1 2018-05-16 Richard Biener <rguenther@suse.de>
3 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6 (check_name): New callback.
7 (dse_classify_store): Track cycles via a visited bitmap of PHI
8 defs and simplify handling of in-loop and across loop dead stores
9 and properly fail for loop-variant refs. Handle byte-tracking with
10 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
13 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
15 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
16 (vect_get_mask_type_for_stmt): Likewise.
17 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
20 to determine the statement's vector type and the vector type that
21 should be used for calculating nunits. Deal with cases in which
22 the type has to be deferred.
23 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
24 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
25 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
26 (vect_determine_vf_for_stmt): New functions, split out from...
27 (vect_determine_vectorization_factor): ...here.
28 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
29 (vect_get_mask_type_for_stmt): New functions, split out from
30 vect_determine_vectorization_factor.
32 2018-05-16 Richard Biener <rguenther@suse.de>
34 * tree-cfg.c (verify_gimple_assign_ternary): Properly
35 verify the [VEC_]COND_EXPR embedded comparison.
37 2018-05-15 Martin Sebor <msebor@redhat.com>
39 PR tree-optimization/85753
40 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
41 RECORD_TYPE in addition to ARRAY_TYPE.
43 2018-05-15 Martin Sebor <msebor@redhat.com>
46 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
48 2018-05-15 Richard Biener <rguenther@suse.de>
50 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
51 add by_clobber_p one. Change algorithm to collect all defs
52 representing uses we need to walk and try reducing them to
53 a single one before failing.
54 (dse_dom_walker::dse_optimize_stmt): Adjust.
56 2018-05-13 Mark Wielaard <mark@klomp.org>
58 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
59 (size_of_loc_descr): Likewise.
60 (output_loc_operands): Likewise.
61 (output_loc_operands_raw): Likewise.
62 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
63 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
64 (hash_loc_operands): Likewise.
65 (compare_loc_operands): Likewise.
67 2018-05-14 Mark Wielaard <mark@klomp.org>
69 * dwarf2out.c (count_index_addrs): New function.
70 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
72 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
74 PR tree-optimization/83648
75 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
76 return value as malloc candidate.
78 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
81 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
82 as true in call to suggest_attribute.
84 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
86 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
89 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
91 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
93 (aarch64_crypto_aesd_fused): Likewise.
95 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
97 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
98 (movsi_aarch64): Likewise.
99 (load_pairsi): Likewise.
100 (load_pairdi): Likewise.
101 (store_pairsi): Likewise.
102 (store_pairdi): Likewise.
103 (load_pairsf): Likewise.
104 (load_pairdf): Likewise.
105 (store_pairsf): Likewise.
106 (store_pairdf): Likewise.
107 (zero_extend): Likewise.
108 (trunc): Swap alternatives.
109 (fcvt_target): Add '?' to prefer w over r.
111 2018-05-14 Jakub Jelinek <jakub@redhat.com>
114 * config/i386/i386.md: Disallow non-commutative arithmetics in
115 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
116 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
117 in the peephole2 before it.
119 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
121 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
122 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
123 (ix86_handle_option): Handle -mcldemote.
124 * config.gcc: New header.
125 * config/i386/cldemoteintrin.h: New file.
126 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
127 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
129 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
130 OPTION_MASK_ISA_CLDEMOTE.
131 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
132 (ix86_valid_target_attribute_inner_p): Ditto.
133 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
134 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
135 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
136 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
137 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
139 * config/i386/i386.opt: Add -mcldemote.
140 * config/i386/x86intrin.h: New header.
141 * doc/invoke.texi: Add -mcldemote.
143 2018-05-14 Richard Biener <rguenther@suse.de>
145 * doc/match-and-simplify.texi: Adjust :s documentation.
147 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
149 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
150 intended memcpy size.
151 (REORDER_45): Likewise.
153 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
156 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
157 * vec.c (qsort_chk): Use gcc_qsort.
