P): New macro.
6156 (output_asm_line_debug_info): New.
6157 (struct var_loc_node): Add view.
6158 (add_AT_view_list, AT_loc_list): New.
6159 (add_var_loc_to_decl): Add view param. Test it against last.
6160 (new_loc_list): Add view params. Record them.
6161 (AT_loc_list_ptr): Handle loc and view lists.
6162 (view_list_to_loc_list_val_node): New.
6163 (print_dw_val): Handle dw_val_class_view_list.
6164 (size_of_die): Likewise.
6165 (value_format): Likewise.
6166 (loc_list_has_views): New.
6167 (gen_llsym): Set vl_symbol too.
6168 (maybe_gen_llsym, skip_loc_list_entry): New.
6169 (dwarf2out_maybe_output_loclist_view_pair): New.
6170 (output_loc_list): Output view list or entries too.
6171 (output_view_list_offset): New.
6172 (output_die): Handle dw_val_class_view_list.
6173 (output_dwarf_version): New.
6174 (output_compilation_unit_header): Use it.
6175 (output_skeleton_debug_sections): Likewise.
6176 (output_rnglists, output_line_info): Likewise.
6177 (output_pubnames, output_aranges): Update version comments.
6178 (output_one_line_info_table): Output view numbers in asm comments.
6179 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6180 Call maybe_gen_llsym.
6181 (loc_list_from_tree_1): Adjust.
6182 (add_AT_location_description): Create view list attribute if
6183 needed, check it's absent otherwise.
6184 (convert_cfa_to_fb_loc_list): Adjust.
6185 (maybe_emit_file): Call output_asm_line_debug_info for test.
6186 (dwarf2out_var_location): Reset views as needed. Precompute
6187 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6188 attribute. Set view.
6189 (new_line_info_table): Reset next view.
6190 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6191 (dwarf2out_source_line): Likewise. Output view resets and labels to
6192 the assembler, or select appropriate line info opcodes.
6193 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6194 (optimize_string_length): Catch it. Adjust.
6195 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6196 dw_val_class_view_list, and remove it if no longer needed.
6197 (hash_loc_list): Hash view numbers.
6198 (loc_list_hasher::equal): Compare them.
6199 (optimize_location_lists): Check whether a view list symbol is
6200 needed, and whether the locview attribute is present, and
6201 whether they match. Remove the locview attribute if no longer
6203 (index_location_lists): Call skip_loc_list_entry for test.
6204 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6205 Use output_dwarf_version.
6206 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6207 (struct dw_val_node): Add val_view_list.
6208 * final.c (SEEN_NEXT_VIEW): New.
6209 (set_next_view_needed): New.
6210 (clear_next_view_needed): New.
6211 (maybe_output_next_view): New.
6212 (final_start_function): Rename to...
6213 (final_start_function_1): ... this. Take pointer to FIRST,
6214 add SEEN parameter. Emit param bindings in the initial view.
6215 (final_start_function): Reintroduce SEEN-less interface.
6216 (final): Rename to...
6217 (final_1): ... this. Take SEEN parameter. Output final pending
6218 next view at the end.
6219 (final): Reintroduce seen-less interface.
6220 (final_scan_insn): Output pending next view before switching
6221 sections or ending a block. Mark the next view as needed when
6222 outputting variable locations. Notify debug backend of section
6223 changes, and of location view changes.
6224 (rest_of_handle_final): Adjust.
6225 * toplev.c (process_options): Autodetect value for debug variable
6226 location views option. Warn on incompat5 without -gdwarf-5.
6227 * doc/invoke.texi (gvariable-location-views): New.
6228 (gvariable-location-views=incompat5): New.
6229 (gno-variable-location-views): New.
6231 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6233 PR tree-optimization/84136
6234 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6235 that the result of find_edge is non-NULL.
6237 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6240 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6241 storing integer register in SImode. Fix cost of 256 and 512
6242 byte aligned SSE register store.
6244 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6246 * config/i386/i386.c (ix86_multiplication_cost): Fix
6247 multiplication cost for TARGET_AVX512DQ.
6249 2018-02-08 Marek Polacek <polacek@redhat.com>
6251 PR tree-optimization/84238
6252 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6255 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6257 PR tree-optimization/84265
6258 * tree-vect-stmts.c (vectorizable_store): Don't treat
6259 VMAT_CONTIGUOUS accesses as grouped.
6260 (vectorizable_load): Likewise.
6262 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6264 PR tree-optimization/81635
6265 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6266 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6267 (test_round_for_mask): New functions.
6268 (wide_int_cc_tests): Call test_round_for_mask.
6269 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6270 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6271 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6272 range returned by get_range_info.
6274 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6277 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6278 * symtab.c: Include builtins.h
6279 (symtab_node::output_to_lto_symbol_table_p): Move here
6280 from lto-streamer-out.c:output_symbol_p.
6281 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6282 (output_symbol_p): Move all logic to symtab.c
6283 (produce_symtab): Update.
6285 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6287 * config/s390/s390-opts.h (enum indirect_branch): Define.
6288 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6289 (s390_indirect_branch_via_thunk)
6290 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6291 (enum s390_indirect_branch_type): Define.
6292 * config/s390/s390.c (struct s390_frame_layout, struct
6293 machine_function): Remove.
6294 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6295 (indirect_branch_table_label_no, indirect_branch_table_name):
6297 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6298 (enum s390_indirect_branch_option): Define.
6299 (s390_return_addr_from_memory): New function.
6300 (s390_handle_string_attribute): New function.
6301 (s390_attribute_table): Add new attribute handler.
6302 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6303 (s390_indirect_branch_via_thunk): New function.
6304 (s390_indirect_branch_via_inline_thunk): New function.
6305 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6306 sibling call optimization for non z10 compiles.
6307 (s390_emit_call): Force indirect branch target to be a single
6308 register. Add r1 clobber for non-z10 compiles.
6309 (s390_emit_epilogue): Emit return jump via return_use expander.
6310 (s390_reorg): Handle JUMP_INSNs as execute targets.
6311 (s390_option_override_internal): Perform validity checks for the
6312 new command line options.
6313 (s390_indirect_branch_attrvalue): New function.
6314 (s390_indirect_branch_settings): New function.
6315 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6316 (s390_output_indirect_thunk_function): New function.
6317 (s390_code_end): Implement target hook.
6318 (s390_case_values_threshold): Implement target hook.
6319 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6321 * config/s390/s390.h (struct s390_frame_layout)
6322 (struct machine_function): Move here from s390.c.
6323 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6324 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6325 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6326 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6327 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6328 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6329 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6330 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6331 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6332 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6333 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6334 (mnemonic attribute): Add values which aren't recognized
6336 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6337 pattern for branch conversion. Fix mnemonic attribute.
6338 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6339 indirect branch via thunk if requested.
6340 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6341 ("*indirect_jump"): Disable for branch conversion using out of
6343 ("indirect_jump_via_thunk<mode>_z10")
6344 ("indirect_jump_via_thunk<mode>")
6345 ("indirect_jump_via_inlinethunk<mode>_z10")
6346 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6347 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6348 ("casesi_jump_via_inlinethunk<mode>_z10")
6349 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6350 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6351 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6352 ("*indirect2_jump"): Disable for branch conversion.
6353 ("casesi_jump"): Turn into expander and expand patterns for branch
6355 ("return_use"): New expander.
6356 ("*return"): Emit return via thunk and rename it to ...
6357 ("*return<mode>"): ... this one.
6358 * config/s390/s390.opt: Add new options and and enum for the
6361 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6363 * lra-constraints.c (match_reload): Unconditionally use
6364 gen_lowpart_SUBREG, rather than selecting between that
6365 and equivalent gen_rtx_SUBREG code.
6367 2018-02-08 Richard Biener <rguenther@suse.de>
6369 PR tree-optimization/84233
6370 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6371 changed flag instead of boguously re-using phi_inserted.
6373 2018-02-08 Martin Jambor <mjambor@suse.cz>
6375 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6376 static local variables.
6378 2018-02-08 Richard Biener <rguenther@suse.de>
6380 PR tree-optimization/84278
6381 * tree-vect-stmts.c (vectorizable_store): When looking for
6382 smaller vector types to perform grouped strided loads/stores
6383 make sure the mode is supported by the target.
6384 (vectorizable_load): Likewise.
6386 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6388 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6389 Increase LDP/STP opportunities by adding adjacent callee-saves.
6391 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6393 PR rtl-optimization/84068
6394 PR rtl-optimization/83459
6395 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6397 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6399 PR tree-optimization/84224
6400 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6401 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6404 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6407 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6408 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6409 count, remove one USE.
6411 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6413 * doc/install.texi (Configuration): Document the
6414 --with-long-double-format={ibm,ieee} PowerPC configuration
6418 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6419 Convert from define_expand to be define_insn_and_split. Rework
6420 float/double/_Float128 conversions to QI/HI/SImode to work with
6421 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6422 conversions to QI/HImode types did a store and then a load to
6423 truncate the value. For conversions to VSX registers, don't split
6424 the insn, instead emit the code directly. Use the code iterator
6425 any_fix to combine signed and unsigned conversions.
6426 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6427 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6428 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6429 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6430 (fix_<mode>di2_hw): Likewise.
6431 (fixuns_<mode>di2_hw): Likewise.
6432 (fix_<mode>si2_hw): Likewise.
6433 (fixuns_<mode>si2_hw): Likewise.
6434 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6435 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6436 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6437 fix<uns>_trunc<SFDF:mode>si2_p8.
6438 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6440 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6441 (fix<uns>_<mode>_mem): Likewise.
6442 (fctiw<u>z_<mode>_mem): Likewise.
6443 (fix<uns>_<mode>_mem): Likewise.
6444 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6445 the register allocator from doing a direct move to the GPRs to do
6446 a store, and instead use the ISA 3.0 store byte/half-word from
6447 vector register instruction. For IEEE 128-bit floating point,
6448 also optimize stores of 32-bit ints.
6449 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6451 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6453 * genextract.c (push_pathstr_operand): New function to support
6455 (walk_rtx): Call push_pathstr_operand.
6456 (print_path): Support [a-zA-Z].
6458 2018-02-07 Richard Biener <rguenther@suse.de>
6460 PR tree-optimization/84037
6461 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6462 (cse_and_gimplify_to_preheader): Declare.
6463 (vect_get_place_in_interleaving_chain): Likewise.
6464 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6466 (_loop_vec_info::~_loop_vec_info): Delete it.
6467 (cse_and_gimplify_to_preheader): New function.
6468 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6469 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6470 (vectorizable_load): Likewise. For grouped stores always base
6471 the IV on the first element.
6472 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6473 condition before gimplifying.
6475 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6477 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6478 *DIV_EXPR and *MOD_EXPR.
6480 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6483 * config/i386/i386.c (ix86_option_override_internal): Mask out
6484 the CF_SET bit when checking -fcf-protection.
6486 2018-02-07 Tom de Vries <tom@codesourcery.com>
6489 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6492 2018-02-07 Richard Biener <rguenther@suse.de>
6494 PR tree-optimization/84204
6495 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6498 PR tree-optimization/84205
6499 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6500 special-case isl_ast_op_zdiv_r.
6502 PR tree-optimization/84223
6503 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6504 Only add conditions from within the region.
6505 (gather_bbs::after_dom_children): Adjust.
6507 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6510 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6511 * config/avr/avr.md: Only post-reload split REG-REG moves if
6512 either register is GENERAL_REG_P.
6514 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6516 PR tree-optimization/84235
6517 * tree-ssa-scopedtables.c
6518 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6519 if the subtraction is performed in floating point type where NaNs are
6520 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6521 build 1. Formatting fix.
6523 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6526 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6527 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6528 and skip it regardless of bb boundaries. Use CALL_P macro,
6529 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6530 unnecessarily, formatting fix.
6532 2018-02-06 Michael Collison <michael.collison@arm.com>
6534 * config/arm/thumb2.md:
6535 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6536 (*thumb_mov_notscc): Ditto.
6538 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6541 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6542 unsigned_fix, not "s".
6544 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6546 * configure.ac (gcc_fn_eh_frame_ro): New function.
6547 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6548 correct .eh_frame permissions.
6549 * configure: Regenerate.
6551 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6553 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6556 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6558 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6559 Display warning message for -mno-speculate-indirect-jumps.
6561 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6563 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6565 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6567 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6569 PR tree-optimization/84225
6570 * tree-eh.c (find_trapping_overflow): Only call
6571 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6573 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6576 * config/i386/i386.c: Reimplement the check of possible options
6577 -mibt/-mshstk conbination. Change error messages.
6578 * doc/invoke.texi: Fix a typo: remove extra '='.
6580 2018-02-06 Marek Polacek <polacek@redhat.com>
6582 PR tree-optimization/84228
6583 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6585 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6588 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6589 emitted arch directives.
6590 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6591 __ARM_FEATURE_COPROC before changing architectures.
6593 2018-02-06 Richard Biener <rguenther@suse.de>
6595 * config/i386/i386.c (print_reg): Fix typo.
6596 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6598 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6600 * configure: Regenerate.
6602 2018-02-05 Martin Sebor <msebor@redhat.com>
6604 PR tree-optimization/83369
6605 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6608 2018-02-05 Martin Liska <mliska@suse.cz>
6610 * doc/invoke.texi: Cherry-pick upstream r323995.
6612 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6614 * ira.c (ira_init_register_move_cost): Adjust comment.
6616 2018-02-05 Martin Liska <mliska@suse.cz>
6618 PR gcov-profile/84137
6619 * doc/gcov.texi: Fix typo in documentation.
6621 2018-02-05 Martin Liska <mliska@suse.cz>
6623 PR gcov-profile/83879
6624 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6625 using dlopen functionality.
6627 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6629 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6630 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6631 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6632 _mm_maskz_range_round_ss): New intrinsics.
6633 (__builtin_ia32_rangesd128_round)
6634 (__builtin_ia32_rangess128_round): Remove.
6635 (__builtin_ia32_rangesd128_mask_round,
6636 __builtin_ia32_rangess128_mask_round): New builtins.
6637 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6638 __builtin_ia32_rangess128_round): Remove.
6639 (__builtin_ia32_rangesd128_mask_round,
6640 __builtin_ia32_rangess128_mask_round): New builtins.
6641 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6642 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6643 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6644 "<round_saeonly_constraint>")): Changed to ...
6645 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6646 "<round_saeonly_scalar_constraint>")): ... this.
6647 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6648 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6649 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6650 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6651 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6653 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6655 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6657 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6658 Remove all values except native, 8540 and 8548.
6660 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6662 * config/i386/i386.c (ix86_output_function_return): Pass
6663 INVALID_REGNUM, instead of -1, as invalid register number to
6664 indirect_thunk_name and output_indirect_thunk.
6666 2018-02-02 Julia Koval <julia.koval@intel.com>
6668 * config.gcc: Add -march=icelake.
6669 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6670 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6671 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6672 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6673 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6674 (processor_target_table): Add icelake.
6675 (ix86_option_override_internal): Handle new PTAs.
6676 (get_builtin_code_for_version): Handle icelake.
6677 (M_INTEL_COREI7_ICELAKE): New.
6678 (fold_builtin_cpu): Handle icelake.
6679 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6680 * doc/invoke.texi: Add -march=icelake.
6682 2018-02-02 Julia Koval <julia.koval@intel.com>
6684 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6685 to wide_int_bitmask.
6686 * wide-int-bitmask.h: New.
6688 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6691 * config/i386/i386.md: Replace Pmode with word_mode in
6692 builtin_setjmp_setup and builtin_longjmp to support x32.
6694 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6698 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6700 (rs6000_supported_cpu_names): New static variable.
6701 (linux_cpu_translation_table): Likewise.
6702 (elf_platform) <cpu>: Define new static variable and use it.
6703 Translate kernel AT_PLATFORM name to canonical name if needed.
6704 Error if platform name is unknown.
6706 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6709 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6711 2018-02-01 Jeff Law <law@redhat.com>
6714 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6715 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6716 the scratch if RELEASE_VIA_POP is false.
6717 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6718 If we have to save a temporary register, decrement SIZE appropriately.
6719 Pass new arguments to release_scratch_register_on_entry.
6720 (ix86_adjust_stack_and_probe): Likewise.
6721 (ix86_emit_probe_stack_range): Pass new arguments to
6722 release_scratch_register_on_entry.
6724 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6726 PR rtl-optimization/84157
6727 * combine.c (change_zero_ext): Use REG_P predicate in
6728 front of HARD_REGISTER_P predicate.
6730 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6732 * config/avr/avr.c (avr_option_override): Move disabling of
6733 -fdelete-null-pointer-checks to...
6734 * common/config/avr/avr-common.c (avr_option_optimization_table):
6737 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6739 PR tree-optimization/81635
6740 * tree-data-ref.c (split_constant_offset_1): For types that
6741 wrap on overflow, try to use range info to prove that wrapping
6744 2018-02-01 Renlin Li <renlin.li@arm.com>
6747 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6749 (aarch64_register_move_cost): Likewise.
6750 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6752 (REG_CLASS_NAMES): Likewise.
6753 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6754 TAILCALL_ADDR_REGS. Remove IP registers.
6755 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6757 2018-02-01 Richard Biener <rguenther@suse.de>
6759 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6760 for specifying RPO order and allow NULL for that.
6761 * domwalk.c (dom_walker::dom_walker): Likewise.
6762 (dom_walker::walk): Handle NULL RPO order.
6763 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6765 (rewrite_update_dom_walker): Likewise.
6766 (mark_def_dom_walker): Likewise.
6768 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6770 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6771 (aarch64_maybe_expand_sve_subreg_move): Declare.
6772 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6773 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6775 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6776 that are semantically a reverse operation.
6777 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6778 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6779 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6781 (aarch64_can_change_mode_class): For big-endian, forbid changes
6782 between two SVE modes if they have different element sizes.
6784 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6786 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6787 the TImode handling for big-endian targets.
6789 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6791 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6792 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6794 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6795 Remove BSWAP handing for big-endian targets and use the form of
6796 LD1RQ appropariate for the mode.
6798 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6800 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6801 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6804 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6807 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6808 check for operands that need to go through aarch64_sve_reload_be.
6810 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6812 PR tree-optimization/81661
6813 PR tree-optimization/84117
6814 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6815 * tree-eh.c: Include gimplify.h.
6816 (find_trapping_overflow, replace_trapping_overflow,
6817 rewrite_to_non_trapping_overflow): New functions.
6818 * tree-vect-loop.c: Include tree-eh.h.
6819 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6820 * tree-data-ref.c: Include tree-eh.h.
6821 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6823 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6825 PR rtl-optimization/84123
6826 * combine.c (change_zero_ext): Check if hard register satisfies
6827 can_change_dest_mode before calling gen_lowpart_SUBREG.
6829 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6832 * ira.c (ira_init_register_move_cost): Remove assert.
6834 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6836 PR rtl-optimization/84071
6837 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6838 * doc/tm.texi: Regenerate.
6840 2018-01-31 Richard Biener <rguenther@suse.de>
6842 PR tree-optimization/84132
6843 * tree-data-ref.c (analyze_miv_subscript): Properly
6844 check whether evolution_function_is_affine_multivariate_p
6845 before calling gcd_of_steps_may_divide_p.
6847 2018-01-31 Julia Koval <julia.koval@intel.com>
6850 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6851 * config/i386/i386.md (rdpid_rex64) New.
6852 (rdpid): Make 32bit only.
6854 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6857 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6858 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6860 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6863 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6865 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6867 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6869 PR rtl-optimization/84071
6870 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6871 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6873 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6875 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6876 (arc_attribute_table): Add 'aux' attribute.
6877 (arc_in_small_data_p): Consider aux like variables.
6878 (arc_is_aux_reg_p): New function.
6879 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6880 (arc_get_aux_arg): New function.
6881 (prepare_move_operands): Handle aux-register access.
6882 (arc_handle_aux_attribute): New function.
6883 * doc/extend.texi (ARC Variable attributes): Add subsection.
6885 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6887 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6888 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6889 (arc_attribute_table): Add 'uncached' attribute.
6890 (arc_print_operand): Print '.di' flag for uncached memory
6892 (arc_in_small_data_p): Do not consider for small data the uncached
6894 (arc_is_uncached_mem_p): New function.
6895 * config/arc/predicates.md (compact_store_memory_operand): Check
6896 for uncached memory accesses.
6897 (nonvol_nonimm_operand): Likewise.
6898 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6900 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6903 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6904 falign-loops=): Add Optimization flag.
6906 2018-01-30 Jeff Law <law@redhat.com>
6909 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6910 INT_REGISTERS_SAVED. Check it prior to calling
6911 get_scratch_register_on_entry.
6912 (ix86_adjust_stack_and_probe): Similarly.
6913 (ix86_emit_probe_stack_range): Similarly.
6914 (ix86_expand_prologue): Corresponding changes.
6916 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6919 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6920 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6922 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6925 * lra-constraints.c (curr_insn_transform): Process AND in the
6928 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6930 PR rtl-optimization/83986
6931 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6932 dependence against last_pending_memory_flush in addition to
6935 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6937 PR tree-optimization/81611
6938 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6941 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6944 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6947 2018-01-30 Richard Biener <rguenther@suse.de>
6948 Jakub Jelinek <jakub@redhat.com>
6950 PR tree-optimization/84111
6951 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6952 inner loops added during recursion, as they don't have up-to-date
6955 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6958 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6959 (can_inline_edge_by_limits_p): ... here.
6960 (can_early_inline_edge_p, check_callers,
6961 update_caller_keys, update_callee_keys, recursive_inlining,
6962 add_new_edges_to_heap, speculation_useful_p,
6963 inline_small_functions,
6964 inline_small_functions, flatten_function,
6965 inline_to_all_callers_1): Update.
6967 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6969 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6972 2018-01-30 Richard Biener <rguenther@suse.de>
6974 PR tree-optimization/83008
6975 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6976 invariant and constant vector uses in stmts when they need
6979 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6982 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6983 * configure: Regenerate.
6985 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6987 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6989 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6990 Use gen_rtx_REG rather than gen_lowpart.
6992 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6994 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6995 rather than 0 when creating partial subregs.
6997 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6999 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7002 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7005 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7006 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7007 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7008 flags. This restores the settings used before the 2017-07-24.
7009 Turning off pre increment/decrement/modify allows IVOPTS to
7010 optimize DF/SF loops where the index is an int.
7012 2018-01-29 Richard Biener <rguenther@suse.de>
7013 Kelvin Nilsen <kelvin@gcc.gnu.org>
7016 * tree-vect-stmts.c (vectorizable_call): Don't call
7017 targetm.vectorize_builtin_md_vectorized_function if callee is
7020 2018-01-22 Carl Love <cel@us.ibm.com>
7022 * doc/extend.tex: Fix typo in second arg in
7023 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7025 2018-01-29 Richard Biener <rguenther@suse.de>
7027 PR tree-optimization/84086
7028 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7029 (flush_ssaname_freelist): When SSA names were released reset
7030 the SCEV hash table.
7032 2018-01-29 Richard Biener <rguenther@suse.de>
7034 PR tree-optimization/84057
7035 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7036 removed paths when removing edges.