158 * Makefile.in (OBJS-libcommon): Add sort.o.
159 (build/sort.o): New target. Use it...
160 (BUILD_RTL): ... here, and...
161 (build/gencfn-macros): ... here, and...
162 (build/genmatch): ... here.
164 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
165 Chung-Ju Wu <jasonwucj@gmail.com>
167 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
168 * config/nds32/nds32-graywolf.md: New file.
169 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
170 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
172 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
173 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
174 * config/nds32/nds32.md (pipeline_model): Add graywolf.
175 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
176 * config/nds32/pipelines.md: Include n15 settings.
178 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
179 Chung-Ju Wu <jasonwucj@gmail.com>
181 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
182 * config/nds32/nds32-n13.md: New file.
183 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
184 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
186 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
187 * config/nds32/nds32.md (pipeline_model): Add n13.
188 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
189 * config/nds32/pipelines.md: Include n13 settings.
191 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
192 Chung-Ju Wu <jasonwucj@gmail.com>
194 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
195 * config/nds32/nds32-n10.md: New file.
196 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
197 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
199 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
200 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
201 * config/nds32/nds32.md (pipeline_model): Add n10.
202 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
203 * config/nds32/pipelines.md: Include n10 settings.
205 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
206 Kito Cheng <kito.cheng@gmail.com>
207 Chung-Ju Wu <jasonwucj@gmail.com>
209 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
210 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
211 Add enum values for DSP extension instructions.
212 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
214 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
215 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
217 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
218 * config/nds32/nds32-dspext.md: New file for DSP implementation.
219 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
220 * config/nds32/nds32-intrinsic.md: Likewise.
221 * config/nds32/nds32_intrinsic.h: Likewise.
222 * config/nds32/nds32-md-auxiliary.c: Likewise.
223 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
224 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
225 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
226 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
227 * config/nds32/nds32-protos.h: New declarations for DSP extension.
228 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
229 TYPE_DMAC in switch statement.
230 * config/nds32/nds32.c: New checking and implementation for DSP
231 extension instructions.
232 * config/nds32/nds32.h: Likewise.
233 * config/nds32/nds32.md: Likewise.
234 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
235 * config/nds32/predicates.md: Implement new predicates for DSP
238 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
240 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
241 Reformat alternatives and attributes so it is easier to identify
242 which constraints/attributes go with which instruction.
243 (mov<mode>_hardfloat32, FMOVE64): Likewise.
244 (mov<mode>_softfloat32, FMOVE64): Likewise.
245 (mov<mode>_hardfloat64, FMOVE64): Likewise.
246 (mov<mode>_softfloat64, FMOVE64): Likewise.
248 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
250 * doc/extend.texi (PowerPC Built-in Functions): Rename this
252 (Basic PowerPC Built-in Functions): The new name of the
253 subsection previously known as "PowerPC Built-in Functions".
254 (Basic PowerPC Built-in Functions Available on all Configurations):
256 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
257 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
258 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
259 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
261 2018-05-11 Martin Jambor <mjambor@suse.cz>
264 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
267 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
270 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
272 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
274 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
275 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
276 (ix86_handle_option): Handle -mwaitpkg.
277 * config.gcc: New header.
278 * config/i386/cpuid.h (bit_WAITPKG): New bit.
279 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
280 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
282 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
283 OPTION_MASK_ISA_WAITPKG.
284 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
285 (ix86_option_override_internal): Add PTA_WAITPKG.
286 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
287 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
289 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
290 __builtin_ia32_umwait and __builtin_ia32_tpause.
291 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
292 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
293 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
294 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
295 UNSPECV_TPAUSE): New.
296 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
297 * config/i386/i386.opt: Add -mwaitpkg.
298 * config/i386/waitpkgintrin.h: New file.
299 * config/i386/x86intrin.h: New header.
300 * doc/invoke.texi: Add -mwaitpkg.
302 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
305 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
307 (cortex-m0): Use armv6s-m isa.