7038 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7040 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7041 -mfunction-return=@var{choice}.
7043 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7046 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7048 (layout::print_source_line): Likewise.
7049 (test_get_line_width_without_trailing_whitespace): Add test cases.
7051 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7054 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7057 2018-01-26 Jim Wilson <jimw@sifive.com>
7059 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7061 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7064 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7066 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7067 and CMP + SUB-immediate -> SUBS.
7069 2018-01-26 Martin Sebor <msebor@redhat.com>
7071 PR tree-optimization/83896
7072 * tree-ssa-strlen.c (get_string_len): Rename...
7073 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7074 Avoid assuming length is constant.
7075 (handle_char_store): Use HOST_WIDE_INT for string length.
7077 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7080 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7081 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7083 2018-01-26 Richard Biener <rguenther@suse.de>
7085 PR rtl-optimization/84003
7086 * dse.c (record_store): Only record redundant stores when
7087 the earlier store aliases at least all accesses the later one does.
7089 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7091 PR rtl-optimization/83985
7092 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7093 REG_CFA_RESTORE insns.
7094 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7095 REG_CFA_RESTORE insns here.
7098 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7099 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7101 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7103 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7104 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7105 (arc_init): Likewise.
7106 (arc_override_options): Likewise.
7107 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7109 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7111 * config/arc/arc.h (TARGET_DBNZ): Define.
7112 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7113 properly set the tune attribute.
7114 (dbnz): Use TARGET_DBNZ guard.
7115 * config/arc/arc.opt (mtune): Add core3 option.
7117 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7119 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7120 recognize new pic like addresses.
7121 (arc_delegitimize_address): Clean up.
7123 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7125 * config/arc/arc-arches.def: Option mrf16 valid for all
7127 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7128 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7129 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7130 * config/arc/arc-tables.opt: Regenerate.
7131 * config/arc/arc.c (arc_conditional_register_usage): Handle
7132 reduced register file case.
7133 (arc_file_start): Set must have build attributes.
7134 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7136 * config/arc/arc.opt (mrf16): Add new option.
7137 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7138 * config/arc/genmultilib.awk: Handle new mrf16 option.
7139 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7140 * config/arc/t-multilib: Regenerate.
7141 * doc/invoke.texi (ARC Options): Document mrf16 option.
7143 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7145 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7146 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7147 (arc_attribute_table): Add 'secure_call' attribute.
7148 (arc_print_operand): Print secure call operand.
7149 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7151 (arc_is_secure_call_p): New function. * config/arc/arc.md
7152 (call_i): Add support for sjli instruction.
7153 (call_value_i): Likewise.
7154 * config/arc/constraints.md (Csc): New constraint.
7156 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7157 John Eric Martin <John.Martin@emmicro-us.com>
7159 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7160 * config/arc/arc.c (_arc_jli_section): New struct.
7161 (arc_jli_section): New type.
7162 (rc_jli_sections): New static variable.
7163 (arc_handle_jli_attribute): New function.
7164 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7165 (arc_file_end): New function.
7166 (TARGET_ASM_FILE_END): Define.
7167 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7168 (arc_add_jli_section): New function.
7169 (jli_call_scan): Likewise.
7170 (arc_reorg): Call jli_call_scan.
7171 (arc_output_addsi): Remove 'S' from printing asm operand.
7172 (arc_is_jli_call_p): New function.
7173 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7175 (movhi_insn): Likewise.
7176 (movsi_insn): Likewise.
7177 (movsi_set_cc_insn): Likewise.
7178 (loadqi_update): Likewise.
7179 (load_zeroextendqisi_update): Likewise.
7180 (load_signextendqisi_update): Likewise.
7181 (loadhi_update): Likewise.
7182 (load_zeroextendhisi_update): Likewise.
7183 (load_signextendhisi_update): Likewise.
7184 (loadsi_update): Likewise.
7185 (loadsf_update): Likewise.
7186 (movsicc_insn): Likewise.
7187 (bset_insn): Likewise.
7188 (bxor_insn): Likewise.
7189 (bclr_insn): Likewise.
7190 (bmsk_insn): Likewise.
7191 (bicsi3_insn): Likewise.
7192 (cmpsi_cc_c_insn): Likewise.
7193 (movsi_ne): Likewise.
7194 (movsi_cond_exec): Likewise.
7195 (clrsbsi2): Likewise.
7206 (call_i): Remove 'S' asm letter, add jli instruction.
7207 (call_value_i): Likewise.
7208 * config/arc/arc.op (mjli-always): New option.
7209 * config/arc/constraints.md (Cji): New constraint.
7210 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7212 (subsf3_fpx): Likewise.
7213 (mulsf3_fpx): Likewise.
7214 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7216 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7218 * doc/invoke.texi (ARC): Document mjli-always option.
7220 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7222 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7223 avoid addition with 0 and use incw and decw where possible.
7225 2018-01-26 Richard Biener <rguenther@suse.de>
7227 PR tree-optimization/81082
7228 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7229 association if it requires casting to unsigned.
7230 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7231 from fold_plusminus_mult_expr to catch important cases late when
7232 range info is available.
7234 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7236 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7237 * configure.ac (hidden_linkonce): New test.
7238 * configure: Regenerate.
7239 * config.in: Regenerate.
7241 2018-01-26 Julia Koval <julia.koval@intel.com>
7243 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7244 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7245 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7246 _mm_mask_bitshuffle_epi64_mask): Fix type.
7247 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7248 USI_FTYPE_V4DI_V4DI_USI): Remove.
7249 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7250 __builtin_ia32_vpshufbitqmb256_mask,
7251 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7252 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7253 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7255 2018-01-26 Alan Modra <amodra@gmail.com>
7258 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7259 UNSPEC_VBPERMQ. Sort other unspecs.
7261 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7263 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7265 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7268 * predict.c (drop_profile): Do not push/pop cfun; update also
7270 (handle_missing_profiles): Fix logic looking for zero profiles.
7272 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7275 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7276 on functions with #pragma omp declare simd or functions with simd
7278 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7279 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7280 Remove trailing \n from warning_at calls.
7282 2018-01-25 Tom de Vries <tom@codesourcery.com>
7285 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7286 for neutered workers.
7288 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7291 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7292 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7294 2018-01-24 Jeff Law <law@redhat.com>
7297 * i386.c (get_probe_interval): Move to earlier point.
7298 (ix86_compute_frame_layout): If -fstack-clash-protection and
7299 the frame is larger than the probe interval, then use pushes
7300 to save registers rather than reg->mem moves.
7301 (ix86_expand_prologue): Remove conditional for int_registers_saved
7304 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7307 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7308 min/max for never referenced object.
7310 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7313 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7315 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7316 attributes from DECL_ATTRIBUTES (decl) without affecting
7317 DECL_ATTRIBUTES (current_function_decl).
7318 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7319 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7321 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7323 PR tree-optimization/83979
7324 * fold-const.c (fold_comparison): Use constant_boolean_node
7325 instead of boolean_{true,false}_node.
7327 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7329 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7332 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7334 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7335 Simplify the clause that sets the length attribute.
7336 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7337 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7338 clause that sets the length attribute.
7339 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7341 2018-01-24 Tom de Vries <tom@codesourcery.com>
7344 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7345 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7346 Add strict parameter.
7347 (prevent_branch_around_nothing): Insert dummy insn between branch to
7348 label and label with no ptx insn inbetween.
7349 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7351 2018-01-24 Tom de Vries <tom@codesourcery.com>
7354 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7355 for neutered threads in warp.
7356 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7358 2018-01-24 Richard Biener <rguenther@suse.de>
7360 PR tree-optimization/83176
7361 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7364 2018-01-24 Richard Biener <rguenther@suse.de>
7366 PR tree-optimization/82819
7367 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7368 code generating pluses that are no-ops in the target precision.
7370 2018-01-24 Richard Biener <rguenther@suse.de>
7373 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7375 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7377 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7378 to merge probabilities.
7379 * predict.c (probably_never_executed): Also mark as cold functions
7380 with global 0 profile and guessed local profile.
7381 * profile-count.c (profile_probability::combine_with_count): New
7383 * profile-count.h (profile_probability::operator*,
7384 profile_probability::operator*=, profile_probability::operator/,
7385 profile_probability::operator/=): Reduce precision to adjusted
7386 and set value to guessed on contradictory divisions.
7387 (profile_probability::combine_with_freq): Remove.
7388 (profile_probability::combine_wiht_count): Declare.
7389 (profile_count::force_nonzero):: Set to adjusted.
7390 (profile_count::probability_in):: Set quality to adjusted.
7391 * tree-ssa-tail-merge.c (replace_block_by): Use
7394 2018-01-23 Andrew Waterman <andrew@sifive.com>
7395 Jim Wilson <jimw@sifive.com>
7397 * config/riscv/riscv.c (riscv_stack_boundary): New.
7398 (riscv_option_override): Set riscv_stack_boundary. Handle
7399 riscv_preferred_stack_boundary_arg.
7400 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7401 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7402 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7403 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7404 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7405 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7407 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7410 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7411 of struct ix86_frame.
7412 (ix86_expand_epilogue): Likewise. Add a local variable for
7413 the reg_save_offset field in struct ix86_frame.
7415 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7417 PR tree-optimization/82604
7418 * tree-loop-distribution.c (enum partition_kind): New enum item
7419 PKIND_PARTIAL_MEMSET.
7420 (partition_builtin_p): Support above new enum item.
7421 (generate_code_for_partition): Ditto.
7422 (compute_access_range): Differentiate cases that equality can be
7423 proven at all loops, the innermost loops or no loops.
7424 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7425 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7426 (finalize_partitions, distribute_loop): Don't fuse partition of
7427 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7428 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7431 2018-01-23 Martin Liska <mliska@suse.cz>
7433 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7434 order to ignore the predictor.
7435 (PRED_POLYMORPHIC_CALL): Likewise.
7436 (PRED_RECURSIVE_CALL): Likewise.
7438 2018-01-23 Martin Liska <mliska@suse.cz>
7440 * tree-profile.c (tree_profiling): Print function header to
7441 aware reader which function we are working on.
7442 * value-prof.c (gimple_find_values_to_profile): Do not print
7443 not interesting value histograms.
7445 2018-01-23 Martin Liska <mliska@suse.cz>
7447 * profile-count.h (enum profile_quality): Add
7448 profile_uninitialized as the first value. Do not number values
7449 as they are zero based.
7450 (profile_count::verify): Update sanity check.
7451 (profile_probability::verify): Likewise.
7453 2018-01-23 Nathan Sidwell <nathan@acm.org>
7455 * doc/invoke.texi (ffor-scope): Deprecate.
7457 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7459 PR tree-optimization/83510
7460 * domwalk.c (set_all_edges_as_executable): New function.
7461 (dom_walker::dom_walker): Convert bool param
7462 "skip_unreachable_blocks" to enum reachability. Move setup of
7463 edge flags to set_all_edges_as_executable and only do it when
7464 reachability is REACHABLE_BLOCKS.
7465 * domwalk.h (enum dom_walker::reachability): New enum.
7466 (dom_walker::dom_walker): Convert bool param
7467 "skip_unreachable_blocks" to enum reachability.
7468 (set_all_edges_as_executable): New decl.
7469 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7470 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7472 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7473 but converting true to REACHABLE_BLOCKS.
7474 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7476 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7477 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7478 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7480 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7481 if check_all_array_refs will be called.
7483 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7485 * tree.c (selftest::test_location_wrappers): Add more test
7488 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7490 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7491 (selftest::test_bit_in_range): Likewise.
7493 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7496 * doc/sourcebuild.texi (vect_float): Say that the selector
7497 only describes the situation when -funsafe-math-optimizations is on.
7498 (vect_float_strict): Document.
7500 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7502 PR tree-optimization/83965
7503 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7504 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7505 instead of checking only for a reduction.
7506 (vect_recog_widen_sum_pattern): Likewise.
7508 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7510 * predict.c (probably_never_executed): Only use precise profile info.
7511 (compute_function_frequency): Skip after inlining hack since we now
7512 have quality checking.
7514 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7516 * profile-count.h (profile_probability::very_unlikely,
7517 profile_probability::unlikely, profile_probability::even): Set
7518 precision to guessed.
7520 2018-01-23 Richard Biener <rguenther@suse.de>
7522 PR tree-optimization/83963
7523 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7524 Properly terminate dominator walk when crossing the exit edge not
7525 when visiting its source block.
7527 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7530 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7531 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7533 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7535 PR tree-optimization/83957
7536 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7537 semicolon after for body surrounded by braces.
7539 PR tree-optimization/83081
7540 * profile-count.h (profile_probability::split): New method.
7541 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7542 Use profile_probability::split.
7543 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7544 when splitting a single conditional jump into 2.
7546 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7548 PR tree-optimization/69452
7549 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7552 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7554 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7555 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7556 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7558 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7560 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7561 * config/rl78/rl78.md (movdi): New define_expand.
7562 * config/rl78/rl78.c (rl78_split_movdi): New function.
7564 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7567 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7569 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7570 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7571 128-bit to produce an UNSPEC move to get the double word with the
7572 signbit and then a shift directly to do signbit.
7573 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7574 implementation with a new version that just does either a direct
7575 move or a regular move. Move memory interface to separate insns.
7576 Move insns so they are next to the expander.
7577 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7578 with signbit move. Split big and little endian case.
7579 (signbit<mode>2_dm_mem_le): Likewise.
7580 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7581 (signbit<mode>2_dm2): Likewise.
7583 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7585 * config/rl78/rl78.md (anddi3): New define_expand.
7587 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7589 * config/rl78/rl78.md (umindi3): New define_expand.
7591 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7593 * config/rl78/rl78.md (smindi3): New define_expand.
7595 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7597 * config/rl78/rl78.md (smaxdi3): New define_expand.
7599 2018-01-22 Carl Love <cel@us.ibm.com>
7601 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7602 LVX_V1TI): Add macro expansion.
7603 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7604 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7605 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7606 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7607 Change check to determine if the instruction is a byte reversing
7608 entry. Fix typo in comment.
7609 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7610 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7611 Add def_builtin calls for new builtins.
7612 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7613 Add define_insn expansion.
7615 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7617 * config/rl78/rl78.md (umaxdi3): New define_expand.
7619 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7621 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7622 for non-QImode registers.
7624 2018-01-22 Richard Biener <rguenther@suse.de>
7626 PR tree-optimization/83963
7627 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7628 including the loop exit block.
7629 (scop_detection::merge_sese): Likewise.
7630 (scop_detection::add_scop): Do it here instead.
7632 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7634 * doc/sourcebuild.texi (arm_softfloat): Document.
7636 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7639 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7640 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7641 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7643 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7644 David Edelsohn <dje.gcc@gmail.com>
7647 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7648 Change "crset eq" to "crset 2".
7649 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7650 (*call_indirect_aix<mode>_nospec): Likewise.
7651 (*call_value_indirect_aix<mode>_nospec): Likewise.
7652 (*call_indirect_elfv2<mode>_nospec): Likewise.
7653 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7654 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7655 change assembly output from . to $.
7656 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7657 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7658 (*tablejump<mode>_internal1_nospec): Likewise.
7660 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7663 * config/sh/sh_optimize_sett_clrt.cc:
7664 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7666 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7668 PR tree-optimization/83940
7669 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7670 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7671 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7672 use it to pass back the definition type.
7673 (vect_check_store_rhs): Likewise rhs_dt_out.
7674 (vect_build_gather_load_calls): Add a mask_dt argument and use
7675 it instead of a call to vect_is_simple_use.
7676 (vectorizable_store): Update calls to vect_check_load_store_mask
7677 and vect_check_store_rhs. Use the dt returned by the latter instead
7678 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7679 instead of calls to vect_is_simple_use. Pass the scalar rather
7680 than the vector operand to vect_is_simple_use when handling
7681 second and subsequent copies of an rhs value.
7682 (vectorizable_load): Update calls to vect_check_load_store_mask
7683 and vect_build_gather_load_calls. Use the cached mask_dt and
7684 gs_info.offset_dt instead of calls to vect_is_simple_use.
7686 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7689 * tree-emutls.c: Include gimplify.h.
7690 (lower_emutls_2): New function.
7691 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7692 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7693 it before further processing.
7696 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7697 UINTVAL (trueop1) instead of INTVAL (op1).
7699 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7703 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7704 INCOMING_FRAME_SP_OFFSET if not defined.
7705 (scan_trace): Add ENTRY argument. If true and
7706 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7707 emit a note to adjust the CFA offset.
7708 (create_cfi_notes): Adjust scan_trace callers.
7709 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7710 INCOMING_FRAME_SP_OFFSET in the CIE.
7711 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7712 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7714 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7715 * doc/tm.texi: Regenerated.
7717 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7719 PR rtl-optimization/83147
7720 * lra-constraints.c (remove_inheritance_pseudos): Use
7721 lra_substitute_pseudo_within_insn.
7723 2018-01-19 Tom de Vries <tom@codesourcery.com>
7724 Cesar Philippidis <cesar@codesourcery.com>
7727 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7729 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7732 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7733 spaces for function labels.
7735 2018-01-19 Martin Liska <mliska@suse.cz>
7737 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7738 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7739 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7740 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7741 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7742 (PRED_CONST_RETURN): Change from 69 to 65.
7743 (PRED_NULL_RETURN): Change from 91 to 71.
7744 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7745 (PRED_LOOP_GUARD): Change from 66 to 73.
7747 2018-01-19 Martin Liska <mliska@suse.cz>
7749 * predict.c (predict_insn_def): Add new assert.
7750 (struct branch_predictor): Change type to signed integer.
7751 (test_prediction_value_range): Amend test to cover
7753 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7754 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7755 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7756 (PRED_LOOP_IV_COMPARE): Likewise.
7757 * predict.h (PROB_UNINITIALIZED): Define new constant.
7759 2018-01-19 Martin Liska <mliska@suse.cz>
7761 * predict.c (dump_prediction): Add new format for
7762 analyze_brprob.py script which is enabled with -details
7764 * profile-count.h (precise_p): New function.
7766 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7768 PR tree-optimization/83922
7769 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7770 there are no statements that need masking.
7771 (vect_active_double_reduction_p): New function.
7772 (vect_analyze_loop_operations): Use it when handling phis that
7773 are not in the loop header.
7775 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7777 PR tree-optimization/83914
7778 * tree-vect-loop.c (vectorizable_induction): Don't convert
7779 init_expr or apply the peeling adjustment for inductions
7780 that are nested within the vectorized loop.
7782 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7784 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7787 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7791 * function.h (gimplify_parameters): Add gimple_seq * argument.
7792 * function.c: Include gimple.h and options.h.
7793 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7794 for the added local temporaries if needed.
7795 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7796 if there are any parameter cleanups, wrap whole body into a
7797 try/finally with the cleanups.
7799 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7802 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7803 Use GET_MODE_CLASS for scalar floating point.
7805 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7809 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7810 Fix call of call_cgraph_insertion_hooks.
7812 2018-01-18 Martin Sebor <msebor@redhat.com>
7814 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7816 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7819 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7822 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7825 * common.opt: (-ffile-prefix-map): New option.
7826 * opts.c (common_handle_option): Defer it.
7827 * opts-global.c (handle_common_deferred_options): Handle it.
7828 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7829 * file-prefix-map.h: New file.
7830 (remap_debug_filename, add_debug_prefix_map): ...here.
7831 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7832 * final.c (debug_prefix_map, add_debug_prefix_map
7833 remap_debug_filename): Move to...
7834 * file-prefix-map.c: New file.
7835 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7836 generalize, get rid of alloca(), use strrchr() instead of strchr().
7837 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7838 Implement in terms of add_prefix_map().
7839 (remap_macro_filename, remap_debug_filename): Implement in term of
7841 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7842 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7843 * dbxout.c: Include file-prefix-map.h.
7844 * varasm.c: Likewise.
7845 * vmsdbgout.c: Likewise.
7846 * xcoffout.c: Likewise.
7847 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7848 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7849 * doc/invoke.texi (-ffile-prefix-map): Document.
7850 (-fdebug-prefix-map): Update description.
7852 2018-01-18 Martin Liska <mliska@suse.cz>
7854 * config/i386/i386.c (indirect_thunk_name): Document that also
7856 (output_indirect_thunk): Document why both instructions
7857 (pause and lfence) are generated.
7859 2018-01-18 Richard Biener <rguenther@suse.de>
7861 PR tree-optimization/83887
7862 * graphite-scop-detection.c
7863 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7864 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7865 (scop_detection::merge_sese): Re-implement with a flood-fill
7866 algorithm that properly finds a SESE region if it exists.
7868 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7871 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7872 pointer_diff optimizations use view_convert instead of convert.
7874 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7876 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7877 Generate different code for -mno-speculate-indirect-jumps.
7878 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7879 (*call_indirect_aix<mode>): Disable for
7880 -mno-speculate-indirect-jumps.
7881 (*call_indirect_aix<mode>_nospec): New define_insn.
7882 (*call_value_indirect_aix<mode>): Disable for
7883 -mno-speculate-indirect-jumps.
7884 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7885 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7886 -mno-speculate-indirect-jumps.
7887 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7889 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7891 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7892 long double type, set the flags for noting the default long double
7893 type, even if we don't pass or return a long double type.
7895 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7898 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7901 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7903 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7904 support for merge[hl].
7905 (fold_mergehl_helper): New helper function.
7906 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7907 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7908 (altivec_vmrglw_direct): Add xxmrglw insn.
7910 2018-01-17 Andrew Waterman <andrew@sifive.com>
7912 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7913 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7915 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7918 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7919 call the lto_location_cache before reading the
7920 DECL_SOURCE_LOCATION of the types.
7922 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7923 Richard Sandiford <richard.sandiford@linaro.org>
7925 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7926 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7927 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7928 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7929 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7931 * config/aarch64/constraints.md (aarch64_movti_operand):
7933 * config/aarch64/predicates.md (Uti): Add new constraint.
7935 2018-01-17 Carl Love <cel@us.ibm.com>
7937 * config/rs6000/vsx.md (define_expand xl_len_r,
7938 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7939 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7941 (define_expand, define_insn): Move the shift left from the
7942 define_insn to the define_expand for lxvl and stxvl instructions.
7943 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7944 and XL_LEN_R definitions to PURE.
7946 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7948 * config/i386/i386.c (indirect_thunk_name): Declare regno
7949 as unsigned int. Compare regno with INVALID_REGNUM.
7950 (output_indirect_thunk): Ditto.
7951 (output_indirect_thunk_function): Ditto.
7952 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7953 in the call to output_indirect_thunk_function.
7955 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7958 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7959 rather than the size of inner_type to determine the stack slot size
7960 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7962 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7965 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7968 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7970 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7971 endian Linux systems to optionally enable multilibs for selecting
7972 the long double type if the user configured an explicit type.
7973 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7974 have no long double multilibs if not defined.
7975 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7976 warn if the user used -mabi={ieee,ibm}longdouble and we built
7977 multilibs for long double.
7978 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7979 appropriate multilib option.
7980 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7982 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7983 for building long double multilibs.
7984 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7986 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7988 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7991 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7993 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7996 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7999 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8002 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8004 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8005 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8007 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8009 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8010 different rtl trees depending on TARGET_64BIT.
8011 (rs6000_gen_lvx): Likewise.
8013 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8015 * config/visium/visium.md (nop): Tweak comment.
8016 (hazard_nop): Likewise.
8018 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8020 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8021 -mspeculate-indirect-jumps.
8022 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8023 for -mno-speculate-indirect-jumps.
8024 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8025 (*call_value_indirect_elfv2<mode>): Disable for
8026 -mno-speculate-indirect-jumps.
8027 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8028 (indirect_jump): Emit different RTL for
8029 -mno-speculate-indirect-jumps.
8030 (*indirect_jump<mode>): Disable for
8031 -mno-speculate-indirect-jumps.
8032 (*indirect_jump<mode>_nospec): New define_insn.
8033 (tablejump): Emit different RTL for
8034 -mno-speculate-indirect-jumps.
8035 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8036 (tablejumpsi_nospec): New define_expand.
8037 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8038 (tablejumpdi_nospec): New define_expand.
8039 (*tablejump<mode>_internal1): Disable for
8040 -mno-speculate-indirect-jumps.
8041 (*tablejump<mode>_internal1_nospec): New define_insn.
8042 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8045 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8047 * caller-save.c (insert_save): Drop unnecessary parameter. All
8050 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8051 Richard Biener <rguenth@suse.de>
8054 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8055 return early, inline manually is_gimple_sizepos. Make sure if we
8056 call gimplify_expr we don't end up with a gimple constant.
8057 * tree.c (variably_modified_type_p): Don't return true for
8058 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8059 * gimplify.h (is_gimple_sizepos): Remove.
8061 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8063 PR tree-optimization/83857
8064 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8065 vectorizable_live_operation for pure SLP statements.
8066 (vectorizable_live_operation): Handle PHIs.
8068 2018-01-16 Richard Biener <rguenther@suse.de>
8070 PR tree-optimization/83867
8071 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8072 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8074 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8077 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8078 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8079 If off is not INTEGER_CST, issue a may not be aligned warning
8080 rather than isn't aligned. Use isn%'t rather than isn't.
8081 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8083 <case MULT_EXPR>: Improve the case when bottom and one of the
8084 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8085 operand, in that case check if the other operand is multiple of
8086 bottom divided by the INTEGER_CST operand.
8088 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8091 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8092 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8093 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8094 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8095 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8096 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8097 (pa_function_arg_size): New function.
8099 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8101 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8102 in a separate statement.
8104 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8106 PR tree-optimization/83847
8107 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8108 group gathers and scatters.
8110 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8112 PR rtl-optimization/86620
8113 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8115 PR rtl-optimization/83213
8116 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8117 to last if both are JUMP_INSNs.
8119 PR tree-optimization/83843
8120 * gimple-ssa-store-merging.c
8121 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8122 store_immediate_info for bswap/nop orig_stores.
8124 2018-01-15 Andrew Waterman <andrew@sifive.com>
8126 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8128 <UDIV>: Increase cost if !TARGET_DIV.
8130 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8132 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8133 (define_attr "cr_logical_3op"): New.
8134 (cceq_ior_compare): Adjust.
8135 (cceq_ior_compare_complement): Adjust.
8136 (*cceq_rev_compare): Adjust.
8137 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8138 (is_cracked_insn): Adjust.
8139 (insn_must_be_first_in_group): Adjust.
8140 * config/rs6000/40x.md: Adjust.
8141 * config/rs6000/440.md: Adjust.
8142 * config/rs6000/476.md: Adjust.
8143 * config/rs6000/601.md: Adjust.
8144 * config/rs6000/603.md: Adjust.
8145 * config/rs6000/6xx.md: Adjust.
8146 * config/rs6000/7450.md: Adjust.
8147 * config/rs6000/7xx.md: Adjust.
8148 * config/rs6000/8540.md: Adjust.
8149 * config/rs6000/cell.md: Adjust.
8150 * config/rs6000/e300c2c3.md: Adjust.
8151 * config/rs6000/e500mc.md: Adjust.
8152 * config/rs6000/e500mc64.md: Adjust.
8153 * config/rs6000/e5500.md: Adjust.
8154 * config/rs6000/e6500.md: Adjust.
8155 * config/rs6000/mpc.md: Adjust.
8156 * config/rs6000/power4.md: Adjust.
8157 * config/rs6000/power5.md: Adjust.
8158 * config/rs6000/power6.md: Adjust.
8159 * config/rs6000/power7.md: Adjust.
8160 * config/rs6000/power8.md: Adjust.
8161 * config/rs6000/power9.md: Adjust.
8162 * config/rs6000/rs64.md: Adjust.
8163 * config/rs6000/titan.md: Adjust.
8165 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8167 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8168 ix86_indirect_branch_register logic.
8170 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8172 * config/i386/constraints.md (Bs): Update
8173 ix86_indirect_branch_register check. Don't check
8174 ix86_indirect_branch_register with GOT_memory_operand.
8176 * config/i386/predicates.md (GOT_memory_operand): Don't check
8177 ix86_indirect_branch_register here.
8178 (GOT32_symbol_operand): Likewise.
8180 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8182 * config/i386/predicates.md (constant_call_address_operand):
8183 Rewrite ix86_indirect_branch_register logic.
8184 (sibcall_insn_operand): Likewise.
8186 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8188 * config/i386/constraints.md (Bs): Replace
8189 ix86_indirect_branch_thunk_register with
8190 ix86_indirect_branch_register.
8192 * config/i386/i386.md (indirect_jump): Likewise.
8193 (tablejump): Likewise.
8194 (*sibcall_memory): Likewise.
8195 (*sibcall_value_memory): Likewise.
8196 Peepholes of indirect call and jump via memory: Likewise.
8197 * config/i386/i386.opt: Likewise.
8198 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8199 (GOT_memory_operand): Likewise.
8200 (call_insn_operand): Likewise.
8201 (sibcall_insn_operand): Likewise.
8202 (GOT32_symbol_operand): Likewise.
8204 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8207 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8208 type rather than type addr's type points to.
8209 (expand_omp_atomic_mutex): Likewise.
8210 (expand_omp_atomic): Likewise.
8212 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8215 * config/i386/i386.c (output_indirect_thunk_function): Use
8216 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8217 for __x86_return_thunk.
8219 2018-01-15 Richard Biener <rguenther@suse.de>
8222 * expmed.c (extract_bit_field_1): Fix typo.
8224 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8227 * config/arm/iterators.md (VF): New mode iterator.
8228 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8229 Remove integer-related logic from pattern.
8230 (neon_vabd<mode>_3): Likewise.
8232 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8235 * common.opt (fstrict-overflow): No longer an alias.
8236 (fwrapv-pointer): New option.
8237 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8238 also for pointer types based on flag_wrapv_pointer.
8239 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8240 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8241 opts->x_flag_wrapv got set.
8242 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8243 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8244 POINTER_TYPE_OVERFLOW_UNDEFINED.
8245 * match.pd: Likewise in address comparison pattern.
8246 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8248 2018-01-15 Richard Biener <rguenther@suse.de>
8251 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8252 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8253 Reset type names to their identifier if their TYPE_DECL doesn't
8254 have linkage (and thus is used for ODR and devirt).
8255 (save_debug_info_for_decl): Remove.
8256 (save_debug_info_for_type): Likewise.
8257 (add_tree_to_fld_list): Adjust.
8258 * tree-pretty-print.c (dump_generic_node): Make dumping of
8259 type names more robust.
8261 2018-01-15 Richard Biener <rguenther@suse.de>
8263 * BASE-VER: Bump to 8.0.1.
8265 2018-01-14 Martin Sebor <msebor@redhat.com>
8268 * builtins.c (check_access): Avoid warning when the no-warning bit
8271 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8273 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8274 * ira-color (allocno_hard_regs_compare): Likewise.
8276 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8279 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8280 Use .pushsection/.popsection.
8282 2018-01-14 Martin Sebor <msebor@redhat.com>
8285 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8287 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8289 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8290 entry from extra_headers.
8291 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8292 extra_headers, make the list bitwise identical to the i?86-*-* one.
8294 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8296 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8297 -mcmodel=large with -mindirect-branch=thunk,
8298 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8299 -mfunction-return=thunk-extern.
8300 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8301 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8302 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8304 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8306 * config/i386/i386.c (print_reg): Print the name of the full
8307 integer register without '%'.
8308 (ix86_print_operand): Handle 'V'.
8309 * doc/extend.texi: Document 'V' modifier.
8311 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8313 * config/i386/constraints.md (Bs): Disallow memory operand for
8314 -mindirect-branch-register.
8316 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8317 (GOT_memory_operand): Likewise.
8318 (call_insn_operand): Likewise.
8319 (sibcall_insn_operand): Likewise.
8320 (GOT32_symbol_operand): Likewise.
8321 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8322 for -mindirect-branch-register.
8323 (tablejump): Likewise.
8324 (*sibcall_memory): Likewise.
8325 (*sibcall_value_memory): Likewise.
8326 Disallow peepholes of indirect call and jump via memory for
8327 -mindirect-branch-register.
8328 (*call_pop): Replace m with Bw.
8329 (*call_value_pop): Likewise.
8330 (*sibcall_pop_memory): Replace m with Bs.
8331 * config/i386/i386.opt (mindirect-branch-register): New option.
8332 * doc/invoke.texi: Document -mindirect-branch-register option.
8334 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8336 * config/i386/i386-protos.h (ix86_output_function_return): New.
8337 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8338 set function_return_type.
8339 (indirect_thunk_name): Add ret_p to indicate thunk for function
8341 (output_indirect_thunk_function): Pass false to
8342 indirect_thunk_name.
8343 (ix86_output_indirect_branch_via_reg): Likewise.
8344 (ix86_output_indirect_branch_via_push): Likewise.
8345 (output_indirect_thunk_function): Create alias for function
8346 return thunk if regno < 0.
8347 (ix86_output_function_return): New function.
8348 (ix86_handle_fndecl_attribute): Handle function_return.
8349 (ix86_attribute_table): Add function_return.
8350 * config/i386/i386.h (machine_function): Add
8351 function_return_type.
8352 * config/i386/i386.md (simple_return_internal): Use
8353 ix86_output_function_return.
8354 (simple_return_internal_long): Likewise.
8355 * config/i386/i386.opt (mfunction-return=): New option.
8356 (indirect_branch): Mention -mfunction-return=.
8357 * doc/extend.texi: Document function_return function attribute.
8358 * doc/invoke.texi: Document -mfunction-return= option.
8360 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8362 * config/i386/i386-opts.h (indirect_branch): New.
8363 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8364 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8365 with local indirect jump when converting indirect call and jump.
8366 (ix86_set_indirect_branch_type): New.
8367 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8368 (indirectlabelno): New.
8369 (indirect_thunk_needed): Likewise.
8370 (indirect_thunk_bnd_needed): Likewise.
8371 (indirect_thunks_used): Likewise.
8372 (indirect_thunks_bnd_used): Likewise.
8373 (INDIRECT_LABEL): Likewise.
8374 (indirect_thunk_name): Likewise.
8375 (output_indirect_thunk): Likewise.
8376 (output_indirect_thunk_function): Likewise.
8377 (ix86_output_indirect_branch_via_reg): Likewise.
8378 (ix86_output_indirect_branch_via_push): Likewise.
8379 (ix86_output_indirect_branch): Likewise.
8380 (ix86_output_indirect_jmp): Likewise.
8381 (ix86_code_end): Call output_indirect_thunk_function if needed.
8382 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8384 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8385 (ix86_attribute_table): Add indirect_branch.
8386 * config/i386/i386.h (machine_function): Add indirect_branch_type
8387 and has_local_indirect_jump.
8388 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8390 (tablejump): Likewise.
8391 (*indirect_jump): Use ix86_output_indirect_jmp.
8392 (*tablejump_1): Likewise.
8393 (simple_return_indirect_internal): Likewise.
8394 * config/i386/i386.opt (mindirect-branch=): New option.
8395 (indirect_branch): New.
8398 (thunk-inline): Likewise.
8399 (thunk-extern): Likewise.
8400 * doc/extend.texi: Document indirect_branch function attribute.
8401 * doc/invoke.texi: Document -mindirect-branch= option.
8403 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8406 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8408 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8410 * ipa-inline.c (want_inline_small_function_p): Return false if
8411 inlining has already failed with CIF_FINAL_ERROR.
8412 (update_caller_keys): Call want_inline_small_function_p before
8414 (update_callee_keys): Likewise.
8416 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8418 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8420 (rs6000_quadword_masked_address_p): Likewise.
8421 (quad_aligned_load_p): Likewise.
8422 (quad_aligned_store_p): Likewise.
8423 (const_load_sequence_p): Add comment to describe the outer-most loop.
8424 (mimic_memory_attributes_and_flags): New function.
8425 (rs6000_gen_stvx): Likewise.
8426 (replace_swapped_aligned_store): Likewise.
8427 (rs6000_gen_lvx): Likewise.
8428 (replace_swapped_aligned_load): Likewise.
8429 (replace_swapped_load_constant): Capitalize argument name in
8430 comment describing this function.
8431 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8432 and stores that access quad-word aligned addresses and replace
8433 with stvx or lvx instructions when appropriate.
8434 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8435 New function prototype.
8436 (rs6000_quadword_masked_address_p): Likewise.
8437 (rs6000_gen_lvx): Likewise.
8438 (rs6000_gen_stvx): Likewise.
8439 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8440 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8441 when memory address is aligned.
8442 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8443 this split to select lvx instruction when memory address is aligned.
8444 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8445 instruction when memory address is aligned.
8446 (*vsx_le_perm_load_v16qi): Likewise.
8447 (four unnamed splitters): Modify to select the stvx instruction
8448 when memory is aligned.
8450 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8452 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8453 which appears in the queue multiple times.
8455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8456 Alan Hayward <alan.hayward@arm.com>
8457 David Sherwood <david.sherwood@arm.com>
8459 * tree-vectorizer.h (vec_lower_bound): New structure.
8460 (_loop_vec_info): Add check_nonzero and lower_bounds.
8461 (LOOP_VINFO_CHECK_NONZERO): New macro.
8462 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8463 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8464 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8465 fields. Make seg_len the distance travelled, not including the
8467 (dr_direction_indicator): Declare.
8468 (dr_zero_step_indicator): Likewise.
8469 (dr_known_forward_stride_p): Likewise.
8470 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8472 (runtime_alias_check_p): Allow runtime alias checks with
8474 (operator ==): Compare access_size and align.
8475 (prune_runtime_alias_test_list): Rework for new distinction between
8476 the access_size and seg_len.
8477 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8479 (get_segment_min_max): New function.
8480 (create_intersect_range_checks): Use it.
8481 (dr_step_indicator): New function.
8482 (dr_direction_indicator): Likewise.
8483 (dr_zero_step_indicator): Likewise.
8484 (dr_known_forward_stride_p): Likewise.
8485 * tree-loop-distribution.c (data_ref_segment_size): Return
8486 DR_STEP * (niters - 1).
8487 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8489 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8490 (vect_preserves_scalar_order_p): New function, split out from...
8491 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8492 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8493 (vect_vfa_access_size): New function.
8494 (vect_vfa_align): Likewise.
8495 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8496 (dump_lower_bound): New function.
8497 (vect_check_lower_bound): Likewise.
8498 (vect_small_gap_p): Likewise.
8499 (vectorizable_with_step_bound_p): Likewise.
8500 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8501 depencies if the vectorization factor is 1. Convert the checks
8502 for nonzero steps into checks on the bounds of DR_STEP. Try using
8503 a bunds check for variable steps if the minimum required step is
8504 relatively small. Update calls to the dr_with_seg_len
8505 constructor and to vect_compile_time_alias.
8506 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8508 (vect_loop_versioning): Call it.
8509 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8511 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8514 Alan Hayward <alan.hayward@arm.com>
8515 David Sherwood <david.sherwood@arm.com>
8517 * doc/sourcebuild.texi (vect_scatter_store): Document.
8518 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8520 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8522 * genopinit.c (main): Add supports_vec_scatter_store and
8523 supports_vec_scatter_store_cached to target_optabs.
8524 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8525 IFN_MASK_SCATTER_STORE.
8526 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8528 * internal-fn.h (internal_store_fn_p): Declare.
8529 (internal_fn_stored_value_index): Likewise.
8530 * internal-fn.c (scatter_store_direct): New macro.
8531 (expand_scatter_store_optab_fn): New function.
8532 (direct_scatter_store_optab_supported_p): New macro.
8533 (internal_store_fn_p): New function.
8534 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8535 IFN_MASK_SCATTER_STORE.
8536 (internal_fn_mask_index): Likewise.
8537 (internal_fn_stored_value_index): New function.
8538 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8540 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8541 * optabs-query.c (supports_vec_scatter_store_p): New function.
8542 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8543 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8544 true for scatter stores.
8545 (vect_gather_scatter_fn_p): Handle scatter stores too.
8546 (vect_check_gather_scatter): Consider using scatter stores if
8547 supports_vec_scatter_store_p.
8548 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8550 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8551 internal_fn_stored_value_index.
8552 (check_load_store_masking): Handle scatter stores too.
8553 (vect_get_store_rhs): Make public.
8554 (vectorizable_call): Use internal_store_fn_p.
8555 (vectorizable_store): Handle scatter store internal functions.
8556 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8557 when deciding whether the end of the group has been reached.
8558 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8559 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8560 (mask_scatter_store<mode>): New insns.
8562 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8563 Alan Hayward <alan.hayward@arm.com>
8564 David Sherwood <david.sherwood@arm.com>
8566 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8567 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8568 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8570 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8571 Use vect_truncate_gather_scatter_offset if we can't treat the
8572 operation as a normal gather load or scatter store.
8573 (get_group_load_store_type): Take the gather_scatter_info
8574 as argument. Try using a gather load or scatter store for
8575 single-element groups.
8576 (get_load_store_type): Update calls to get_group_load_store_type
8577 and vect_use_strided_gather_scatters_p.
8579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8580 Alan Hayward <alan.hayward@arm.com>
8581 David Sherwood <david.sherwood@arm.com>
8583 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8584 optional tree argument.
8585 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8587 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8588 but continue to use the current value as a fallback.
8589 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8590 to compare the updates.
8591 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8592 (get_load_store_type): Use it when handling a strided access.
8593 (vect_get_strided_load_store_ops): New function.
8594 (vect_get_data_ptr_increment): Likewise.
8595 (vectorizable_load): Handle strided gather loads. Always pass
8596 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8598 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8599 Alan Hayward <alan.hayward@arm.com>
8600 David Sherwood <david.sherwood@arm.com>
8602 * doc/md.texi (gather_load@var{m}): Document.
8603 (mask_gather_load@var{m}): Likewise.
8604 * genopinit.c (main): Add supports_vec_gather_load and
8605 supports_vec_gather_load_cached to target_optabs.
8606 * optabs-tree.c (init_tree_optimization_optabs): Use
8607 ggc_cleared_alloc to allocate target_optabs.
8608 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8609 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8611 * internal-fn.h (internal_load_fn_p): Declare.
8612 (internal_gather_scatter_fn_p): Likewise.
8613 (internal_fn_mask_index): Likewise.
8614 (internal_gather_scatter_fn_supported_p): Likewise.
8615 * internal-fn.c (gather_load_direct): New macro.
8616 (expand_gather_load_optab_fn): New function.
8617 (direct_gather_load_optab_supported_p): New macro.
8618 (direct_internal_fn_optab): New function.
8619 (internal_load_fn_p): Likewise.
8620 (internal_gather_scatter_fn_p): Likewise.
8621 (internal_fn_mask_index): Likewise.
8622 (internal_gather_scatter_fn_supported_p): Likewise.
8623 * optabs-query.c (supports_at_least_one_mode_p): New function.
8624 (supports_vec_gather_load_p): Likewise.
8625 * optabs-query.h (supports_vec_gather_load_p): Declare.
8626 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8627 and memory_type field.
8628 (NUM_PATTERNS): Bump to 15.
8629 * tree-vect-data-refs.c: Include internal-fn.h.
8630 (vect_gather_scatter_fn_p): New function.
8631 (vect_describe_gather_scatter_call): Likewise.
8632 (vect_check_gather_scatter): Try using internal functions for
8633 gather loads. Recognize existing calls to a gather load function.
8634 (vect_analyze_data_refs): Consider using gather loads if
8635 supports_vec_gather_load_p.
8636 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8637 (vect_get_gather_scatter_offset_type): Likewise.
8638 (vect_convert_mask_for_vectype): Likewise.
8639 (vect_add_conversion_to_patterm): Likewise.
8640 (vect_try_gather_scatter_pattern): Likewise.
8641 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8642 (vect_vect_recog_func_ptrs): Add it.
8643 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8644 internal_fn_mask_index and internal_gather_scatter_fn_p.
8645 (check_load_store_masking): Take the gather_scatter_info as an
8646 argument and handle gather loads.
8647 (vect_get_gather_scatter_ops): New function.
8648 (vectorizable_call): Check internal_load_fn_p.
8649 (vectorizable_load): Likewise. Handle gather load internal
8651 (vectorizable_store): Update call to check_load_store_masking.
8652 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8653 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8654 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8655 (aarch64_gather_scale_operand_d): New predicates.
8656 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8657 (mask_gather_load<mode>): New insns.
8659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8660 Alan Hayward <alan.hayward@arm.com>
8661 David Sherwood <david.sherwood@arm.com>
8663 * optabs.def (fold_left_plus_optab): New optab.
8664 * doc/md.texi (fold_left_plus_@var{m}): Document.
8665 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8666 * internal-fn.c (fold_left_direct): Define.
8667 (expand_fold_left_optab_fn): Likewise.
8668 (direct_fold_left_optab_supported_p): Likewise.
8669 * fold-const-call.c (fold_const_fold_left): New function.
8670 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8671 * tree-parloops.c (valid_reduction_p): New function.
8672 (gather_scalar_reductions): Use it.
8673 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8674 (vect_finish_replace_stmt): Declare.
8675 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8676 (needs_fold_left_reduction_p): New function, split out from...
8677 (vect_is_simple_reduction): ...here. Accept reductions that
8678 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8679 (vect_force_simple_reduction): Also store the reduction type in
8680 the assignment's STMT_VINFO_REDUC_TYPE.
8681 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8682 (merge_with_identity): New function.
8683 (vect_expand_fold_left): Likewise.
8684 (vectorize_fold_left_reduction): Likewise.
8685 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8686 scalar phi in place for it. Check for target support and reject
8687 cases that would reassociate the operation. Defer the transform
8688 phase to vectorize_fold_left_reduction.
8689 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8690 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8691 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8693 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8695 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8696 call to ifc_temp_var.
8698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8699 Alan Hayward <alan.hayward@arm.com>
8700 David Sherwood <david.sherwood@arm.com>
8702 * target.def (legitimize_address_displacement): Take the original
8703 offset as a poly_int.
8704 * targhooks.h (default_legitimize_address_displacement): Update
8706 * targhooks.c (default_legitimize_address_displacement): Likewise.