308 (cortex-m0plus): Likewise.
309 (cortex-m1): Likewise.
310 (cortex-m0.small-multiply): Likewise.
311 (cortex-m0plus.small-multiply): Likewise.
312 (cortex-m1.small-multiply): Likewise.
314 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
315 Jakub Jelinek <jakub@redhat.com>
317 PR tree-optimization/85692
318 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
319 source permute as well.
321 2018-05-11 Martin Liska <mliska@suse.cz>
324 * doc/extend.texi: Document LLVM style format for no_sanitize
327 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
329 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
330 mode_supports_vsx_dform_quad to mode_supports_dq_form.
331 (mode_supports_vsx_dform_quad): Likewise.
332 (mode_supports_vmx_dform): Move these functions to be next to the
333 other mode_supports functions.
334 (mode_supports_dq_form): Likewise.
335 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
336 mode_supports_dq_form.
337 (reg_offset_addressing_ok_p): Likewise.
338 (offsettable_ok_by_alignment): Likewise.
339 (rs6000_legitimate_offset_address_p): Likewise.
340 (legitimate_lo_sum_address_p): Likewise.
341 (rs6000_legitimize_address): Likewise.
342 (rs6000_legitimize_reload_address): Likewise.
343 (rs6000_secondary_reload_inner): Likewise.
344 (rs6000_preferred_reload_class): Likewise.
345 (rs6000_output_move_128bit): Likewise.
347 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
349 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
350 Generate SImode target register for null target.
351 <case IX86_BUILTIN_XGETBV>: Ditto.
352 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
353 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
355 2018-05-10 Carl Love <cel@us.ibm.com>
357 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
358 dcbtt and dcbtstt if operands[2] is 0.
360 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
363 * config/i386/sse.md (usadv64qi): New expander.
365 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
367 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
368 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
369 -maltivec=be support.
370 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
371 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
372 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
373 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
374 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
375 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
376 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
377 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
378 altivec_vsumsws): Adjust.
379 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
380 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
381 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
383 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
384 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
385 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
386 (altivec_lve<VI_char>x): Delete expand.
387 (*altivec_lve<VI_char>x_internal): Rename to...
388 (altivec_lve<VI_char>x): ... this.
389 (altivec_lvxl_<mode>): Delete expand.
390 (*altivec_lvxl_<mode>_internal): Rename to ...
391 (altivec_lvxl_<mode>): ... this.
392 (altivec_stvxl_<mode>): Delete expand.
393 (*altivec_stvxl_<mode>_internal): Rename to ...
394 (altivec_stvxl_<mode>): ... this.
395 (altivec_stve<VI_char>x): Delete expand.
396 (*altivec_stve<VI_char>x_internal): Rename to ...
397 (altivec_stve<VI_char>x): ... this.
398 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
399 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
400 reduc_plus_scal_<mode>): Adjust.
401 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
403 (rs6000_cpu_cpp_builtins): Adjust.
404 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
405 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
406 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
407 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
408 -maltivec=be support.
409 (rs6000_split_vec_extract_var): Adjust.
410 (rs6000_split_v4si_init): Adjust.
411 (swap_selector_for_mode): Delete.
412 (altivec_expand_lvx_be, altivec_expand_stvx_be,
413 altivec_expand_stvex_be): Delete.
414 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
415 -maltivec=be support.
416 (rs6000_gimple_fold_builtin): Ditto.
417 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
419 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
420 (TARGET_DIRECT_MOVE_64BIT): Adjust.
421 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
422 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
423 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
424 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
425 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
426 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
427 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
428 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
429 anonymous split): Adjust.
430 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
431 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
433 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
435 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
436 when --with-gxx-include-dir is also specified.
437 * configure: Regenerate.
439 2018-05-09 Jim Wilson <jimw@sifive.com>
442 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
443 * config/riscv/t-withmultilib: New.
444 * config/riscv/withmultilib.h: New.