8707 * doc/tm.texi: Regenerate.
8708 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8709 as an argument, moving assert of ad->disp == ad->disp_term to...
8710 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8711 Try calling targetm.legitimize_address_displacement before expanding
8712 the address rather than afterwards, and adjust for the new interface.
8713 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8714 Match the new hook interface. Handle SVE addresses.
8715 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8720 * Makefile.in (OBJS): Add early-remat.o.
8721 * target.def (select_early_remat_modes): New hook.
8722 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8723 * doc/tm.texi: Regenerate.
8724 * targhooks.h (default_select_early_remat_modes): Declare.
8725 * targhooks.c (default_select_early_remat_modes): New function.
8726 * timevar.def (TV_EARLY_REMAT): New timevar.
8727 * passes.def (pass_early_remat): New pass.
8728 * tree-pass.h (make_pass_early_remat): Declare.
8729 * early-remat.c: New file.
8730 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8732 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8734 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8735 Alan Hayward <alan.hayward@arm.com>
8736 David Sherwood <david.sherwood@arm.com>
8738 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8739 vfm1 with a bound_epilog parameter.
8740 (vect_do_peeling): Update calls accordingly, and move the prologue
8741 call earlier in the function. Treat the base bound_epilog as 0 for
8742 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8743 this base when peeling for gaps.
8744 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8745 with fully-masked loops.
8746 (vect_estimate_min_profitable_iters): Handle the single peeled
8747 iteration in that case.
8749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8750 Alan Hayward <alan.hayward@arm.com>
8751 David Sherwood <david.sherwood@arm.com>
8753 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8754 single-element interleaving even if the size is not a power of 2.
8755 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8756 accesses for single-element interleaving if the group size is
8759 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8760 Alan Hayward <alan.hayward@arm.com>
8761 David Sherwood <david.sherwood@arm.com>
8763 * doc/md.texi (fold_extract_last_@var{m}): Document.
8764 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8765 * optabs.def (fold_extract_last_optab): New optab.
8766 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8767 * internal-fn.c (fold_extract_direct): New macro.
8768 (expand_fold_extract_optab_fn): Likewise.
8769 (direct_fold_extract_optab_supported_p): Likewise.
8770 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8771 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8772 EXTRACT_LAST_REDUCTION.
8773 (get_initial_def_for_reduction): Do not create an initial vector
8774 for EXTRACT_LAST_REDUCTION reductions.
8775 (vectorizable_reduction): Leave the scalar phi in place for
8776 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8777 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8778 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8779 transform phase to vectorizable_condition.
8780 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8782 (vect_finish_stmt_generation): ...here.
8783 (vect_finish_replace_stmt): New function.
8784 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8785 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8787 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8790 Alan Hayward <alan.hayward@arm.com>
8791 David Sherwood <david.sherwood@arm.com>
8793 * doc/md.texi (extract_last_@var{m}): Document.
8794 * optabs.def (extract_last_optab): New optab.
8795 * internal-fn.def (EXTRACT_LAST): New internal function.
8796 * internal-fn.c (cond_unary_direct): New macro.
8797 (expand_cond_unary_optab_fn): Likewise.
8798 (direct_cond_unary_optab_supported_p): Likewise.
8799 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8800 loops using EXTRACT_LAST.
8801 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8802 (extract_last_<mode>): ...this optab.
8803 (vec_extract<mode><Vel>): Update accordingly.
8805 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8806 Alan Hayward <alan.hayward@arm.com>
8807 David Sherwood <david.sherwood@arm.com>
8809 * target.def (empty_mask_is_expensive): New hook.
8810 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8811 * doc/tm.texi: Regenerate.
8812 * targhooks.h (default_empty_mask_is_expensive): Declare.
8813 * targhooks.c (default_empty_mask_is_expensive): New function.
8814 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8815 if the target says that empty masks are expensive.
8816 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8818 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8821 Alan Hayward <alan.hayward@arm.com>
8822 David Sherwood <david.sherwood@arm.com>
8824 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8825 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8826 (vect_use_loop_mask_for_alignment_p): New function.
8827 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8828 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8829 niters_skip argument. Make sure that the first niters_skip elements
8830 of the first iteration are inactive.
8831 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8832 Update call to vect_set_loop_masks_directly.
8833 (get_misalign_in_elems): New function, split out from...
8834 (vect_gen_prolog_loop_niters): ...here.
8835 (vect_update_init_of_dr): Take a code argument that specifies whether
8836 the adjustment should be added or subtracted.
8837 (vect_update_init_of_drs): Likewise.
8838 (vect_prepare_for_masked_peels): New function.
8839 (vect_do_peeling): Skip prologue peeling if we're using a mask
8840 instead. Update call to vect_update_inits_of_drs.
8841 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8843 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8844 alignment. Do not include the number of peeled iterations in
8845 the minimum threshold in that case.
8846 (vectorizable_induction): Adjust the start value down by
8847 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8848 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8849 Take the number of skipped iterations into account when calculating
8851 * tree-vect-stmts.c (vect_gen_while_not): New function.
8853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8854 Alan Hayward <alan.hayward@arm.com>
8855 David Sherwood <david.sherwood@arm.com>
8857 * doc/sourcebuild.texi (vect_fully_masked): Document.
8858 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8860 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8862 (vect_analyze_loop_2): ...here. Don't check the vectorization
8863 factor against the number of loop iterations if the loop is
8866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8867 Alan Hayward <alan.hayward@arm.com>
8868 David Sherwood <david.sherwood@arm.com>
8870 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8871 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8872 (dump_groups): Update accordingly.
8873 (iv_use::mem_type): New member variable.
8874 (address_p): New function.
8875 (record_use): Add a mem_type argument and initialize the new
8877 (record_group_use): Add a mem_type argument. Use address_p.
8878 Remove obsolete null checks of base_object. Update call to record_use.
8879 (find_interesting_uses_op): Update call to record_group_use.
8880 (find_interesting_uses_cond): Likewise.
8881 (find_interesting_uses_address): Likewise.
8882 (get_mem_type_for_internal_fn): New function.
8883 (find_address_like_use): Likewise.
8884 (find_interesting_uses_stmt): Try find_address_like_use before
8885 calling find_interesting_uses_op.
8886 (addr_offset_valid_p): Use the iv mem_type field as the type
8887 of the addressed memory.
8888 (add_autoinc_candidates): Likewise.
8889 (get_address_cost): Likewise.
8890 (split_small_address_groups_p): Use address_p.
8891 (split_address_groups): Likewise.
8892 (add_iv_candidate_for_use): Likewise.
8893 (autoinc_possible_for_pair): Likewise.
8894 (rewrite_groups): Likewise.
8895 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8896 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8897 (get_alias_ptr_type_for_ptr_address): New function.
8898 (rewrite_use_address): Rewrite address uses in calls that were
8899 identified by find_address_like_use.
8901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8902 Alan Hayward <alan.hayward@arm.com>
8903 David Sherwood <david.sherwood@arm.com>
8905 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8907 * gimple-expr.h (is_gimple_addressable: Likewise.
8908 * gimple-expr.c (is_gimple_address): Likewise.
8909 * internal-fn.c (expand_call_mem_ref): New function.
8910 (expand_mask_load_optab_fn): Use it.
8911 (expand_mask_store_optab_fn): Likewise.
8913 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8914 Alan Hayward <alan.hayward@arm.com>
8915 David Sherwood <david.sherwood@arm.com>
8917 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8918 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8919 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8920 (cond_umax@var{mode}): Document.
8921 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8922 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8923 (cond_umin_optab, cond_umax_optab): New optabs.
8924 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8925 (COND_IOR, COND_XOR): New internal functions.
8926 * internal-fn.h (get_conditional_internal_fn): Declare.
8927 * internal-fn.c (cond_binary_direct): New macro.
8928 (expand_cond_binary_optab_fn): Likewise.
8929 (direct_cond_binary_optab_supported_p): Likewise.
8930 (get_conditional_internal_fn): New function.
8931 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8932 Cope with reduction statements that are vectorized as calls rather
8934 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8935 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8936 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8937 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8938 (UNSPEC_COND_EOR): New unspecs.
8939 (optab): Add mappings for them.
8940 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8941 (sve_int_op, sve_fp_op): New int attributes.
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 (while_ult_optab): New optab.
8948 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8949 * internal-fn.def (WHILE_ULT): New internal function.
8950 * internal-fn.h (direct_internal_fn_supported_p): New override
8951 that takes two types as argument.
8952 * internal-fn.c (while_direct): New macro.
8953 (expand_while_optab_fn): New function.
8954 (convert_optab_supported_p): Likewise.
8955 (direct_while_optab_supported_p): New macro.
8956 * wide-int.h (wi::udiv_ceil): New function.
8957 * tree-vectorizer.h (rgroup_masks): New structure.
8958 (vec_loop_masks): New typedef.
8959 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8961 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8962 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8963 (vect_max_vf): New function.
8964 (slpeel_make_loop_iterate_ntimes): Delete.
8965 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8966 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8967 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8968 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8969 internal-fn.h, stor-layout.h and optabs-query.h.
8970 (vect_set_loop_mask): New function.
8971 (add_preheader_seq): Likewise.
8972 (add_header_seq): Likewise.
8973 (interleave_supported_p): Likewise.
8974 (vect_maybe_permute_loop_masks): Likewise.
8975 (vect_set_loop_masks_directly): Likewise.
8976 (vect_set_loop_condition_masked): Likewise.
8977 (vect_set_loop_condition_unmasked): New function, split out from
8978 slpeel_make_loop_iterate_ntimes.
8979 (slpeel_make_loop_iterate_ntimes): Rename to..
8980 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8981 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8982 (vect_do_peeling): Update call accordingly.
8983 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8985 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8986 mask_compare_type, can_fully_mask_p and fully_masked_p.
8987 (release_vec_loop_masks): New function.
8988 (_loop_vec_info): Use it to free the loop masks.
8989 (can_produce_all_loop_masks_p): New function.
8990 (vect_get_max_nscalars_per_iter): Likewise.
8991 (vect_verify_full_masking): Likewise.
8992 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8993 retries, and free the mask rgroups before retrying. Check loop-wide
8994 reasons for disallowing fully-masked loops. Make the final decision
8995 about whether use a fully-masked loop or not.
8996 (vect_estimate_min_profitable_iters): Do not assume that peeling
8997 for the number of iterations will be needed for fully-masked loops.
8998 (vectorizable_reduction): Disable fully-masked loops.
8999 (vectorizable_live_operation): Likewise.
9000 (vect_halve_mask_nunits): New function.
9001 (vect_double_mask_nunits): Likewise.
9002 (vect_record_loop_mask): Likewise.
9003 (vect_get_loop_mask): Likewise.
9004 (vect_transform_loop): Handle the case in which the final loop
9005 iteration might handle a partial vector. Call vect_set_loop_condition
9006 instead of slpeel_make_loop_iterate_ntimes.
9007 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9008 (check_load_store_masking): New function.
9009 (prepare_load_store_mask): Likewise.
9010 (vectorizable_store): Handle fully-masked loops.
9011 (vectorizable_load): Likewise.
9012 (supportable_widening_operation): Use vect_halve_mask_nunits for
9014 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9015 (vect_gen_while): New function.
9016 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9017 (aarch64_uqdec<mode>): New insn.
9019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9020 Alan Hayward <alan.hayward@arm.com>
9021 David Sherwood <david.sherwood@arm.com>
9023 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9024 (reduc_xor_scal_optab): New optabs.
9025 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9026 (reduc_xor_scal_@var{m}): Document.
9027 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9028 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9030 * fold-const-call.c (fold_const_call): Handle them.
9031 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9032 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9033 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9034 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9035 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9036 (UNSPEC_XORV): New unspecs.
9037 (optab): Add entries for them.
9038 (BITWISEV): New int iterator.
9039 (bit_reduc_op): New int attributes.
9041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9042 Alan Hayward <alan.hayward@arm.com>
9043 David Sherwood <david.sherwood@arm.com>
9045 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9046 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9047 * optabs.def (vec_shl_insert_optab): New optab.
9048 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9049 (duplicate_and_interleave): Likewise.
9050 * tree-vect-loop.c: Include internal-fn.h.
9051 (neutral_op_for_slp_reduction): New function, split out from
9052 get_initial_defs_for_reduction.
9053 (get_initial_def_for_reduction): Handle option 2 for variable-length
9054 vectors by loading the neutral value into a vector and then shifting
9055 the initial value into element 0.
9056 (get_initial_defs_for_reduction): Replace the code argument with
9057 the neutral value calculated by neutral_op_for_slp_reduction.
9058 Use gimple_build_vector for constant-length vectors.
9059 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9060 but the first group_size elements have a neutral value.
9061 Use duplicate_and_interleave otherwise.
9062 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9063 Update call to get_initial_defs_for_reduction. Handle SLP
9064 reductions for variable-length vectors by creating one vector
9065 result for each scalar result, with the elements associated
9066 with other scalar results stubbed out with the neutral value.
9067 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9068 Require IFN_VEC_SHL_INSERT for double reductions on
9069 variable-length vectors, or SLP reductions that have
9070 a neutral value. Require can_duplicate_and_interleave_p
9071 support for variable-length unchained SLP reductions if there
9072 is no neutral value, such as for MIN/MAX reductions. Also require
9073 the number of vector elements to be a multiple of the number of
9074 SLP statements when doing variable-length unchained SLP reductions.
9075 Update call to vect_create_epilog_for_reduction.
9076 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9077 and remove initial values.
9078 (duplicate_and_interleave): Make public.
9079 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9080 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9082 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9083 Alan Hayward <alan.hayward@arm.com>
9084 David Sherwood <david.sherwood@arm.com>
9086 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9087 (can_duplicate_and_interleave_p): New function.
9088 (vect_get_and_check_slp_defs): Take the vector of statements
9089 rather than just the current one. Remove excess parentheses.
9090 Restriction rejectinon of vect_constant_def and vect_external_def
9091 for variable-length vectors to boolean types, or types for which
9092 can_duplicate_and_interleave_p is false.
9093 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9094 (duplicate_and_interleave): New function.
9095 (vect_get_constant_vectors): Use gimple_build_vector for
9096 constant-length vectors and suitable variable-length constant
9097 vectors. Use duplicate_and_interleave for other variable-length
9098 vectors. Don't defer the update when inserting new statements.
9100 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9101 Alan Hayward <alan.hayward@arm.com>
9102 David Sherwood <david.sherwood@arm.com>
9104 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9105 min_profitable_iters doesn't go negative.
9107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9108 Alan Hayward <alan.hayward@arm.com>
9109 David Sherwood <david.sherwood@arm.com>
9111 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9112 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9113 * optabs.def (vec_mask_load_lanes_optab): New optab.
9114 (vec_mask_store_lanes_optab): Likewise.
9115 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9116 (MASK_STORE_LANES): Likewise.
9117 * internal-fn.c (mask_load_lanes_direct): New macro.
9118 (mask_store_lanes_direct): Likewise.
9119 (expand_mask_load_optab_fn): Handle masked operations.
9120 (expand_mask_load_lanes_optab_fn): New macro.
9121 (expand_mask_store_optab_fn): Handle masked operations.
9122 (expand_mask_store_lanes_optab_fn): New macro.
9123 (direct_mask_load_lanes_optab_supported_p): Likewise.
9124 (direct_mask_store_lanes_optab_supported_p): Likewise.
9125 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9127 (vect_load_lanes_supported): Likewise.
9128 * tree-vect-data-refs.c (strip_conversion): New function.
9129 (can_group_stmts_p): Likewise.
9130 (vect_analyze_data_ref_accesses): Use it instead of checking
9131 for a pair of assignments.
9132 (vect_store_lanes_supported): Take a masked_p parameter.
9133 (vect_load_lanes_supported): Likewise.
9134 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9135 vect_store_lanes_supported and vect_load_lanes_supported.
9136 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9137 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9138 parameter. Don't allow gaps for masked accesses.
9139 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9140 and vect_load_lanes_supported.
9141 (get_load_store_type): Take a masked_p parameter and update
9142 call to get_group_load_store_type.
9143 (vectorizable_store): Update call to get_load_store_type.
9144 Handle IFN_MASK_STORE_LANES.
9145 (vectorizable_load): Update call to get_load_store_type.
9146 Handle IFN_MASK_LOAD_LANES.
9148 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9149 Alan Hayward <alan.hayward@arm.com>
9150 David Sherwood <david.sherwood@arm.com>
9152 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9154 * config/aarch64/aarch64-protos.h
9155 (aarch64_sve_struct_memory_operand_p): Declare.
9156 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9157 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9158 (VPRED, vpred): Handle SVE structure modes.
9159 * config/aarch64/constraints.md (Utx): New constraint.
9160 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9161 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9162 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9163 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9164 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9165 structure modes. Split into pieces after RA.
9166 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9167 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9169 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9170 SVE structure modes.
9171 (aarch64_classify_address): Likewise.
9172 (sizetochar): Move earlier in file.
9173 (aarch64_print_operand): Handle SVE register lists.
9174 (aarch64_array_mode): New function.
9175 (aarch64_sve_struct_memory_operand_p): Likewise.
9176 (TARGET_ARRAY_MODE): Redefine.
9178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9179 Alan Hayward <alan.hayward@arm.com>
9180 David Sherwood <david.sherwood@arm.com>
9182 * target.def (array_mode): New target hook.
9183 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9184 * doc/tm.texi: Regenerate.
9185 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9186 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9187 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9189 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9192 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9193 Alan Hayward <alan.hayward@arm.com>
9194 David Sherwood <david.sherwood@arm.com>
9196 * fold-const.c (fold_binary_loc): Check the argument types
9197 rather than the result type when testing for a vector operation.
9199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9201 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9202 * doc/tm.texi: Regenerate.
9204 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9205 Alan Hayward <alan.hayward@arm.com>
9206 David Sherwood <david.sherwood@arm.com>
9208 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9209 (sve): Document new AArch64 extension.
9210 * doc/md.texi (w): Extend the description of the AArch64
9211 constraint to include SVE vectors.
9212 (Upl, Upa): Document new AArch64 predicate constraints.
9213 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9215 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9216 (msve-vector-bits=): New option.
9217 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9218 SVE when these are disabled.
9219 (sve): New extension.
9220 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9221 modes. Adjust their number of units based on aarch64_sve_vg.
9222 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9223 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9224 aarch64_addr_query_type.
9225 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9226 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9227 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9228 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9229 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9230 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9231 (aarch64_simd_imm_zero_p): Delete.
9232 (aarch64_check_zero_based_sve_index_immediate): Declare.
9233 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9234 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9235 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9236 (aarch64_sve_float_mul_immediate_p): Likewise.
9237 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9239 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9240 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9241 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9242 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9243 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9244 (aarch64_regmode_natural_size): Likewise.
9245 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9246 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9248 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9249 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9250 for VG and the SVE predicate registers.
9251 (V_ALIASES): Add a "z"-prefixed alias.
9252 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9253 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9254 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9255 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9256 (REG_CLASS_NAMES): Add entries for them.
9257 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9258 and the predicate registers.
9259 (aarch64_sve_vg): Declare.
9260 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9261 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9262 (REGMODE_NATURAL_SIZE): Define.
9263 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9265 * config/aarch64/aarch64.c: Include cfgrtl.h.
9266 (simd_immediate_info): Add a constructor for series vectors,
9267 and an associated step field.
9268 (aarch64_sve_vg): New variable.
9269 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9270 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9271 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9272 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9273 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9274 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9275 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9276 (aarch64_get_mask_mode): New functions.
9277 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9278 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9279 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9280 predicate modes and predicate registers. Explicitly restrict
9281 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9282 to store a vector mode if it is recognized by
9283 aarch64_classify_vector_mode.
9284 (aarch64_regmode_natural_size): New function.
9285 (aarch64_hard_regno_caller_save_mode): Return the original mode
9287 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9288 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9289 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9290 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9292 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9293 does not overlap dest if the function is frame-related. Handle
9295 (aarch64_split_add_offset): New function.
9296 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9297 them aarch64_add_offset.
9298 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9299 and update call to aarch64_sub_sp.
9300 (aarch64_add_cfa_expression): New function.
9301 (aarch64_expand_prologue): Pass extra temporary registers to the
9302 functions above. Handle the case in which we need to emit new
9303 DW_CFA_expressions for registers that were originally saved
9304 relative to the stack pointer, but now have to be expressed
9305 relative to the frame pointer.
9306 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9308 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9309 IP0 and IP1 values for SVE frames.
9310 (aarch64_expand_vec_series): New function.
9311 (aarch64_expand_sve_widened_duplicate): Likewise.
9312 (aarch64_expand_sve_const_vector): Likewise.
9313 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9314 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9315 into the register, rather than emitting a SET directly.
9316 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9317 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9318 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9319 (offset_9bit_signed_scaled_p): New functions.
9320 (aarch64_replicate_bitmask_imm): New function.
9321 (aarch64_bitmask_imm): Use it.
9322 (aarch64_cannot_force_const_mem): Reject expressions involving
9323 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9324 (aarch64_classify_index): Handle SVE indices, by requiring
9325 a plain register index with a scale that matches the element size.
9326 (aarch64_classify_address): Handle SVE addresses. Assert that
9327 the mode of the address is VOIDmode or an integer mode.
9328 Update call to aarch64_classify_symbol.
9329 (aarch64_classify_symbolic_expression): Update call to
9330 aarch64_classify_symbol.
9331 (aarch64_const_vec_all_in_range_p): New function.
9332 (aarch64_print_vector_float_operand): Likewise.
9333 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9334 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9335 and the FP immediates 1.0 and 0.5.
9336 (aarch64_print_address_internal): Handle SVE addresses.
9337 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9338 (aarch64_regno_regclass): Handle predicate registers.
9339 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9341 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9342 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9343 (aarch64_convert_sve_vector_bits): New function.
9344 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9345 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9347 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9348 Handle SVE vector and predicate modes. Accept VL-based constants
9349 that need only one temporary register, and VL offsets that require
9350 no temporary registers.
9351 (aarch64_conditional_register_usage): Mark the predicate registers
9352 as fixed if SVE isn't available.
9353 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9354 Return true for SVE vector and predicate modes.
9355 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9356 rather than an unsigned int. Handle SVE modes.
9357 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9359 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9361 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9362 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9363 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9364 (aarch64_sve_float_mul_immediate_p): New functions.
9365 (aarch64_sve_valid_immediate): New function.
9366 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9367 Explicitly reject structure modes. Check for INDEX constants.
9368 Handle PTRUE and PFALSE constants.
9369 (aarch64_check_zero_based_sve_index_immediate): New function.
9370 (aarch64_simd_imm_zero_p): Delete.
9371 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9372 vector modes. Accept constants in the range of CNT[BHWD].
9373 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9374 ask for an Advanced SIMD mode.
9375 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9376 (aarch64_simd_vector_alignment): Handle SVE predicates.
9377 (aarch64_vectorize_preferred_vector_alignment): New function.
9378 (aarch64_simd_vector_alignment_reachable): Use it instead of
9380 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9381 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9383 (MAX_VECT_LEN): Delete.