445 * doc/install.texi: Document RISC-V --with-multilib-list support.
447 2018-05-09 Richard Biener <rguenther@suse.de>
449 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
451 (vect_bb_vectorization_profitable_p): Adjust. Compute
452 actual scalar cost using the cost vector and the add_stmt_cost
455 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
457 PR rtl-optimization/85645
458 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
459 in the REG_CFA_REGISTER note for LR, don't leave it empty.
461 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
463 PR rtl-optimization/85645
464 * shrink-wrap.c (spread_components): Return a boolean saying if
465 anything was changed.
466 (try_shrink_wrapping_separate): Iterate spread_components until
467 nothing changes anymore.
469 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
471 PR rtl-optimization/85645
472 * regrename.c (build_def_use): Also kill the chains that include the
473 destination of a REG_CFA_REGISTER note.
475 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
477 PR rtl-optimization/85645
478 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
479 insn that has a REG_CFA_REGISTER note.
481 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
483 * cfgexpand.c (expand_clobber): New function.
484 (expand_gimple_stmt_1): Use it.
485 * tree-vect-stmts.c (vect_clobber_variable): New function,
487 (vectorizable_simd_clone_call): ...here.
488 (vectorizable_store): Emit a clobber either side of an
489 IFN_STORE_LANES sequence.
490 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
492 2018-05-09 Tom de Vries <tom@codesourcery.com>
495 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
496 (define_insn "trap_if_false"): Add exit after trap.
498 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
500 PR rtl-optimization/85638
501 * bb-reorder.c: Include common/common-target.h.
502 (create_forwarder_block): New function extracted from...
503 (fix_up_crossing_landing_pad): ...here. Rename into...
504 (dw2_fix_up_crossing_landing_pad): ...this.
505 (sjlj_fix_up_crossing_landing_pad): New function.
506 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
507 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
508 from both partitions and exit the loop after one iteration.
510 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
513 * doc/extend.texi (PowerPC Built-in Functions): Rename this
515 (Basic PowerPC Built-in Functions): The new name of the
516 subsection previously known as "PowerPC Built-in Functions".
517 (Basic PowerPC Built-in Functions Available on all Configurations):
519 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
521 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
523 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
525 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
528 2018-05-08 Jim Wilson <jimw@sifive.com>
530 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
531 (LD_EMUL_SUFFIX): New.
534 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
536 * doc/extend.texi (PowerPC Built-in Functions): Rename this
538 (Basic PowerPC Built-in Functions): The new name of the
539 subsection previously known as "PowerPC Built-in Functions".
540 (Basic PowerPC Built-in Functions Available on all Configurations):
542 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
544 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
546 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
548 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
551 2018-05-08 Jakub Jelinek <jakub@redhat.com>
554 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
555 after cmpelim optimization.
557 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
559 * config.gcc: Support "goldmont".
560 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
561 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
563 * config/i386/i386.c (m_GOLDMONT): Define.
564 (processor_target_table): Add "goldmont".
565 (PTA_GOLDMONT): Define.
566 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
567 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
568 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
569 (fold_builtin_cpu): Add "goldmont".
570 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
571 (ix86_option_override_internal): Add "goldmont".
572 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
573 (processor_type): Add PROCESSOR_GOLDMONT.
574 * config/i386/i386.md: Add CPU "glm".
575 * config/i386/glm.md: New file.
576 * config/i386/x86-tune.def: Add m_GOLDMONT.
577 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
579 2018-05-08 Jakub Jelinek <jakub@redhat.com>
582 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
584 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
585 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
586 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
589 * config/i386/i386.c (ix86_fold_builtin): Handle
590 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
593 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
594 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
596 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
599 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
600 (check_arch): Likewise.
601 (check_fpu): Return the result rather than printing it.
602 (end arch): Fix operator precedence.
604 (END): Print the result from check_fpu.