9384 (expand_vec_perm_d): Add a vec_flags field.
9385 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9386 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9387 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9389 (aarch64_evpc_rev): Rename to...
9390 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9391 (aarch64_evpc_rev_global): New function.
9392 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9393 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9395 (aarch64_evpc_sve_tbl): New function.
9396 (aarch64_expand_vec_perm_const_1): Update after rename of
9397 aarch64_evpc_rev. Handle SVE permutes too, trying
9398 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9399 than aarch64_evpc_tbl.
9400 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9401 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9402 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9403 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9404 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9405 (aarch64_expand_sve_vcond): New functions.
9406 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9407 of aarch64_vector_mode_p.
9408 (aarch64_dwarf_poly_indeterminate_value): New function.
9409 (aarch64_compute_pressure_classes): Likewise.
9410 (aarch64_can_change_mode_class): Likewise.
9411 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9412 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9413 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9414 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9415 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9416 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9417 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9418 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9420 (Dn, Dl, Dr): Accept const as well as const_vector.
9421 (Dz): Likewise. Compare against CONST0_RTX.
9422 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9423 of "vector" where appropriate.
9424 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9425 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9426 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9427 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9428 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9429 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9430 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9431 (v_int_equiv): Extend to SVE modes.
9432 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9434 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9435 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9436 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9437 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9438 (SVE_COND_FP_CMP): New int iterators.
9439 (perm_hilo): Handle the new unpack unspecs.
9440 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9442 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9443 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9444 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9445 (aarch64_equality_operator, aarch64_constant_vector_operand)
9446 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9447 (aarch64_sve_nonimmediate_operand): Likewise.
9448 (aarch64_sve_general_operand): Likewise.
9449 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9450 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9451 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9452 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9453 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9454 (aarch64_sve_float_arith_immediate): Likewise.
9455 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9456 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9457 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9458 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9459 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9460 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9461 (aarch64_sve_float_arith_operand): Likewise.
9462 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9463 (aarch64_sve_float_mul_operand): Likewise.
9464 (aarch64_sve_vec_perm_operand): Likewise.
9465 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9466 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9467 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9468 as well as const_vector.
9469 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9470 in file. Use CONST0_RTX and CONSTM1_RTX.
9471 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9472 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9473 Use aarch64_simd_imm_zero.
9474 * config/aarch64/aarch64-sve.md: New file.
9475 * config/aarch64/aarch64.md: Include it.
9476 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9477 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9478 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9479 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9480 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9481 (sve): New attribute.
9482 (enabled): Disable instructions with the sve attribute unless
9484 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9485 aarch64_expand_mov_immediate.
9486 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9487 CNT[BHSD] immediates.
9488 (movti): Split CONST_POLY_INT moves into two halves.
9489 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9490 Split additions that need a temporary here if the destination
9491 is the stack pointer.
9492 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9493 (*add<mode>3_poly_1): New instruction.
9494 (set_clobber_cc): New expander.
9496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9498 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9499 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9500 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9501 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9502 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9503 Change innermode from fixed_mode_size to machine_mode.
9504 (simplify_subreg): Update call accordingly. Handle a constant-sized
9505 subreg of a variable-length CONST_VECTOR.
9507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9508 Alan Hayward <alan.hayward@arm.com>
9509 David Sherwood <david.sherwood@arm.com>
9511 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9512 (add_offset_to_base): New function, split out from...
9513 (create_mem_ref): ...here. When handling a scale other than 1,
9514 check first whether the address is valid without the offset.
9515 Add it into the base if so, leaving the index and scale as-is.
9517 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9520 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9521 fold_for_warn before checking if arg2 is INTEGER_CST.
9523 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9525 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9526 (store_multiple_operation): Delete.
9527 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9528 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9529 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9530 guarded by TARGET_STRING.
9531 (rs6000_output_load_multiple): Delete.
9532 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9533 OPTION_MASK_STRING / TARGET_STRING handling.
9534 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9535 (const rs6000_opt_masks) <"string">: Change mask to 0.
9536 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9537 (MASK_STRING): Delete.
9538 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9540 (load_multiple): Delete.
9547 (store_multiple): Delete.
9554 (movmemsi_8reg): Delete.
9555 (corresponding unnamed define_insn): Delete.
9556 (movmemsi_6reg): Delete.
9557 (corresponding unnamed define_insn): Delete.
9558 (movmemsi_4reg): Delete.
9559 (corresponding unnamed define_insn): Delete.
9560 (movmemsi_2reg): Delete.
9561 (corresponding unnamed define_insn): Delete.
9562 (movmemsi_1reg): Delete.
9563 (corresponding unnamed define_insn): Delete.
9564 * config/rs6000/rs6000.opt (mno-string): New.
9565 (mstring): Replace by deprecation warning stub.
9566 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9568 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9570 * regrename.c (regrename_do_replace): If replacing the same
9571 reg multiple times, try to reuse last created gen_raw_REG.
9574 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9575 main to workaround a bug in GDB.
9577 2018-01-12 Tom de Vries <tom@codesourcery.com>
9580 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9582 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9584 PR rtl-optimization/80481
9585 * ira-color.c (get_cap_member): New function.
9586 (allocnos_conflict_by_live_ranges_p): Use it.
9587 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9588 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9590 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9593 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9594 (*saddl_se_1): Ditto.
9596 (*ssubl_se_1): Ditto.
9598 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9600 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9601 rather than wi::to_widest for DR_INITs.
9602 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9603 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9604 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9606 (vect_analyze_group_access_1): Note that here.
9608 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9610 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9611 polynomial type sizes.
9613 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9615 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9616 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9617 (gimple_add_tmp_var): Likewise.
9619 2018-01-12 Martin Liska <mliska@suse.cz>
9621 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9622 (gimple_alloc_sizes): Likewise.
9623 (dump_gimple_statistics): Use PRIu64 in printf format.
9624 * gimple.h: Change uint64_t to int.
9626 2018-01-12 Martin Liska <mliska@suse.cz>
9628 * tree-core.h: Use uint64_t instead of int.
9629 * tree.c (tree_node_counts): Likewise.
9630 (tree_node_sizes): Likewise.
9631 (dump_tree_statistics): Use PRIu64 in printf format.
9633 2018-01-12 Martin Liska <mliska@suse.cz>
9635 * Makefile.in: As qsort_chk is implemented in vec.c, add
9636 vec.o to linkage of gencfn-macros.
9637 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9638 passing the info to record_node_allocation_statistics.
9639 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9641 * ggc-common.c (struct ggc_usage): Add operator== and use
9642 it in operator< and compare function.
9643 * mem-stats.h (struct mem_usage): Likewise.
9644 * vec.c (struct vec_usage): Remove operator< and compare
9645 function. Can be simply inherited.
9647 2018-01-12 Martin Jambor <mjambor@suse.cz>
9650 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9651 * tree-ssa-math-opts.c: Include domwalk.h.
9652 (convert_mult_to_fma_1): New function.
9653 (fma_transformation_info): New type.
9654 (fma_deferring_state): Likewise.
9655 (cancel_fma_deferring): New function.
9656 (result_of_phi): Likewise.
9657 (last_fma_candidate_feeds_initial_phi): Likewise.
9658 (convert_mult_to_fma): Added deferring logic, split actual
9659 transformation to convert_mult_to_fma_1.
9660 (math_opts_dom_walker): New type.
9661 (math_opts_dom_walker::after_dom_children): New method, body moved
9662 here from pass_optimize_widening_mul::execute, added deferring logic
9664 (pass_optimize_widening_mul::execute): Moved most of code to
9665 math_opts_dom_walker::after_dom_children.
9666 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9667 * config/i386/i386.c (ix86_option_override_internal): Added
9668 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9670 2018-01-12 Richard Biener <rguenther@suse.de>
9673 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9674 inline instance vars.
9676 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9679 * config/rx/rx.c (rx_is_restricted_memory_address):
9682 2018-01-12 Richard Biener <rguenther@suse.de>
9684 PR tree-optimization/80846
9685 * target.def (split_reduction): New target hook.
9686 * targhooks.c (default_split_reduction): New function.
9687 * targhooks.h (default_split_reduction): Declare.
9688 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9689 target requests first reduce vectors by combining low and high
9691 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9692 (get_vectype_for_scalar_type_and_size): Export.
9693 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9694 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9695 * doc/tm.texi: Regenerate.
9696 * config/i386/i386.c (ix86_split_reduction): Implement
9697 TARGET_VECTORIZE_SPLIT_REDUCTION.
9699 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9702 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9703 in PIC mode except for TARGET_VXWORKS_RTP.
9704 * config/sparc/sparc.c: Include cfgrtl.h.
9705 (TARGET_INIT_PIC_REG): Define.
9706 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9707 (sparc_pic_register_p): New predicate.
9708 (sparc_legitimate_address_p): Use it.
9709 (sparc_legitimize_pic_address): Likewise.
9710 (sparc_delegitimize_address): Likewise.
9711 (sparc_mode_dependent_address_p): Likewise.
9712 (gen_load_pcrel_sym): Remove 4th parameter.
9713 (load_got_register): Adjust call to above. Remove obsolete stuff.
9714 (sparc_expand_prologue): Do not call load_got_register here.
9715 (sparc_flat_expand_prologue): Likewise.
9716 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9717 (sparc_use_pseudo_pic_reg): New function.
9718 (sparc_init_pic_reg): Likewise.
9719 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9720 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9722 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9724 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9725 Add item for branch_cost.
9727 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9729 PR rtl-optimization/83565
9730 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9731 not extend the result to a larger mode for rotate operations.
9732 (num_sign_bit_copies1): Likewise.
9734 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9737 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9739 Use values-Xc.o for -pedantic.
9740 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9742 2018-01-12 Martin Liska <mliska@suse.cz>
9745 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9747 (possible_polymorphic_call_targets): Use it.
9748 (ipa_devirt): Likewise.
9750 2018-01-12 Martin Liska <mliska@suse.cz>
9752 * profile-count.h (enum profile_quality): Use 0 as invalid
9753 enum value of profile_quality.
9755 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9757 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9758 -mext-string options.
9760 2018-01-12 Richard Biener <rguenther@suse.de>
9762 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9763 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9764 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9766 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9768 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9770 * configure.ac (--with-long-double-format): Add support for the
9771 configuration option to change the default long double format on
9773 * config.gcc (powerpc*-linux*-*): Likewise.
9774 * configure: Regenerate.
9775 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9776 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9777 used without modification.
9779 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9781 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9782 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9783 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9784 MISC_BUILTIN_SPEC_BARRIER.
9785 (rs6000_init_builtins): Likewise.
9786 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9788 (speculation_barrier): New define_insn.
9789 * doc/extend.texi: Document __builtin_speculation_barrier.
9791 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9794 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9795 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9796 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9798 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9799 integral modes instead of "ss" and "sd".
9800 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9801 vectors with 32-bit and 64-bit elements.
9802 (vecdupssescalarmodesuffix): New mode attribute.
9803 (vec_dup<mode>): Use it.
9805 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9808 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9809 frame if argument is passed on stack.
9811 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9814 * ree.c (combine_reaching_defs): Optimize also
9815 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9816 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9818 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9821 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9823 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9826 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9827 after they are computed.
9829 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9831 PR tree-optimization/83695
9832 * gimple-loop-linterchange.cc
9833 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9834 reset cached scev information after interchange.
9835 (pass_linterchange::execute): Remove call to scev_reset_htab.
9837 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9839 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9840 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9841 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9842 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9843 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9844 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9845 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9846 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9847 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9848 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9849 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9850 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9851 (V_lane_reg): Likewise.
9852 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9854 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9855 (vfmal_lane_low<mode>_intrinsic,
9856 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9857 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9858 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9859 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9860 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9861 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9863 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9865 * config/arm/arm-cpus.in (fp16fml): New feature.
9866 (ALL_SIMD): Add fp16fml.
9867 (armv8.2-a): Add fp16fml as an option.
9868 (armv8.3-a): Likewise.
9869 (armv8.4-a): Add fp16fml as part of fp16.
9870 * config/arm/arm.h (TARGET_FP16FML): Define.
9871 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9873 * config/arm/arm-modes.def (V2HF): Define.
9874 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9875 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9876 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9877 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9878 vfmsl_low, vfmsl_high): New set of builtins.
9879 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9880 (vfml_op): New code attribute.
9881 (VFMLHALVES): New int iterator.
9882 (VFML, VFMLSEL): New mode attributes.
9883 (V_reg): Define mapping for V2HF.
9884 (V_hi, V_lo): New mode attributes.
9885 (VF_constraint): Likewise.
9886 (vfml_half, vfml_half_selector): New int attributes.
9887 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9889 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9890 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9892 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9893 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9894 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9895 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9897 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9898 Document new effective target and option set.
9900 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9902 * config/arm/arm-cpus.in (armv8_4): New feature.
9903 (ARMv8_4a): New fgroup.
9904 (armv8.4-a): New arch.
9905 * config/arm/arm-tables.opt: Regenerate.
9906 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9907 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9908 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9909 Add matching rules for -march=armv8.4-a and extensions.
9910 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9912 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9915 * config/rx/rx.md (BW): New mode attribute.
9916 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9918 2018-01-11 Richard Biener <rguenther@suse.de>
9920 PR tree-optimization/83435
9921 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9922 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9923 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9925 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9926 Alan Hayward <alan.hayward@arm.com>
9927 David Sherwood <david.sherwood@arm.com>
9929 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9931 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9932 (aarch64_print_address_internal): Use it to check for a zero offset.
9934 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9935 Alan Hayward <alan.hayward@arm.com>
9936 David Sherwood <david.sherwood@arm.com>
9938 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9939 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9940 Return a poly_int64 rather than a HOST_WIDE_INT.
9941 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9942 rather than a HOST_WIDE_INT.
9943 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9944 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9945 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9946 final_offset from HOST_WIDE_INT to poly_int64.
9947 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9948 to_constant when getting the number of units in an Advanced SIMD
9950 (aarch64_builtin_vectorized_function): Check for a constant number
9952 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9954 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9955 attribute instead of GET_MODE_NUNITS.
9956 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9957 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9958 GET_MODE_SIZE for fixed-size registers.
9959 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9960 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9961 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9962 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9963 (aarch64_print_operand, aarch64_print_address_internal)
9964 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9965 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9966 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9967 Handle polynomial GET_MODE_SIZE.
9968 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9969 wider than SImode without modification.
9970 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9971 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9972 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9973 passing and returning SVE modes.
9974 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9975 rather than GEN_INT.
9976 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9977 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9978 (aarch64_allocate_and_probe_stack_space): Likewise.
9979 (aarch64_layout_frame): Cope with polynomial offsets.
9980 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9981 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9983 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9984 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9985 poly_int64 rather than a HOST_WIDE_INT.
9986 (aarch64_get_separate_components, aarch64_process_components)
9987 (aarch64_expand_prologue, aarch64_expand_epilogue)
9988 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9989 (aarch64_anchor_offset): New function, split out from...
9990 (aarch64_legitimize_address): ...here.
9991 (aarch64_builtin_vectorization_cost): Handle polynomial
9992 TYPE_VECTOR_SUBPARTS.
9993 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9995 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9996 number of elements from the PARALLEL rather than the mode.
9997 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9998 rather than GET_MODE_BITSIZE.
9999 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10000 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10001 (aarch64_expand_vec_perm_const_1): Handle polynomial
10002 d->perm.length () and d->perm elements.
10003 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10004 Apply to_constant to d->perm elements.
10005 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10006 polynomial CONST_VECTOR_NUNITS.
10007 (aarch64_move_pointer): Take amount as a poly_int64 rather
10009 (aarch64_progress_pointer): Avoid temporary variable.
10010 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10011 the mode attribute instead of GET_MODE.
10013 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10014 Alan Hayward <alan.hayward@arm.com>
10015 David Sherwood <david.sherwood@arm.com>
10017 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10018 x exists before using it.
10019 (aarch64_add_constant_internal): Rename to...
10020 (aarch64_add_offset_1): ...this. Replace regnum with separate
10021 src and dest rtxes. Handle the case in which they're different,
10022 including when the offset is zero. Replace scratchreg with an rtx.
10023 Use 2 additions if there is no spare register into which we can
10024 move a 16-bit constant.
10025 (aarch64_add_constant): Delete.
10026 (aarch64_add_offset): Replace reg with separate src and dest
10027 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10028 Use aarch64_add_offset_1.
10029 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10030 an rtx rather than an int. Take the delta as a poly_int64
10031 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10032 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10033 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10034 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10035 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10036 and aarch64_add_sp.
10037 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10038 aarch64_add_constant.
10040 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10042 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10043 Use scalar_float_mode.
10045 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10047 * config/aarch64/aarch64-simd.md
10048 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10049 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10050 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10051 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10052 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10053 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10054 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10055 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10056 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10057 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10059 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10062 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10063 targ_options->x_arm_arch_string is non NULL.
10065 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10067 * config/aarch64/aarch64.h
10068 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10070 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10073 * expmed.c (emit_store_flag_force): Swap if const op0
10074 and change VOIDmode to mode of op0.
10076 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10078 PR rtl-optimization/83761
10079 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10080 than bytes to mode_for_size.
10082 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10084 PR middle-end/83189
10085 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10086 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10089 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10091 PR middle-end/83575
10092 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10093 when in layout mode.
10094 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10095 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10098 2018-01-10 Michael Collison <michael.collison@arm.com>
10100 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10101 * config/aarch64/aarch64-option-extension.def: Add
10102 AARCH64_OPT_EXTENSION of 'fp16fml'.
10103 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10104 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10105 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10106 * config/aarch64/constraints.md (Ui7): New constraint.
10107 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10108 (VFMLA_SEL_W): Ditto.
10111 (VFMLA16_LOW): New int iterator.
10112 (VFMLA16_HIGH): Ditto.
10113 (UNSPEC_FMLAL): New unspec.
10114 (UNSPEC_FMLSL): Ditto.
10115 (UNSPEC_FMLAL2): Ditto.
10116 (UNSPEC_FMLSL2): Ditto.
10117 (f16mac): New code attribute.
10118 * config/aarch64/aarch64-simd-builtins.def
10119 (aarch64_fmlal_lowv2sf): Ditto.
10120 (aarch64_fmlsl_lowv2sf): Ditto.
10121 (aarch64_fmlalq_lowv4sf): Ditto.
10122 (aarch64_fmlslq_lowv4sf): Ditto.
10123 (aarch64_fmlal_highv2sf): Ditto.
10124 (aarch64_fmlsl_highv2sf): Ditto.
10125 (aarch64_fmlalq_highv4sf): Ditto.
10126 (aarch64_fmlslq_highv4sf): Ditto.
10127 (aarch64_fmlal_lane_lowv2sf): Ditto.
10128 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10129 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10130 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10131 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10132 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10133 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10134 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10135 (aarch64_fmlal_lane_highv2sf): Ditto.
10136 (aarch64_fmlsl_lane_highv2sf): Ditto.
10137 (aarch64_fmlal_laneq_highv2sf): Ditto.
10138 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10139 (aarch64_fmlalq_lane_highv4sf): Ditto.
10140 (aarch64_fmlsl_lane_highv4sf): Ditto.
10141 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10142 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10143 * config/aarch64/aarch64-simd.md:
10144 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10145 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10146 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10147 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10148 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10149 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10150 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10151 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10152 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10153 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10154 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10155 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10156 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10157 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10158 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10159 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10160 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10161 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10162 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10163 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10164 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10165 (vfmlsl_low_u32): Ditto.
10166 (vfmlalq_low_u32): Ditto.
10167 (vfmlslq_low_u32): Ditto.
10168 (vfmlal_high_u32): Ditto.
10169 (vfmlsl_high_u32): Ditto.
10170 (vfmlalq_high_u32): Ditto.
10171 (vfmlslq_high_u32): Ditto.
10172 (vfmlal_lane_low_u32): Ditto.
10173 (vfmlsl_lane_low_u32): Ditto.
10174 (vfmlal_laneq_low_u32): Ditto.
10175 (vfmlsl_laneq_low_u32): Ditto.
10176 (vfmlalq_lane_low_u32): Ditto.
10177 (vfmlslq_lane_low_u32): Ditto.
10178 (vfmlalq_laneq_low_u32): Ditto.
10179 (vfmlslq_laneq_low_u32): Ditto.
10180 (vfmlal_lane_high_u32): Ditto.
10181 (vfmlsl_lane_high_u32): Ditto.
10182 (vfmlal_laneq_high_u32): Ditto.
10183 (vfmlsl_laneq_high_u32): Ditto.
10184 (vfmlalq_lane_high_u32): Ditto.
10185 (vfmlslq_lane_high_u32): Ditto.
10186 (vfmlalq_laneq_high_u32): Ditto.
10187 (vfmlslq_laneq_high_u32): Ditto.
10188 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10189 (AARCH64_FL_FOR_ARCH8_4): New.
10190 (AARCH64_ISA_F16FML): New ISA flag.
10191 (TARGET_F16FML): New feature flag for fp16fml.
10192 (doc/invoke.texi): Document new fp16fml option.
10194 2018-01-10 Michael Collison <michael.collison@arm.com>
10196 * config/aarch64/aarch64-builtins.c:
10197 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10198 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10199 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10200 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10201 (AARCH64_ISA_SHA3): New ISA flag.
10202 (TARGET_SHA3): New feature flag for sha3.
10203 * config/aarch64/iterators.md (sha512_op): New int attribute.
10204 (CRYPTO_SHA512): New int iterator.
10205 (UNSPEC_SHA512H): New unspec.
10206 (UNSPEC_SHA512H2): Ditto.
10207 (UNSPEC_SHA512SU0): Ditto.
10208 (UNSPEC_SHA512SU1): Ditto.
10209 * config/aarch64/aarch64-simd-builtins.def
10210 (aarch64_crypto_sha512hqv2di): New builtin.
10211 (aarch64_crypto_sha512h2qv2di): Ditto.
10212 (aarch64_crypto_sha512su0qv2di): Ditto.
10213 (aarch64_crypto_sha512su1qv2di): Ditto.
10214 (aarch64_eor3qv8hi): Ditto.
10215 (aarch64_rax1qv2di): Ditto.
10216 (aarch64_xarqv2di): Ditto.
10217 (aarch64_bcaxqv8hi): Ditto.
10218 * config/aarch64/aarch64-simd.md:
10219 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10220 (aarch64_crypto_sha512su0qv2di): Ditto.
10221 (aarch64_crypto_sha512su1qv2di): Ditto.
10222 (aarch64_eor3qv8hi): Ditto.
10223 (aarch64_rax1qv2di): Ditto.
10224 (aarch64_xarqv2di): Ditto.
10225 (aarch64_bcaxqv8hi): Ditto.
10226 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10227 (vsha512h2q_u64): Ditto.
10228 (vsha512su0q_u64): Ditto.
10229 (vsha512su1q_u64): Ditto.
10230 (veor3q_u16): Ditto.
10231 (vrax1q_u64): Ditto.
10232 (vxarq_u64): Ditto.
10233 (vbcaxq_u16): Ditto.
10234 * config/arm/types.md (crypto_sha512): New type attribute.
10235 (crypto_sha3): Ditto.
10236 (doc/invoke.texi): Document new sha3 option.
10238 2018-01-10 Michael Collison <michael.collison@arm.com>
10240 * config/aarch64/aarch64-builtins.c:
10241 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10242 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10243 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10244 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10245 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10246 (AARCH64_ISA_SM4): New ISA flag.
10247 (TARGET_SM4): New feature flag for sm4.
10248 * config/aarch64/aarch64-simd-builtins.def
10249 (aarch64_sm3ss1qv4si): Ditto.
10250 (aarch64_sm3tt1aq4si): Ditto.
10251 (aarch64_sm3tt1bq4si): Ditto.
10252 (aarch64_sm3tt2aq4si): Ditto.
10253 (aarch64_sm3tt2bq4si): Ditto.
10254 (aarch64_sm3partw1qv4si): Ditto.
10255 (aarch64_sm3partw2qv4si): Ditto.
10256 (aarch64_sm4eqv4si): Ditto.
10257 (aarch64_sm4ekeyqv4si): Ditto.
10258 * config/aarch64/aarch64-simd.md:
10259 (aarch64_sm3ss1qv4si): Ditto.
10260 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10261 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10262 (aarch64_sm4eqv4si): Ditto.
10263 (aarch64_sm4ekeyqv4si): Ditto.
10264 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10265 (sm3part_op): Ditto.
10266 (CRYPTO_SM3TT): Ditto.
10267 (CRYPTO_SM3PART): Ditto.
10268 (UNSPEC_SM3SS1): New unspec.
10269 (UNSPEC_SM3TT1A): Ditto.
10270 (UNSPEC_SM3TT1B): Ditto.
10271 (UNSPEC_SM3TT2A): Ditto.
10272 (UNSPEC_SM3TT2B): Ditto.
10273 (UNSPEC_SM3PARTW1): Ditto.
10274 (UNSPEC_SM3PARTW2): Ditto.
10275 (UNSPEC_SM4E): Ditto.
10276 (UNSPEC_SM4EKEY): Ditto.
10277 * config/aarch64/constraints.md (Ui2): New constraint.
10278 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10279 * config/arm/types.md (crypto_sm3): New type attribute.
10280 (crypto_sm4): Ditto.
10281 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10282 (vsm3tt1aq_u32): Ditto.
10283 (vsm3tt1bq_u32): Ditto.
10284 (vsm3tt2aq_u32): Ditto.
10285 (vsm3tt2bq_u32): Ditto.
10286 (vsm3partw1q_u32): Ditto.
10287 (vsm3partw2q_u32): Ditto.
10288 (vsm4eq_u32): Ditto.
10289 (vsm4ekeyq_u32): Ditto.
10290 (doc/invoke.texi): Document new sm4 option.
10292 2018-01-10 Michael Collison <michael.collison@arm.com>
10294 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10295 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10296 (AARCH64_FL_FOR_ARCH8_4): New.
10297 (AARCH64_FL_V8_4): New flag.
10298 (doc/invoke.texi): Document new armv8.4-a option.
10300 2018-01-10 Michael Collison <michael.collison@arm.com>
10302 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10303 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10304 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10305 * config/aarch64/aarch64-option-extension.def: Add
10306 AARCH64_OPT_EXTENSION of 'sha2'.
10307 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10308 (crypto): Disable sha2 and aes if crypto disabled.
10309 (crypto): Enable aes and sha2 if enabled.
10310 (simd): Disable sha2 and aes if simd disabled.
10311 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10313 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10314 (TARGET_SHA2): New feature flag for sha2.
10315 (TARGET_AES): New feature flag for aes.
10316 * config/aarch64/aarch64-simd.md:
10317 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10318 conditional on TARGET_AES.
10319 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10320 (aarch64_crypto_sha1hsi): Make pattern conditional
10322 (aarch64_crypto_sha1hv4si): Ditto.
10323 (aarch64_be_crypto_sha1hv4si): Ditto.
10324 (aarch64_crypto_sha1su1v4si): Ditto.
10325 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10326 (aarch64_crypto_sha1su0v4si): Ditto.
10327 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10328 (aarch64_crypto_sha256su0v4si): Ditto.
10329 (aarch64_crypto_sha256su1v4si): Ditto.
10330 (doc/invoke.texi): Document new aes and sha2 options.
10332 2018-01-10 Martin Sebor <msebor@redhat.com>
10334 PR tree-optimization/83781
10335 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10338 2018-01-11 Martin Sebor <msebor@gmail.com>
10339 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10341 PR tree-optimization/83501
10342 PR tree-optimization/81703
10344 * tree-ssa-strlen.c (get_string_cst): Rename...
10345 (get_string_len): ...to this. Handle global constants.
10346 (handle_char_store): Adjust.
10348 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10349 Jim Wilson <jimw@sifive.com>
10351 * config/riscv/riscv-protos.h (riscv_output_return): New.
10352 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10353 (riscv_attribute_table, riscv_output_return),
10354 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10355 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10356 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10358 (riscv_expand_prologue): Add early return for naked function.
10359 (riscv_expand_epilogue): Likewise.
10360 (riscv_function_ok_for_sibcall): Return false for naked function.
10361 (riscv_set_current_function): New.
10362 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10363 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10364 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10365 * doc/extend.texi (RISC-V Function Attributes): New.
10367 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10369 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10370 check for 128-bit long double before checking TCmode.
10371 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10372 128-bit long doubles before checking TFmode or TCmode.
10373 (FLOAT128_IBM_P): Likewise.
10375 2018-01-10 Martin Sebor <msebor@redhat.com>
10377 PR tree-optimization/83671
10378 * builtins.c (c_strlen): Unconditionally return zero for the empty
10380 Use -Warray-bounds for warnings.
10381 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10382 for non-constant array indices with COMPONENT_REF, arrays of
10383 arrays, and pointers to arrays.
10384 (gimple_fold_builtin_strlen): Determine and set length range for
10385 non-constant character arrays.
10387 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10389 PR middle-end/81897
10390 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10393 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10395 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10397 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10400 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10401 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10402 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10403 indexed_or_indirect_operand predicate.
10404 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10405 (*vsx_le_perm_load_v8hi): Likewise.
10406 (*vsx_le_perm_load_v16qi): Likewise.
10407 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10408 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10409 (*vsx_le_perm_store_v8hi): Likewise.
10410 (*vsx_le_perm_store_v16qi): Likewise.
10411 (eight unnamed splitters): Likewise.
10413 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10415 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10416 * config/rs6000/emmintrin.h: Likewise.
10417 * config/rs6000/mmintrin.h: Likewise.
10418 * config/rs6000/xmmintrin.h: Likewise.
10420 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10423 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10425 * tree.c (tree_nop_conversion): Return true for location wrapper
10427 (maybe_wrap_with_location): New function.
10428 (selftest::check_strip_nops): New function.
10429 (selftest::test_location_wrappers): New function.
10430 (selftest::tree_c_tests): Call it.
10431 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10432 (maybe_wrap_with_location): New decl.
10433 (EXPR_LOCATION_WRAPPER_P): New macro.
10434 (location_wrapper_p): New inline function.
10435 (tree_strip_any_location_wrapper): New inline function.
10437 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10440 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10441 stack_realign_offset for the largest alignment of stack slot
10443 (ix86_find_max_used_stack_alignment): New function.
10444 (ix86_finalize_stack_frame_flags): Use it. Set
10445 max_used_stack_alignment if we don't realign stack.
10446 * config/i386/i386.h (machine_function): Add
10447 max_used_stack_alignment.
10449 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10451 * config/arm/arm.opt (-mbranch-cost): New option.
10452 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10455 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10458 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10459 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10461 2018-01-10 Richard Biener <rguenther@suse.de>
10464 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10465 early out so it also covers the case where we have a non-NULL
10468 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10470 PR tree-optimization/83753
10471 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10472 for non-strided grouped accesses if the number of elements is 1.
10474 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10477 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10478 * i386.h (TARGET_USE_GATHER): Define.
10479 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10481 2018-01-10 Martin Liska <mliska@suse.cz>
10484 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10485 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10487 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10488 CLEANUP_NO_PARTITIONING is not set.
10490 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10492 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10493 for vectors, as a partial revert of r254296.
10494 * rtl.h (const_vec_p): Delete.
10495 (const_vec_duplicate_p): Don't test for vector CONSTs.
10496 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10497 * expmed.c (make_tree): Likewise.
10500 * common.md (E, F): Use CONSTANT_P instead of checking for
10502 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10503 checking for CONST_VECTOR.
10505 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10507 PR middle-end/83575
10508 * predict.c (force_edge_cold): Handle in more sane way edges
10509 with no prediction.
10511 2018-01-09 Carl Love <cel@us.ibm.com>
10513 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10515 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10516 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10517 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10518 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10519 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10520 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10521 * config/rs6000/rs6000-protos.h: Add extern defition for
10522 rs6000_generate_float2_double_code.
10523 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10525 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10526 (float2_v2df): Add define_expand.
10528 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10531 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10532 op_mode in the force_to_mode call.
10534 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10536 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10537 instead of checking each element individually.
10538 (aarch64_evpc_uzp): Likewise.
10539 (aarch64_evpc_zip): Likewise.
10540 (aarch64_evpc_ext): Likewise.
10541 (aarch64_evpc_rev): Likewise.
10542 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10543 instead of checking each element individually. Return true without
10545 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10546 whether all selected elements come from the same input, instead of
10547 checking each element individually. Remove calls to gen_rtx_REG,
10548 start_sequence and end_sequence and instead assert that no rtl is
10551 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10553 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10554 order of HIGH and CONST checks.
10556 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10558 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10559 if the destination isn't an SSA_NAME.
10561 2018-01-09 Richard Biener <rguenther@suse.de>
10563 PR tree-optimization/83668
10564 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10566 (canonicalize_loop_form): ... here, renamed from ...
10567 (canonicalize_loop_closed_ssa_form): ... this and amended to
10568 swap successor edges for loop exit blocks to make us use
10569 the RPO order we need for initial schedule generation.
10571 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10573 PR tree-optimization/64811
10574 * match.pd: When optimizing comparisons with Inf, avoid
10575 introducing or losing exceptions from comparisons with NaN.
10577 2018-01-09 Martin Liska <mliska@suse.cz>
10580 * asan.c (shadow_mem_size): Add gcc_assert.
10582 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10584 Don't save registers in main().
10587 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10588 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10589 * config/avr/avr.c (avr_set_current_function): Don't error if
10590 naked, OS_task or OS_main are specified at the same time.
10591 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10593 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10595 * common/config/avr/avr-common.c (avr_option_optimization_table):
10596 Switch on -mmain-is-OS_task for optimizing compilations.
10598 2018-01-09 Richard Biener <rguenther@suse.de>
10600 PR tree-optimization/83572
10601 * graphite.c: Include cfganal.h.
10602 (graphite_transform_loops): Connect infinite loops to exit
10603 and remove fake edges at the end.
10605 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10607 * ipa-inline.c (edge_badness): Revert accidental checkin.
10609 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10612 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10613 symbols; not inline clones.
10615 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10618 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10619 hard registers. Formatting fixes.
10621 PR preprocessor/83722
10622 * gcc.c (try_generate_repro): Pass
10623 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10624 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10627 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10628 Kito Cheng <kito.cheng@gmail.com>
10630 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10631 (riscv_leaf_function_p): Delete.
10632 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10634 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10636 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10638 (do_ifelse): New function.
10639 (do_isel): New function.
10640 (do_sub3): New function.
10641 (do_add3): New function.
10642 (do_load_mask_compare): New function.
10643 (do_overlap_load_compare): New function.
10644 (expand_compare_loop): New function.
10645 (expand_block_compare): Call expand_compare_loop() when appropriate.
10646 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10647 option description.
10648 (-mblock-compare-inline-loop-limit): New option.
10650 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10653 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10654 Reverse order of second and third operands in first alternative.
10655 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10656 of first and second elements in UNSPEC_VPERMR vector.
10657 (altivec_expand_vec_perm_le): Likewise.
10659 2018-01-08 Jeff Law <law@redhat.com>
10661 PR rtl-optimizatin/81308
10662 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10663 (process_switch): If group_case_labels makes a change, then set
10665 (pass_convert_switch::execute): If a switch is converted, then
10666 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10668 PR rtl-optimization/81308
10669 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10672 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10674 PR target/83663 - Revert r255946
10675 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10676 generation for cases where splatting a value is not useful.
10677 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10678 across a vec_duplicate and a paradoxical subreg forming a vector
10679 mode to a vec_concat.
10681 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10683 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10684 -march=armv8.3-a variants.
10685 * config/arm/t-multilib: Likewise.
10686 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10688 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10690 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10692 (cceq_ior_compare_complement): Give it a name so I can use it, and
10693 change boolean_or_operator predicate to boolean_operator so it can
10694 be used to generate a crand.
10695 (eqne): New code iterator.
10696 (bd/bd_neg): New code_attrs.
10697 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10698 a single define_insn.
10699 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10700 decrement (bdnzt/bdnzf/bdzt/bdzf).
10701 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10702 with the new names of the branch decrement patterns, and added the
10703 names of the branch decrement conditional patterns.
10705 2018-01-08 Richard Biener <rguenther@suse.de>
10707 PR tree-optimization/83563
10708 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10711 2018-01-08 Richard Biener <rguenther@suse.de>
10713 PR middle-end/83713
10714 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10716 2018-01-08 Richard Biener <rguenther@suse.de>
10718 PR tree-optimization/83685
10719 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10720 references to abnormals.
10722 2018-01-08 Richard Biener <rguenther@suse.de>
10725 * dwarf2out.c (output_indirect_strings): Handle empty
10726 skeleton_debug_str_hash.
10727 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10729 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10731 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10732 (emit_store_direct): Likewise.
10733 (arc_trampoline_adjust_address): Likewise.
10734 (arc_asm_trampoline_template): New function.
10735 (arc_initialize_trampoline): Use asm_trampoline_template.
10736 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10737 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10738 * config/arc/arc.md (flush_icache): Delete pattern.
10740 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10742 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10743 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10746 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10749 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10750 by not USED_FOR_TARGET.
10751 (make_pass_resolve_sw_modes): Likewise.
10753 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10755 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10758 2018-01-08 Richard Biener <rguenther@suse.de>
10760 PR middle-end/83580
10761 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10763 2018-01-08 Richard Biener <rguenther@suse.de>
10765 PR middle-end/83517
10766 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10768 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10770 PR middle-end/81897
10771 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10772 basic blocks with a small number of successors.
10773 (convert_control_dep_chain_into_preds): Improve handling of
10775 (dump_predicates): Split apart into...
10776 (dump_pred_chain): ...here...
10777 (dump_pred_info): ...and here.
10778 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10779 (can_chain_union_be_invalidated_p): Improve check for invalidation
10781 (uninit_uses_cannot_happen): Avoid unnecessary if
10782 convert_control_dep_chain_into_preds yielded nothing.
10784 2018-01-06 Martin Sebor <msebor@redhat.com>
10786 PR tree-optimization/83640
10787 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10788 subtracting negative offset from size.
10789 (builtin_access::overlap): Adjust offset bounds of the access to fall
10790 within the size of the object if possible.
10792 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10794 PR rtl-optimization/83699
10795 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10796 extract_bit_field_as_subreg to cases in which the extracted
10797 value is also a vector.
10799 * lra-constraints.c (process_alt_operands): Test for the equivalence
10800 substitutions when detecting a possible reload cycle.
10802 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10805 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10806 by default if flag_selective_schedling{,2}. Formatting fixes.
10808 PR rtl-optimization/83682
10809 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10810 if it has non-VECTOR_MODE element mode.
10811 (vec_duplicate_p): Likewise.
10813 PR middle-end/83694
10814 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10815 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10817 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10820 * config/i386/i386-builtin.def
10821 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10822 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10823 Require also OPTION_MASK_ISA_AVX512F in addition to
10824 OPTION_MASK_ISA_GFNI.
10825 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10826 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10827 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10828 to OPTION_MASK_ISA_GFNI.
10829 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10830 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10831 OPTION_MASK_ISA_AVX512BW.
10832 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10833 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10834 addition to OPTION_MASK_ISA_GFNI.
10835 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10836 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10837 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10838 to OPTION_MASK_ISA_GFNI.
10839 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10840 a requirement for all ISAs rather than any of them with a few
10842 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10844 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10845 bitmasks to be enabled with 3 exceptions, instead of requiring any
10846 enabled ISA with lots of exceptions.
10847 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10848 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10849 Change avx512bw in isa attribute to avx512f.
10850 * config/i386/sgxintrin.h: Add license boilerplate.
10851 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10852 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10853 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10854 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10856 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10857 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10858 temporarily sse2 rather than sse if not enabled already.
10861 * config/i386/sse.md (VI248_VLBW): Rename to ...
10862 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10863 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10864 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10865 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10866 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10867 mode iterator instead of VI248_VLBW.
10869 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10871 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10872 (record_modified): Skip clobbers; add debug output.
10873 (param_change_prob): Use sreal frequencies.
10875 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10877 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10878 punt for user-aligned variables.
10880 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10882 * tree-chrec.c (chrec_contains_symbols): Return true for
10885 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10888 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10889 of (x|y) == x for BICS pattern.
10891 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10893 PR tree-optimization/83605
10894 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10895 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10898 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10900 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10901 * config/epiphany/rtems.h: New file.
10903 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10904 Uros Bizjak <ubizjak@gmail.com>
10907 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10908 QIreg_operand instead of register_operand predicate.
10909 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10910 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10911 comments instead of -fmitigate[-_]rop.
10913 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10916 * cgraphunit.c (symbol_table::compile): Switch to text_section
10917 before calling assembly_start debug hook.
10918 * run-rtl-passes.c (run_rtl_passes): Likewise.
10921 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10923 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10924 range_int_cst_p rather than !symbolic_range_p before calling
10925 extract_range_from_multiplicative_op_1.
10927 2018-01-04 Jeff Law <law@redhat.com>
10929 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10930 redundant test in assertion.
10932 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10934 * doc/rtl.texi: Document machine_mode wrapper classes.
10936 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10938 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10939 using tree_to_uhwi.
10941 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10943 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10944 the VEC_PERM_EXPR fold to fail.
10946 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10949 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10950 to switched_sections.
10952 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10955 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10956 test for d.testing.
10958 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10961 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10962 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10964 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10967 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10968 is BLKmode and bitpos not zero or mode change is needed.
10970 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10973 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10976 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10979 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10980 instead of MULT rtx. Update all corresponding splitters.
10981 (*saddl_se): Ditto.
10982 (*ssub<modesuffix>): Ditto.
10983 (*ssubl_se): Ditto.
10984 (*cmp_sadd_di): Update split patterns.
10985 (*cmp_sadd_si): Ditto.
10986 (*cmp_sadd_sidi): Ditto.
10987 (*cmp_ssub_di): Ditto.
10988 (*cmp_ssub_si): Ditto.
10989 (*cmp_ssub_sidi): Ditto.
10990 * config/alpha/predicates.md (const23_operand): New predicate.
10991 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10992 Look for ASHIFT, not MULT inner operand.
10993 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10995 2018-01-04 Martin Liska <mliska@suse.cz>
10997 PR gcov-profile/83669
10998 * gcov.c (output_intermediate_file): Add version to intermediate
11000 * doc/gcov.texi: Document new field 'version' in intermediate
11001 file format. Fix location of '-k' option of gcov command.
11003 2018-01-04 Martin Liska <mliska@suse.cz>
11006 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11008 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11010 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11012 2018-01-03 Martin Sebor <msebor@redhat.com>
11014 PR tree-optimization/83655
11015 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11016 checking calls with invalid arguments.
11018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11020 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11021 (vectorizable_mask_load_store): Delete.
11022 (vectorizable_call): Return false for masked loads and stores.
11023 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11024 instead of gimple_assign_rhs1.
11025 (vectorizable_load): Handle IFN_MASK_LOAD.
11026 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11030 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11032 (vectorizable_mask_load_store): ...here.
11033 (vectorizable_load): ...and here.
11035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11037 * tree-vect-stmts.c (vect_build_all_ones_mask)
11038 (vect_build_zero_merge_argument): New functions, split out from...
11039 (vectorizable_load): ...here.
11041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11043 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11045 (vectorizable_mask_load_store): ...here.
11046 (vectorizable_store): ...and here.
11048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11050 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11052 (vectorizable_mask_load_store): ...here.
11054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11056 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11057 (vect_model_store_cost): Take a vec_load_store_type instead of a
11059 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11060 (vect_model_store_cost): Take a vec_load_store_type instead of a
11062 (vectorizable_mask_load_store): Update accordingly.
11063 (vectorizable_store): Likewise.
11064 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11068 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11069 IFN_MASK_LOAD calls here rather than...
11070 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11073 Alan Hayward <alan.hayward@arm.com>
11074 David Sherwood <david.sherwood@arm.com>
11076 * expmed.c (extract_bit_field_1): For vector extracts,
11077 fall back to extract_bit_field_as_subreg if vec_extract
11080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11081 Alan Hayward <alan.hayward@arm.com>
11082 David Sherwood <david.sherwood@arm.com>
11084 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11085 they are variable or constant sized.
11086 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11087 slots for constant-sized data.
11089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11090 Alan Hayward <alan.hayward@arm.com>
11091 David Sherwood <david.sherwood@arm.com>
11093 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11094 handling COND_EXPRs with boolean comparisons, try to find a better
11095 basis for the mask type than the boolean itself.
11097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11099 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11100 is calculated and how it can be overridden.
11101 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11102 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11104 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11108 Alan Hayward <alan.hayward@arm.com>
11109 David Sherwood <david.sherwood@arm.com>
11111 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11112 Remove the mode argument.
11113 (aarch64_simd_valid_immediate): Remove the mode and inverse
11115 * config/aarch64/iterators.md (bitsize): New iterator.
11116 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11117 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11118 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11119 aarch64_simd_valid_immediate.
11120 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11121 (aarch64_reg_or_bic_imm): Likewise.
11122 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11123 with an insn_type enum and msl with a modifier_type enum.
11124 Replace element_width with a scalar_mode. Change the shift
11125 to unsigned int. Add constructors for scalar_float_mode and
11126 scalar_int_mode elements.
11127 (aarch64_vect_float_const_representable_p): Delete.
11128 (aarch64_can_const_movi_rtx_p)
11129 (aarch64_simd_scalar_immediate_valid_for_move)
11130 (aarch64_simd_make_constant): Update call to
11131 aarch64_simd_valid_immediate.
11132 (aarch64_advsimd_valid_immediate_hs): New function.
11133 (aarch64_advsimd_valid_immediate): Likewise.
11134 (aarch64_simd_valid_immediate): Remove mode and inverse
11135 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11136 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11137 and aarch64_float_const_representable_p on the result.
11138 (aarch64_output_simd_mov_immediate): Remove mode argument.
11139 Update call to aarch64_simd_valid_immediate and use of
11140 simd_immediate_info.