606 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
607 Alan Hayward <alan.hayward@arm.com>
608 David Sherwood <david.sherwood@arm.com>
610 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
611 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
612 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
613 (*fcmuo<mode>_and): New patterns.
615 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
617 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
618 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
619 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
620 (cmp_op, sve_imm_con): New code attributes.
621 (SVE_COND_INT_CMP, imm_con): Delete.
622 (cmp_op): Remove above unspecs from int attribute.
623 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
625 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
626 comparison-specific unspecs.
627 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
628 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
629 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
630 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
631 (*vec_fcm<cmp_op><mode>): Rename to...
632 (*fcm<cmp_op><mode>): ...this and adjust likewise.
633 (*vec_fcmuo<mode>): Rename to...
634 (*fcmuo<mode>): ...this and adjust likewise.
635 (*pred_fcm<cmp_op><mode>): New pattern.
636 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
637 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
639 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
641 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
642 (aarch64_emit_sve_predicated_cond): New function.
643 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
644 (aarch64_emit_unspec_cond_or): Replace with...
645 (aarch64_emit_sve_or_conds): ...this new function. Use
646 aarch64_emit_sve_ptrue_op for the individual comparisons and
647 aarch64_emit_binop to OR them together.
648 (aarch64_emit_inverted_unspec_cond): Replace with...
649 (aarch64_emit_sve_inverted_cond): ...this new function. Use
650 aarch64_emit_sve_ptrue_op for the comparison and
651 aarch64_emit_unop to invert the result.
652 (aarch64_expand_sve_vec_cmp_float): Update after the above
653 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
655 2018-05-07 Nathan Sidwell <nathan@acm.org>
657 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
658 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
659 (Backwards Compatibility): Likewise.
661 2018-05-07 Luis Machado <luis.machado@linaro.org>
665 2018-05-07 Luis Machado <luis.machado@linaro.org>
667 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
668 <prefetch_dynamic_strides>: New const bool field.
669 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
670 prefetch_dynamic_strides.
671 (exynosm1_prefetch_tune): Likewise.
672 (thunderxt88_prefetch_tune): Likewise.
673 (thunderx_prefetch_tune): Likewise.
674 (thunderx2t99_prefetch_tune): Likewise.
675 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
676 (aarch64_override_options_internal): Update to set
677 PARAM_PREFETCH_DYNAMIC_STRIDES.
678 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
679 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
680 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
681 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
682 prefetch-dynamic-strides setting.
684 2018-05-07 Luis Machado <luis.machado@linaro.org>
686 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
687 <minimum_stride>: New const int field.
688 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
689 minimum_stride field.
690 (exynosm1_prefetch_tune): Likewise.
691 (thunderxt88_prefetch_tune): Likewise.
692 (thunderx_prefetch_tune): Likewise.
693 (thunderx2t99_prefetch_tune): Likewise.
694 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
695 (aarch64_override_options_internal): Update to set
696 PARAM_PREFETCH_MINIMUM_STRIDE.
697 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
698 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
699 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
700 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
701 stride is constant and is below the minimum stride threshold.
703 2018-05-07 Luis Machado <luis.machado@linaro.org>
705 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
708 2018-05-07 Luis Machado <luis.machado@linaro.org>
710 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
711 <prefetch_dynamic_strides>: New const bool field.
712 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
713 prefetch_dynamic_strides.
714 (exynosm1_prefetch_tune): Likewise.
715 (thunderxt88_prefetch_tune): Likewise.
716 (thunderx_prefetch_tune): Likewise.
717 (thunderx2t99_prefetch_tune): Likewise.
718 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
719 (aarch64_override_options_internal): Update to set
720 PARAM_PREFETCH_DYNAMIC_STRIDES.
721 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
722 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
723 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
724 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
725 prefetch-dynamic-strides setting.
727 2018-05-07 Luis Machado <luis.machado@linaro.org>
729 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
730 <minimum_stride>: New const int field.
731 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
732 minimum_stride field.
733 (exynosm1_prefetch_tune): Likewise.