11141 (aarch64_output_scalar_simd_mov_immediate): Update call
11144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11145 Alan Hayward <alan.hayward@arm.com>
11146 David Sherwood <david.sherwood@arm.com>
11148 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11149 (mode_nunits): Likewise CONST_MODE_NUNITS.
11150 * machmode.def (ADJUST_NUNITS): Document.
11151 * genmodes.c (mode_data::need_nunits_adj): New field.
11152 (blank_mode): Update accordingly.
11153 (adj_nunits): New variable.
11154 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11156 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11157 listed in adj_nunits.
11158 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11159 listed in adj_nunits. Don't emit case statements for such modes.
11160 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11161 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11162 nothing if adj_nunits is nonnull.
11163 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11164 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11165 (emit_mode_fbit): Update use of print_maybe_const_decl.
11166 (emit_move_size): Likewise. Treat the array as non-const
11168 (emit_mode_adjustments): Handle adj_nunits.
11170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11172 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11173 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11174 (VECTOR_MODES): Use it.
11175 (make_vector_modes): Take the prefix as an argument.
11177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11178 Alan Hayward <alan.hayward@arm.com>
11179 David Sherwood <david.sherwood@arm.com>
11181 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11182 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11183 for MODE_VECTOR_BOOL.
11184 * machmode.def (VECTOR_BOOL_MODE): Document.
11185 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11186 (make_vector_bool_mode): New function.
11187 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11189 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11190 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11192 * stor-layout.c (int_mode_for_mode): Likewise.
11193 * tree.c (build_vector_type_for_mode): Likewise.
11194 * varasm.c (output_constant_pool_2): Likewise.
11195 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11196 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11197 for MODE_VECTOR_BOOL.
11198 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11199 of mode class checks.
11200 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11201 instead of a list of mode class checks.
11202 (expand_vector_scalar_condition): Likewise.
11203 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11206 Alan Hayward <alan.hayward@arm.com>
11207 David Sherwood <david.sherwood@arm.com>
11209 * machmode.h (mode_size): Change from unsigned short to
11211 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11212 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11213 or if measurement_type is not polynomial.
11214 (fixed_size_mode::includes_p): Check for constant-sized modes.
11215 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11216 return a poly_uint16 rather than an unsigned short.
11217 (emit_mode_size): Change the type of mode_size from unsigned short
11218 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11219 (emit_mode_adjustments): Cope with polynomial vector sizes.
11220 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11222 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11224 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11225 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11226 * caller-save.c (setup_save_areas): Likewise.
11227 (replace_reg_with_saved_mem): Likewise.
11228 * calls.c (emit_library_call_value_1): Likewise.
11229 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11230 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11231 (gen_lowpart_for_combine): Likewise.
11232 * convert.c (convert_to_integer_1): Likewise.
11233 * cse.c (equiv_constant, cse_insn): Likewise.
11234 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11235 (cselib_subst_to_values): Likewise.
11236 * dce.c (word_dce_process_block): Likewise.
11237 * df-problems.c (df_word_lr_mark_ref): Likewise.
11238 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11239 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11240 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11241 (rtl_for_decl_location): Likewise.
11242 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11243 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11244 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11245 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11246 (expand_expr_real_1): Likewise.
11247 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11248 (pad_below): Likewise.
11249 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11250 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11251 * ira.c (get_subreg_tracking_sizes): Likewise.
11252 * ira-build.c (ira_create_allocno_objects): Likewise.
11253 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11254 (ira_sort_regnos_for_alter_reg): Likewise.
11255 * ira-costs.c (record_operand_costs): Likewise.
11256 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11257 (resolve_simple_move): Likewise.
11258 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11259 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11260 (lra_constraints): Likewise.
11261 (CONST_POOL_OK_P): Reject variable-sized modes.
11262 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11263 (add_pseudo_to_slot, lra_spill): Likewise.
11264 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11265 * optabs-query.c (get_best_extraction_insn): Likewise.
11266 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11267 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11268 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11269 * recog.c (offsettable_address_addr_space_p): Likewise.
11270 * regcprop.c (maybe_mode_change): Likewise.
11271 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11272 * regrename.c (build_def_use): Likewise.
11273 * regstat.c (dump_reg_info): Likewise.
11274 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11275 (find_reloads, find_reloads_subreg_address): Likewise.
11276 * reload1.c (eliminate_regs_1): Likewise.
11277 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11278 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11279 (simplify_binary_operation_1, simplify_subreg): Likewise.
11280 * targhooks.c (default_function_arg_padding): Likewise.
11281 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11282 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11283 (verify_gimple_assign_ternary): Likewise.
11284 * tree-inline.c (estimate_move_cost): Likewise.
11285 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11286 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11287 (get_address_cost_ainc): Likewise.
11288 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11289 (vect_supportable_dr_alignment): Likewise.
11290 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11291 (vectorizable_reduction): Likewise.
11292 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11293 (vectorizable_operation, vectorizable_load): Likewise.
11294 * tree.c (build_same_sized_truth_vector_type): Likewise.
11295 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11296 * var-tracking.c (emit_note_insn_var_location): Likewise.
11297 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11298 (ADDR_VEC_ALIGN): 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 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11306 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11307 or if measurement_type is polynomial.
11308 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11309 * combine.c (make_extraction): Likewise.
11310 * dse.c (find_shift_sequence): Likewise.
11311 * dwarf2out.c (mem_loc_descriptor): Likewise.
11312 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11313 (extract_bit_field, extract_low_bits): Likewise.
11314 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11315 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11316 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11317 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11318 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11319 * reload.c (find_reloads): Likewise.
11320 * reload1.c (alter_reg): Likewise.
11321 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11322 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11323 * tree-if-conv.c (predicate_mem_writes): Likewise.
11324 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11325 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11326 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11327 * valtrack.c (dead_debug_insert_temp): Likewise.
11328 * varasm.c (mergeable_constant_section): Likewise.
11329 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11332 Alan Hayward <alan.hayward@arm.com>
11333 David Sherwood <david.sherwood@arm.com>
11335 * expr.c (expand_assignment): Cope with polynomial mode sizes
11336 when assigning to a CONCAT.
11338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11339 Alan Hayward <alan.hayward@arm.com>
11340 David Sherwood <david.sherwood@arm.com>
11342 * machmode.h (mode_precision): Change from unsigned short to
11344 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11346 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11347 or if measurement_type is not polynomial.
11348 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11349 in which the mode is already known to be a scalar_int_mode.
11350 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11351 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11353 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11354 for GET_MODE_PRECISION.
11355 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11356 for GET_MODE_PRECISION.
11357 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11359 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11360 (expand_field_assignment, make_extraction): Likewise.
11361 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11362 (get_last_value): Likewise.
11363 * convert.c (convert_to_integer_1): Likewise.
11364 * cse.c (cse_insn): Likewise.
11365 * expr.c (expand_expr_real_1): Likewise.
11366 * lra-constraints.c (simplify_operand_subreg): Likewise.
11367 * optabs-query.c (can_atomic_load_p): Likewise.
11368 * optabs.c (expand_atomic_load): Likewise.
11369 (expand_atomic_store): Likewise.
11370 * ree.c (combine_reaching_defs): Likewise.
11371 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11372 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11373 * tree.h (type_has_mode_precision_p): Likewise.
11374 * ubsan.c (instrument_si_overflow): Likewise.
11376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11377 Alan Hayward <alan.hayward@arm.com>
11378 David Sherwood <david.sherwood@arm.com>
11380 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11381 polynomial numbers of units.
11382 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11383 (valid_vector_subparts_p): New function.
11384 (build_vector_type): Remove temporary shim and take the number
11385 of units as a poly_uint64 rather than an int.
11386 (build_opaque_vector_type): Take the number of units as a
11387 poly_uint64 rather than an int.
11388 * tree.c (build_vector_from_ctor): Handle polynomial
11389 TYPE_VECTOR_SUBPARTS.
11390 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11391 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11392 (build_vector_from_val): If the number of units is variable,
11393 use build_vec_duplicate_cst for constant operands and
11394 VEC_DUPLICATE_EXPR otherwise.
11395 (make_vector_type): Remove temporary is_constant ().
11396 (build_vector_type, build_opaque_vector_type): Take the number of
11397 units as a poly_uint64 rather than an int.
11398 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11400 * cfgexpand.c (expand_debug_expr): Likewise.
11401 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11402 (store_constructor, expand_expr_real_1): Likewise.
11403 (const_scalar_mask_from_tree): Likewise.
11404 * fold-const-call.c (fold_const_reduction): Likewise.
11405 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11406 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11407 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11408 (fold_relational_const): Likewise.
11409 (native_interpret_vector): Likewise. Change the size from an
11410 int to an unsigned int.
11411 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11412 TYPE_VECTOR_SUBPARTS.
11413 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11414 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11415 duplicating a non-constant operand into a variable-length vector.
11416 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11417 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11418 * ipa-icf.c (sem_variable::equals): Likewise.
11419 * match.pd: Likewise.
11420 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11421 * print-tree.c (print_node): Likewise.
11422 * stor-layout.c (layout_type): Likewise.
11423 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11424 * tree-cfg.c (verify_gimple_comparison): Likewise.
11425 (verify_gimple_assign_binary): Likewise.
11426 (verify_gimple_assign_ternary): Likewise.
11427 (verify_gimple_assign_single): Likewise.
11428 * tree-pretty-print.c (dump_generic_node): Likewise.
11429 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11430 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11431 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11432 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11433 (vect_shift_permute_load_chain): Likewise.
11434 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11435 (expand_vector_condition, optimize_vector_constructor): Likewise.
11436 (lower_vec_perm, get_compute_type): Likewise.
11437 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11438 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11439 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11440 (vect_recog_mask_conversion_pattern): Likewise.
11441 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11442 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11443 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11444 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11445 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11446 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11447 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11448 (supportable_widening_operation): Likewise.
11449 (supportable_narrowing_operation): Likewise.
11450 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11452 * varasm.c (output_constant): Likewise.
11454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11455 Alan Hayward <alan.hayward@arm.com>
11456 David Sherwood <david.sherwood@arm.com>
11458 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11459 so that both the length == 3 and length != 3 cases set up their
11460 own permute vectors. Add comments explaining why we know the
11461 number of elements is constant.
11462 (vect_permute_load_chain): Likewise.
11464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11465 Alan Hayward <alan.hayward@arm.com>
11466 David Sherwood <david.sherwood@arm.com>
11468 * machmode.h (mode_nunits): Change from unsigned char to
11470 (ONLY_FIXED_SIZE_MODES): New macro.
11471 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11472 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11473 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11475 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11476 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11477 or if measurement_type is not polynomial.
11478 * genmodes.c (ZERO_COEFFS): New macro.
11479 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11481 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11482 Use ZERO_COEFFS when emitting initializers.
11483 * data-streamer.h (bp_pack_poly_value): New function.
11484 (bp_unpack_poly_value): Likewise.
11485 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11486 for GET_MODE_NUNITS.
11487 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11488 for GET_MODE_NUNITS.
11489 * tree.c (make_vector_type): Remove temporary shim and make
11490 the real function take the number of units as a poly_uint64
11491 rather than an int.
11492 (build_vector_type_for_mode): Handle polynomial nunits.
11493 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11494 * emit-rtl.c (const_vec_series_p_1): Likewise.
11495 (gen_rtx_CONST_VECTOR): Likewise.
11496 * fold-const.c (test_vec_duplicate_folding): Likewise.
11497 * genrecog.c (validate_pattern): Likewise.
11498 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11499 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11500 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11501 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11502 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11503 * rtlanal.c (subreg_get_info): Likewise.
11504 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11505 (vect_grouped_load_supported): Likewise.
11506 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11507 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11508 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11509 (simplify_const_unary_operation, simplify_binary_operation_1)
11510 (simplify_const_binary_operation, simplify_ternary_operation)
11511 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11512 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11513 instead of CONST_VECTOR_NUNITS.
11514 * varasm.c (output_constant_pool_2): Likewise.
11515 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11516 explicit-encoded elements in the XVEC for variable-length vectors.
11518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11520 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11523 Alan Hayward <alan.hayward@arm.com>
11524 David Sherwood <david.sherwood@arm.com>
11526 * coretypes.h (fixed_size_mode): Declare.
11527 (fixed_size_mode_pod): New typedef.
11528 * builtins.h (target_builtins::x_apply_args_mode)
11529 (target_builtins::x_apply_result_mode): Change type to
11530 fixed_size_mode_pod.
11531 * builtins.c (apply_args_size, apply_result_size, result_vector)
11532 (expand_builtin_apply_args_1, expand_builtin_apply)
11533 (expand_builtin_return): Update accordingly.
11535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11537 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11538 * cselib.c (cselib_hash_rtx): Likewise.
11539 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11540 CONST_VECTOR encoding.
11542 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11543 Jeff Law <law@redhat.com>
11546 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11547 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11548 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11549 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11552 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11553 explicitly probe *sp in a noreturn function if there were any callee
11554 register saves or frame pointer is needed.
11556 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11559 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11560 BLKmode for ternary, binary or unary expressions.
11563 * var-tracking.c (delete_vta_debug_insn): New inline function.
11564 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11565 insns from get_insns () to NULL instead of each bb separately.
11566 Use delete_vta_debug_insn. No longer static.
11567 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11568 delete_vta_debug_insns callers.
11569 * rtl.h (delete_vta_debug_insns): Declare.
11570 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11571 instead of variable_tracking_main.
11573 2018-01-03 Martin Sebor <msebor@redhat.com>
11575 PR tree-optimization/83603
11576 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11577 arguments past the endof the argument list in functions declared
11578 without a prototype.
11579 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11580 Avoid checking when arguments are null.
11582 2018-01-03 Martin Sebor <msebor@redhat.com>
11585 * doc/extend.texi (attribute const): Fix a typo.
11586 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11587 issuing -Wsuggest-attribute for void functions.
11589 2018-01-03 Martin Sebor <msebor@redhat.com>
11591 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11592 offset_int::from instead of wide_int::to_shwi.
11593 (maybe_diag_overlap): Remove assertion.
11594 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11595 * gimple-ssa-sprintf.c (format_directive): Same.
11596 (parse_directive): Same.
11597 (sprintf_dom_walker::compute_format_length): Same.
11598 (try_substitute_return_value): Same.
11600 2018-01-03 Jeff Law <law@redhat.com>
11602 PR middle-end/83654
11603 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11604 non-constant residual for zero at runtime and avoid probing in
11605 that case. Reorganize code for trailing problem to mirror handling
11608 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11610 PR tree-optimization/83501
11611 * tree-ssa-strlen.c (get_string_cst): New.
11612 (handle_char_store): Call get_string_cst.
11614 2018-01-03 Martin Liska <mliska@suse.cz>
11616 PR tree-optimization/83593
11617 * tree-ssa-strlen.c: Include tree-cfg.h.
11618 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11619 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11620 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11622 (strlen_dom_walker::before_dom_children): Call
11623 gimple_purge_dead_eh_edges. Dump tranformation with details
11625 (strlen_dom_walker::before_dom_children): Update call by adding
11626 new argument cleanup_eh.
11627 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11629 2018-01-03 Martin Liska <mliska@suse.cz>
11632 * cif-code.def (VARIADIC_THUNK): New enum value.
11633 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11636 2018-01-03 Jan Beulich <jbeulich@suse.com>
11638 * sse.md (mov<mode>_internal): Tighten condition for when to use
11639 vmovdqu<ssescalarsize> for TI and OI modes.
11641 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11643 Update copyright years.
11645 2018-01-03 Martin Liska <mliska@suse.cz>
11648 * ipa-visibility.c (function_and_variable_visibility): Skip
11649 functions with noipa attribure.
11651 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11653 * gcc.c (process_command): Update copyright notice dates.
11654 * gcov-dump.c (print_version): Ditto.
11655 * gcov.c (print_version): Ditto.
11656 * gcov-tool.c (print_version): Ditto.
11657 * gengtype.c (create_file): Ditto.
11658 * doc/cpp.texi: Bump @copying's copyright year.
11659 * doc/cppinternals.texi: Ditto.
11660 * doc/gcc.texi: Ditto.
11661 * doc/gccint.texi: Ditto.
11662 * doc/gcov.texi: Ditto.
11663 * doc/install.texi: Ditto.
11664 * doc/invoke.texi: Ditto.
11666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11668 * vector-builder.h (vector_builder::m_full_nelts): Change from
11669 unsigned int to poly_uint64.
11670 (vector_builder::full_nelts): Update prototype accordingly.
11671 (vector_builder::new_vector): Likewise.
11672 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11673 (vector_builder::operator ==): Likewise.
11674 (vector_builder::finalize): Likewise.
11675 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11676 Take the number of elements as a poly_uint64 rather than an
11678 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11679 from unsigned int to poly_uint64.
11680 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11681 (vec_perm_indices::new_vector): Likewise.
11682 (vec_perm_indices::length): Likewise.
11683 (vec_perm_indices::nelts_per_input): Likewise.
11684 (vec_perm_indices::input_nelts): Likewise.
11685 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11686 number of elements per input as a poly_uint64 rather than an
11687 unsigned int. Use the original encoding for variable-length
11688 vectors, rather than clamping each individual element.
11689 For the second and subsequent elements in each pattern,
11690 clamp the step and base before clamping their sum.
11691 (vec_perm_indices::series_p): Handle polynomial element counts.
11692 (vec_perm_indices::all_in_range_p): Likewise.
11693 (vec_perm_indices_to_tree): Likewise.
11694 (vec_perm_indices_to_rtx): Likewise.
11695 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11696 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11697 (tree_vector_builder::new_binary_operation): Handle polynomial
11698 element counts. Return false if we need to know the number
11699 of elements at compile time.
11700 * fold-const.c (fold_vec_perm): Punt if the number of elements
11701 isn't known at compile time.
11703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11705 * vec-perm-indices.h (vec_perm_builder): Change element type
11706 from HOST_WIDE_INT to poly_int64.
11707 (vec_perm_indices::element_type): Update accordingly.
11708 (vec_perm_indices::clamp): Handle polynomial element_types.
11709 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11710 (vec_perm_indices::all_in_range_p): Likewise.
11711 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11713 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11714 polynomial vec_perm_indices element types.
11715 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11716 * fold-const.c (fold_vec_perm): Likewise.
11717 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11718 * tree-vect-generic.c (lower_vec_perm): Likewise.
11719 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11720 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11721 element type to HOST_WIDE_INT.
11723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11724 Alan Hayward <alan.hayward@arm.com>
11725 David Sherwood <david.sherwood@arm.com>
11727 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11728 rather than an int. Use plus_constant.
11729 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11730 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11733 Alan Hayward <alan.hayward@arm.com>
11734 David Sherwood <david.sherwood@arm.com>
11736 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11737 a HOST_WIDE_INT to a poly_int64.
11739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11740 Alan Hayward <alan.hayward@arm.com>
11741 David Sherwood <david.sherwood@arm.com>
11743 * calls.c (load_register_parameters): Cope with polynomial
11744 mode sizes. Require a constant size for BLKmode parameters
11745 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11746 forces a parameter to be padded at the lsb end in order to
11747 fill a complete number of words, require the parameter size
11748 to be ordered wrt UNITS_PER_WORD.
11750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11751 Alan Hayward <alan.hayward@arm.com>
11752 David Sherwood <david.sherwood@arm.com>
11754 * reload1.c (spill_stack_slot_width): Change element type
11755 from unsigned int to poly_uint64_pod.
11756 (alter_reg): Treat mode sizes as polynomial.
11758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11759 Alan Hayward <alan.hayward@arm.com>
11760 David Sherwood <david.sherwood@arm.com>
11762 * reload.c (complex_word_subreg_p): New function.
11763 (reload_inner_reg_of_subreg, push_reload): Use it.
11765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11766 Alan Hayward <alan.hayward@arm.com>
11767 David Sherwood <david.sherwood@arm.com>
11769 * lra-constraints.c (process_alt_operands): Reject matched
11770 operands whose sizes aren't ordered.
11771 (match_reload): Refer to this check here.
11773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11774 Alan Hayward <alan.hayward@arm.com>
11775 David Sherwood <david.sherwood@arm.com>
11777 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11778 that the mode size is in the set {1, 2, 4, 8, 16}.
11780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11781 Alan Hayward <alan.hayward@arm.com>
11782 David Sherwood <david.sherwood@arm.com>
11784 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11785 Use plus_constant instead of gen_rtx_PLUS.
11787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11788 Alan Hayward <alan.hayward@arm.com>
11789 David Sherwood <david.sherwood@arm.com>
11791 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11792 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11793 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11794 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11795 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11796 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11797 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11798 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11799 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11800 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11802 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11803 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11804 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11805 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11806 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11807 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11808 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11809 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11810 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11811 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11813 * expr.c (emit_move_resolve_push): Treat the input and result
11814 of PUSH_ROUNDING as a poly_int64.
11815 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11816 (emit_push_insn): Likewise.
11817 * lra-eliminations.c (mark_not_eliminable): Likewise.
11818 * recog.c (push_operand): Likewise.
11819 * reload1.c (elimination_effects): Likewise.
11820 * rtlanal.c (nonzero_bits1): Likewise.
11821 * calls.c (store_one_arg): Likewise. Require the padding to be
11822 known at compile time.
11824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11825 Alan Hayward <alan.hayward@arm.com>
11826 David Sherwood <david.sherwood@arm.com>
11828 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11829 Use plus_constant instead of gen_rtx_PLUS.
11831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11832 Alan Hayward <alan.hayward@arm.com>
11833 David Sherwood <david.sherwood@arm.com>
11835 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11836 rather than an int.
11838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11839 Alan Hayward <alan.hayward@arm.com>
11840 David Sherwood <david.sherwood@arm.com>
11842 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11843 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11844 via stack temporaries. Treat the mode size as polynomial too.
11846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11847 Alan Hayward <alan.hayward@arm.com>
11848 David Sherwood <david.sherwood@arm.com>
11850 * expr.c (expand_expr_real_2): When handling conversions involving
11851 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11852 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11853 as a poly_uint64 too.
11855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11856 Alan Hayward <alan.hayward@arm.com>
11857 David Sherwood <david.sherwood@arm.com>
11859 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11862 Alan Hayward <alan.hayward@arm.com>
11863 David Sherwood <david.sherwood@arm.com>
11865 * combine.c (can_change_dest_mode): Handle polynomial
11866 REGMODE_NATURAL_SIZE.
11867 * expmed.c (store_bit_field_1): Likewise.
11868 * expr.c (store_constructor): Likewise.
11869 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11870 and polynomial REGMODE_NATURAL_SIZE.
11871 (gen_lowpart_common): Likewise.
11872 * reginfo.c (record_subregs_of_mode): Likewise.
11873 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11880 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
11888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11889 Alan Hayward <alan.hayward@arm.com>
11890 David Sherwood <david.sherwood@arm.com>
11892 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11893 in a POINTER_PLUS_EXPR.
11895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11896 Alan Hayward <alan.hayward@arm.com>
11897 David Sherwood <david.sherwood@arm.com>
11899 * omp-simd-clone.c (simd_clone_subparts): New function.
11900 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11901 (ipa_simd_modify_function_body): Likewise.
11903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11904 Alan Hayward <alan.hayward@arm.com>
11905 David Sherwood <david.sherwood@arm.com>
11907 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11908 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11909 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11910 (expand_vector_condition, vector_element): Likewise.
11911 (subparts_gt): New function.
11912 (get_compute_type): Use subparts_gt.
11913 (count_type_subparts): Delete.
11914 (expand_vector_operations_1): Use subparts_gt instead of
11915 count_type_subparts.
11917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11918 Alan Hayward <alan.hayward@arm.com>
11919 David Sherwood <david.sherwood@arm.com>
11921 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11922 (vect_compile_time_alias): ...this new function. Do the calculation
11923 on poly_ints rather than trees.
11924 (vect_prune_runtime_alias_test_list): Update call accordingly.
11926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11927 Alan Hayward <alan.hayward@arm.com>
11928 David Sherwood <david.sherwood@arm.com>
11930 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11932 (vect_schedule_slp_instance): Likewise.
11934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11935 Alan Hayward <alan.hayward@arm.com>
11936 David Sherwood <david.sherwood@arm.com>
11938 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11939 constant and extern definitions for variable-length vectors.
11940 (vect_get_constant_vectors): Note that the number of units
11941 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11948 of units as polynomial. Choose between WIDE and NARROW based
11951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11952 Alan Hayward <alan.hayward@arm.com>
11953 David Sherwood <david.sherwood@arm.com>
11955 * tree-vect-stmts.c (simd_clone_subparts): New function.
11956 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11959 Alan Hayward <alan.hayward@arm.com>
11960 David Sherwood <david.sherwood@arm.com>
11962 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11963 vectors as polynomial. Use build_index_vector for
11964 IFN_GOMP_SIMD_LANE.
11966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11967 Alan Hayward <alan.hayward@arm.com>
11968 David Sherwood <david.sherwood@arm.com>
11970 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11971 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11972 for variable-length vectors.
11973 (vectorizable_mask_load_store): Treat the number of units as
11974 polynomial, asserting that it is constant if the condition has
11975 already been enforced.
11976 (vectorizable_store, vectorizable_load): Likewise.
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 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11983 of units as polynomial. Punt if we can't tell at compile time
11984 which vector contains the final result.
11986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11987 Alan Hayward <alan.hayward@arm.com>
11988 David Sherwood <david.sherwood@arm.com>
11990 * tree-vect-loop.c (vectorizable_induction): Treat the number
11991 of units as polynomial. Punt on SLP inductions. Use an integer
11992 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11993 cast of such a series for variable-length floating-point
11996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11997 Alan Hayward <alan.hayward@arm.com>
11998 David Sherwood <david.sherwood@arm.com>
12000 * tree.h (build_index_vector): Declare.
12001 * tree.c (build_index_vector): New function.
12002 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12003 of units as polynomial, forcibly converting it to a constant if
12004 vectorizable_reduction has already enforced the condition.
12005 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12006 to create a {1,2,3,...} vector.
12007 (vectorizable_reduction): Treat the number of units as polynomial.
12008 Choose vectype_in based on the largest scalar element size rather
12009 than the smallest number of units. Enforce the restrictions
12012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12013 Alan Hayward <alan.hayward@arm.com>
12014 David Sherwood <david.sherwood@arm.com>
12016 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12017 number of units as polynomial.
12019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12020 Alan Hayward <alan.hayward@arm.com>
12021 David Sherwood <david.sherwood@arm.com>
12023 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12024 * target.def (autovectorize_vector_sizes): Return the vector sizes
12025 by pointer, using vector_sizes rather than a bitmask.
12026 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12027 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12028 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12030 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12031 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12032 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12033 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12034 * omp-general.c (omp_max_vf): Likewise.
12035 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12036 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12037 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12038 * tree-vect-slp.c (vect_slp_bb): Likewise.
12039 * doc/tm.texi: Regenerate.
12040 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12042 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12043 the vector size as a poly_uint64 rather than an unsigned int.
12044 (current_vector_size): Change from an unsigned int to a poly_uint64.
12045 (get_vectype_for_scalar_type): Update accordingly.
12046 * tree.h (build_truth_vector_type): Take the size and number of
12047 units as a poly_uint64 rather than an unsigned int.
12048 (build_vector_type): Add a temporary overload that takes
12049 the number of units as a poly_uint64 rather than an unsigned int.
12050 * tree.c (make_vector_type): Likewise.
12051 (build_truth_vector_type): Take the number of units as a poly_uint64
12052 rather than an unsigned int.
12054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12055 Alan Hayward <alan.hayward@arm.com>
12056 David Sherwood <david.sherwood@arm.com>
12058 * target.def (get_mask_mode): Take the number of units and length
12059 as poly_uint64s rather than unsigned ints.
12060 * targhooks.h (default_get_mask_mode): Update accordingly.
12061 * targhooks.c (default_get_mask_mode): Likewise.
12062 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12063 * doc/tm.texi: Regenerate.
12065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12066 Alan Hayward <alan.hayward@arm.com>
12067 David Sherwood <david.sherwood@arm.com>
12069 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12070 * omp-general.c (omp_max_vf): Likewise.
12071 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12072 (expand_omp_simd): Handle polynomial safelen.
12073 * omp-low.c (omplow_simd_context): Add a default constructor.
12074 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12075 (lower_rec_simd_input_clauses): Update accordingly.
12076 (lower_rec_input_clauses): Likewise.
12078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12079 Alan Hayward <alan.hayward@arm.com>
12080 David Sherwood <david.sherwood@arm.com>
12082 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12083 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12084 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12085 (vect_analyze_slp_cost): Likewise.
12086 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12087 (vect_model_load_cost): Likewise.
12089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12090 Alan Hayward <alan.hayward@arm.com>
12091 David Sherwood <david.sherwood@arm.com>
12093 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12094 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12095 from an unsigned int * to a poly_uint64_pod *.
12096 (calculate_unrolling_factor): New function.
12097 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12100 Alan Hayward <alan.hayward@arm.com>
12101 David Sherwood <david.sherwood@arm.com>
12103 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12104 from an unsigned int to a poly_uint64.
12105 (_loop_vec_info::slp_unrolling_factor): Likewise.
12106 (_loop_vec_info::vectorization_factor): Change from an int
12108 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12109 (vect_get_num_vectors): New function.
12110 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12111 (vect_get_num_copies): Use vect_get_num_vectors.
12112 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12113 to an unsigned int *.
12114 (vect_analyze_data_refs): Change min_vf from an int * to a
12116 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12117 than an unsigned HOST_WIDE_INT.
12118 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12119 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12120 to an unsigned int *.
12121 (vect_analyze_data_ref_dependences): Likewise.
12122 (vect_compute_data_ref_alignment): Handle polynomial vf.
12123 (vect_enhance_data_refs_alignment): Likewise.
12124 (vect_prune_runtime_alias_test_list): Likewise.
12125 (vect_shift_permute_load_chain): Likewise.
12126 (vect_supportable_dr_alignment): Likewise.
12127 (dependence_distance_ge_vf): Take the vectorization factor as a
12128 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12129 (vect_analyze_data_refs): Change min_vf from an int * to a
12131 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12132 vfm1 as a poly_uint64 rather than an int. Make the same change
12133 for the returned bound_scalar.
12134 (vect_gen_vector_loop_niters): Handle polynomial vf.
12135 (vect_do_peeling): Likewise. Update call to
12136 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12137 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12139 * tree-vect-loop.c (vect_determine_vectorization_factor)
12140 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12141 (vect_get_known_peeling_cost): Likewise.
12142 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12143 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12144 (vect_transform_loop): Likewise. Use the lowest possible VF when
12145 updating the upper bounds of the loop.
12146 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12147 rather than an int.
12148 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12149 polynomial unroll factors.
12150 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12151 (vect_make_slp_decision): Likewise.
12152 (vect_supported_load_permutation_p): Likewise, and polynomial
12154 (vect_analyze_slp_cost): Handle polynomial vf.
12155 (vect_slp_analyze_node_operations): Likewise.
12156 (vect_slp_analyze_bb_1): Likewise.
12157 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12158 than an unsigned HOST_WIDE_INT.
12159 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12160 (vectorizable_load): Handle polynomial vf.
12161 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12163 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12166 Alan Hayward <alan.hayward@arm.com>
12167 David Sherwood <david.sherwood@arm.com>
12169 * match.pd: Handle bit operations involving three constants
12170 and try to fold one pair.
12172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12174 * tree-vect-loop-manip.c: Include gimple-fold.h.
12175 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12176 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12177 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12178 Add a path that uses a step of VF instead of 1, but disable it
12180 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12181 and niters_no_overflow parameters. Update calls to
12182 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12183 Create a new SSA name if the latter choses to use a ste other
12184 than zero, and return it via niters_vector_mult_vf_var.
12185 * tree-vect-loop.c (vect_transform_loop): Update calls to
12186 vect_do_peeling, vect_gen_vector_loop_niters and
12187 slpeel_make_loop_iterate_ntimes.
12188 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12189 (vect_gen_vector_loop_niters): Update declarations after above changes.
12191 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12193 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12194 128-bit round to integer instructions.
12195 (ceil<mode>2): Likewise.
12196 (btrunc<mode>2): Likewise.
12197 (round<mode>2): Likewise.
12199 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12201 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12202 unaligned VSX load/store on P8/P9.
12203 (expand_block_clear): Allow the use of unaligned VSX
12204 load/store on P8/P9.
12206 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12208 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12210 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12211 swap associated with both a load and a store.
12213 2018-01-02 Andrew Waterman <andrew@sifive.com>
12215 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12216 * config/riscv/riscv.md (clear_cache): Use it.
12218 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12220 * web.c: Remove out-of-date comment.
12222 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12224 * expr.c (fixup_args_size_notes): Check that any existing
12225 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12226 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12227 (emit_single_push_insn): ...here.
12229 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12231 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12232 (const_vector_encoded_nelts): New function.
12233 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12234 (const_vector_int_elt, const_vector_elt): Declare.
12235 * emit-rtl.c (const_vector_int_elt_1): New function.
12236 (const_vector_elt): Likewise.
12237 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12238 of CONST_VECTOR_ELT.
12240 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12242 * expr.c: Include rtx-vector-builder.h.
12243 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12244 directly on the tree encoding.
12245 (const_vector_from_tree): Likewise.
12246 * optabs.c: Include rtx-vector-builder.h.
12247 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12248 sequence of "u" values.
12249 * vec-perm-indices.c: Include rtx-vector-builder.h.
12250 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12251 directly on the vec_perm_indices encoding.
12253 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12255 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12256 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12257 * rtx-vector-builder.h: New file.
12258 * rtx-vector-builder.c: Likewise.
12259 * rtl.h (rtx_def::u2): Add a const_vector field.
12260 (CONST_VECTOR_NPATTERNS): New macro.
12261 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12262 (CONST_VECTOR_DUPLICATE_P): Likewise.
12263 (CONST_VECTOR_STEPPED_P): Likewise.
12264 (CONST_VECTOR_ENCODED_ELT): Likewise.
12265 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12266 (unwrap_const_vec_duplicate): Likewise.
12267 (const_vec_series_p): Check for a non-duplicated vector encoding.
12268 Say that the function only returns true for integer vectors.
12269 * emit-rtl.c: Include rtx-vector-builder.h.
12270 (gen_const_vec_duplicate_1): Delete.
12271 (gen_const_vector): Call gen_const_vec_duplicate instead of
12272 gen_const_vec_duplicate_1.
12273 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12274 (gen_const_vec_duplicate): Use rtx_vector_builder.
12275 (gen_const_vec_series): Likewise.
12276 (gen_rtx_CONST_VECTOR): Likewise.
12277 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12278 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12279 Build a new vector rather than modifying a CONST_VECTOR in-place.
12280 (handle_special_swappables): Update call accordingly.
12281 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12282 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12283 Build a new vector rather than modifying a CONST_VECTOR in-place.
12284 (handle_special_swappables): Update call accordingly.
12286 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12288 * simplify-rtx.c (simplify_const_binary_operation): Use
12289 CONST_VECTOR_ELT instead of XVECEXP.
12291 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12293 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12294 the selector elements to be different from the data elements
12295 if the selector is a VECTOR_CST.
12296 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12297 ssizetype for the selector.
12299 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12301 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12302 before testing each element individually.
12303 * tree-vect-generic.c (lower_vec_perm): Likewise.
12305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12307 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12308 * selftest-run-tests.c (selftest::run_tests): Call it.
12309 * vector-builder.h (vector_builder::operator ==): New function.
12310 (vector_builder::operator !=): Likewise.
12311 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12312 (vec_perm_indices::all_from_input_p): New function.
12313 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12314 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12315 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12316 instead of reading the VECTOR_CST directly. Detect whether both
12317 vector inputs are the same before constructing the vec_perm_indices,
12318 and update the number of inputs argument accordingly. Use the
12319 utility functions added above. Only construct sel2 if we need to.
12321 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12323 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12324 the broadcast of the low byte.
12325 (expand_mult_highpart): Use an explicit encoding for the permutes.
12326 * optabs-query.c (can_mult_highpart_p): Likewise.
12327 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12328 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12329 (vectorizable_bswap): Likewise.
12330 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12331 explicit encoding for the power-of-2 permutes.
12332 (vect_permute_store_chain): Likewise.
12333 (vect_grouped_load_supported): Likewise.
12334 (vect_permute_load_chain): Likewise.
12336 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12338 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12339 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12340 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12341 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12342 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12343 (vect_gen_perm_mask_any): Likewise.
12345 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12347 * int-vector-builder.h: New file.
12348 * vec-perm-indices.h: Include int-vector-builder.h.
12349 (vec_perm_indices): Redefine as an int_vector_builder.
12350 (auto_vec_perm_indices): Delete.
12351 (vec_perm_builder): Redefine as a stand-alone class.
12352 (vec_perm_indices::vec_perm_indices): New function.
12353 (vec_perm_indices::clamp): Likewise.
12354 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12355 (vec_perm_indices::new_vector): New function.
12356 (vec_perm_indices::new_expanded_vector): Update for new
12357 vec_perm_indices class.
12358 (vec_perm_indices::rotate_inputs): New function.
12359 (vec_perm_indices::all_in_range_p): Operate directly on the
12360 encoded form, without computing elided elements.
12361 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12362 encoding. Update for new vec_perm_indices class.
12363 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12364 the given vec_perm_builder.
12365 (expand_vec_perm_var): Update vec_perm_builder constructor.
12366 (expand_mult_highpart): Use vec_perm_builder instead of
12367 auto_vec_perm_indices.
12368 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12369 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12370 or double series encoding as appropriate.
12371 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12372 vec_perm_indices instead of auto_vec_perm_indices.
12373 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12374 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12375 (vect_permute_store_chain): Likewise.
12376 (vect_grouped_load_supported): Likewise.
12377 (vect_permute_load_chain): Likewise.
12378 (vect_shift_permute_load_chain): Likewise.
12379 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12380 (vect_transform_slp_perm_load): Likewise.
12381 (vect_schedule_slp_instance): Likewise.
12382 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12383 (vectorizable_mask_load_store): Likewise.
12384 (vectorizable_bswap): Likewise.
12385 (vectorizable_store): Likewise.
12386 (vectorizable_load): Likewise.
12387 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12388 vec_perm_indices instead of auto_vec_perm_indices. Use
12389 tree_to_vec_perm_builder to read the vector from a tree.
12390 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12391 vec_perm_builder instead of a vec_perm_indices.
12392 (have_whole_vector_shift): Use vec_perm_builder and
12393 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12394 truncation to calc_vec_perm_mask_for_shift.
12395 (vect_create_epilog_for_reduction): Likewise.
12396 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12397 from auto_vec_perm_indices to vec_perm_indices.
12398 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12399 instead of changing individual elements.
12400 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12401 the vector in d.perm.
12402 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12403 from auto_vec_perm_indices to vec_perm_indices.
12404 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12405 instead of changing individual elements.
12406 (arm_vectorize_vec_perm_const): Use new_vector to install
12407 the vector in d.perm.
12408 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12409 Update vec_perm_builder constructor.
12410 (rs6000_expand_interleave): Likewise.
12411 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12412 (rs6000_expand_interleave): Likewise.
12414 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12416 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12417 to qimode could truncate the indices.
12418 * optabs.c (expand_vec_perm_var): Likewise.
12420 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12422 * Makefile.in (OBJS): Add vec-perm-indices.o.
12423 * vec-perm-indices.h: New file.
12424 * vec-perm-indices.c: Likewise.
12425 * target.h (vec_perm_indices): Replace with a forward class
12427 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12428 * optabs.h: Include vec-perm-indices.h.
12429 (expand_vec_perm): Delete.
12430 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12431 (expand_vec_perm_const): Declare.
12432 * target.def (vec_perm_const_ok): Replace with...
12433 (vec_perm_const): ...this new hook.
12434 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12435 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12436 * doc/tm.texi: Regenerate.
12437 * optabs.def (vec_perm_const): Delete.
12438 * doc/md.texi (vec_perm_const): Likewise.
12439 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12440 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12441 expand_vec_perm for constant permutation vectors. Assert that
12442 the mode of variable permutation vectors is the integer equivalent
12443 of the mode that is being permuted.
12444 * optabs-query.h (selector_fits_mode_p): Declare.
12445 * optabs-query.c: Include vec-perm-indices.h.
12446 (selector_fits_mode_p): New function.
12447 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12448 is defined, instead of checking whether the vec_perm_const_optab
12449 exists. Use targetm.vectorize.vec_perm_const instead of
12450 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12451 fit in the vector mode before using a variable permute.
12452 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12453 vec_perm_indices instead of an rtx.
12454 (expand_vec_perm): Replace with...
12455 (expand_vec_perm_const): ...this new function. Take the selector
12456 as a vec_perm_indices rather than an rtx. Also take the mode of
12457 the selector. Update call to shift_amt_for_vec_perm_mask.
12458 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12459 Use vec_perm_indices::new_expanded_vector to expand the original
12460 selector into bytes. Check whether the indices fit in the vector
12461 mode before using a variable permute.
12462 (expand_vec_perm_var): Make global.
12463 (expand_mult_highpart): Use expand_vec_perm_const.
12464 * fold-const.c: Includes vec-perm-indices.h.
12465 * tree-ssa-forwprop.c: Likewise.
12466 * tree-vect-data-refs.c: Likewise.
12467 * tree-vect-generic.c: Likewise.
12468 * tree-vect-loop.c: Likewise.
12469 * tree-vect-slp.c: Likewise.
12470 * tree-vect-stmts.c: Likewise.
12471 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12473 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12474 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12475 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12476 (aarch64_vectorize_vec_perm_const): ...this new function.
12477 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12478 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12479 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12480 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12481 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12482 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12483 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12485 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12486 check for NEON modes.
12487 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12488 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12489 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12490 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12492 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12493 the old VEC_PERM_CONST conditions.
12494 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12495 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12496 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12497 (ia64_vectorize_vec_perm_const_ok): Merge into...
12498 (ia64_vectorize_vec_perm_const): ...this new function.
12499 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12500 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12501 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12502 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12503 * config/mips/mips.c (mips_expand_vec_perm_const)
12504 (mips_vectorize_vec_perm_const_ok): Merge into...
12505 (mips_vectorize_vec_perm_const): ...this new function.
12506 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12507 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12508 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12509 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12510 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12511 (rs6000_expand_vec_perm_const): Delete.
12512 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12514 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12515 (altivec_expand_vec_perm_const_le): Take each operand individually.
12516 Operate on constant selectors rather than rtxes.
12517 (altivec_expand_vec_perm_const): Likewise. Update call to
12518 altivec_expand_vec_perm_const_le.
12519 (rs6000_expand_vec_perm_const): Delete.
12520 (rs6000_vectorize_vec_perm_const_ok): Delete.
12521 (rs6000_vectorize_vec_perm_const): New function.
12522 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12523 an element count and rtx array.
12524 (rs6000_expand_extract_even): Update call accordingly.
12525 (rs6000_expand_interleave): Likewise.
12526 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12527 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12528 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12529 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12530 (rs6000_expand_vec_perm_const): Delete.
12531 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12532 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12533 (altivec_expand_vec_perm_const_le): Take each operand individually.
12534 Operate on constant selectors rather than rtxes.
12535 (altivec_expand_vec_perm_const): Likewise. Update call to
12536 altivec_expand_vec_perm_const_le.
12537 (rs6000_expand_vec_perm_const): Delete.
12538 (rs6000_vectorize_vec_perm_const_ok): Delete.
12539 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12540 reference to the SPE evmerge intructions.
12541 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12542 an element count and rtx array.
12543 (rs6000_expand_extract_even): Update call accordingly.
12544 (rs6000_expand_interleave): Likewise.
12545 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12546 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12548 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12550 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12552 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12553 vector mode and that that mode matches the mode of the data
12555 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12556 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12557 directly using expand_vec_perm_1 when forcing selectors into
12559 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12561 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12563 * optabs-query.h (can_vec_perm_p): Delete.
12564 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12565 * optabs-query.c (can_vec_perm_p): Split into...
12566 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12567 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12568 particular selector is valid.
12569 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12570 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12571 (vect_grouped_load_supported): Likewise.
12572 (vect_shift_permute_load_chain): Likewise.
12573 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12574 (vect_transform_slp_perm_load): Likewise.
12575 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12576 (vectorizable_bswap): Likewise.
12577 (vect_gen_perm_mask_checked): Likewise.
12578 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12579 implementations of variable permutation vectors into account
12580 when deciding which selector to use.
12581 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12582 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12583 with a false third argument.
12584 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12585 to test whether the constant selector is valid and can_vec_perm_var_p
12586 to test whether a variable selector is valid.
12588 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12590 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12591 * optabs-query.c (can_vec_perm_p): Likewise.
12592 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12593 instead of vec_perm_indices.
12594 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12595 (vect_gen_perm_mask_checked): Likewise,
12596 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12597 (vect_gen_perm_mask_checked): Likewise,
12599 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12601 * optabs-query.h (qimode_for_vec_perm): Declare.
12602 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12603 (qimode_for_vec_perm): ...this new function.
12604 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12606 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12608 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12609 does not have a conditional at the top.
12611 2018-01-02 Richard Biener <rguenther@suse.de>
12613 * ipa-inline.c (big_speedup_p): Fix expression.
12615 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12618 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12621 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12625 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12626 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12627 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12628 cond_taken_branch_cost 3->4.
12630 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12632 PR tree-optimization/83581
12633 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12634 TODO_cleanup_cfg if any changes have been made.
12636 PR middle-end/83608
12637 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12638 convert_modes if target mode has the right side, but different mode
12641 PR middle-end/83609
12642 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12643 last argument when extracting from CONCAT. If either from_real or
12644 from_imag is NULL, use expansion through memory. If result is not
12645 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12646 the parts directly to inner mode, if even that fails, use expansion
12649 PR middle-end/83623
12650 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12651 check for bswap in mode rather than HImode and use that in expand_unop
12654 Copyright (C) 2018 Free Software Foundation, Inc.
12656 Copying and distribution of this file, with or without modification,
12657 are permitted in any medium without royalty provided the copyright
12658 notice and this notice are preserved.