734 (thunderxt88_prefetch_tune): Likewise.
735 (thunderx_prefetch_tune): Likewise.
736 (thunderx2t99_prefetch_tune): Likewise.
737 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
738 (aarch64_override_options_internal): Update to set
739 PARAM_PREFETCH_MINIMUM_STRIDE.
740 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
741 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
742 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
743 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
744 stride is constant and is below the minimum stride threshold.
746 2018-05-06 Jakub Jelinek <jakub@redhat.com>
749 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
750 the type is addressable. Don't force op into register if it has
753 2018-05-05 Roland McGrath <mcgrathr@google.com>
756 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
757 any section for which we don't know a specific type it should have,
758 regardless of name. Previously this was done only for the exact
759 names ".init_array", ".fini_array", and ".preinit_array".
760 (default_elf_asm_named_section): Add comment about
761 relationship with default_section_type_flags and SECTION_NOTYPE.
762 (get_section): Don't consider it a type conflict if one side has
763 SECTION_NOTYPE and the other doesn't, as long as neither has the
764 SECTION_BSS et al used in the default_section_type_flags logic.
766 2018-05-05 Tom de Vries <tom@codesourcery.com>
769 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
770 (workaround_barsyncs): New function.
771 (nvptx_reorg): Use workaround_barsyncs.
772 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
773 (define_expand "nvptx_membar_cta"): New define_expand.
774 (define_insn "*nvptx_membar_cta"): New insn.
776 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
778 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
779 To improve optimization opportunities.
780 * builtin-types.def: The new needed builtin types for the above.
782 2018-05-04 Richard Biener <rguenther@suse.de>
784 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
785 * gimple-ssa-store-merging.c
786 (imm_store_chain_info::output_merged_store): Remove redundant create,
787 release split_store vector contents on failure.
788 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
789 scalar stmt vector on cache hit.
791 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
793 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
795 * config.gcc (powerpc-xilinx-eabi*): Remove.
796 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
798 (fusion_addis_mem_combo_load): Ditto.
799 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
801 (rs6000_cpu_cpp_builtins): Ditto.
802 * config/rs6000/rs6000-linux.c
803 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
804 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
805 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
807 (rs6000_setup_reg_addr_masks): Ditto.
808 (rs6000_init_hard_regno_mode_ok): Ditto.
809 (rs6000_option_override_internal): Ditto.
810 (legitimate_lo_sum_address_p): Ditto.
811 (rs6000_legitimize_address): Ditto.
812 (rs6000_legitimize_reload_address): Ditto.
813 (rs6000_legitimate_address_p): Ditto.
814 (abi_v4_pass_in_fpr): Ditto.
815 (setup_incoming_varargs): Ditto.
816 (rs6000_gimplify_va_arg): Ditto.
817 (rs6000_split_multireg_move): Ditto.
818 (rs6000_savres_strategy): Ditto.
819 (rs6000_emit_prologue_components): Ditto.
820 (rs6000_emit_epilogue_components): Ditto.
821 (rs6000_emit_prologue): Ditto.
822 (rs6000_emit_epilogue): Ditto.
823 (rs6000_elf_file_end): Ditto.
824 (rs6000_function_value): Ditto.
825 (rs6000_libcall_value): Ditto.
826 * config/rs6000/rs6000.h: Ditto.
827 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
828 (TARGET_MINMAX): ... this. New.
829 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
830 * config/rs6000/rs6000.md: Remove Xilinx FP support.
831 (*movsi_internal1_single): Delete.
832 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
833 mfpu=, mxilinx-fpu): Delete.
834 * config/rs6000/singlefp.h: Delete.
835 * config/rs6000/sysv4.h: Remove Xilinx FP support.
836 * config/rs6000/t-rs6000: Ditto.
837 * config/rs6000/t-xilinx: Delete.
838 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
839 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
840 (VStype_simple): Delete.
841 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
842 * config/rs6000/xfpu.h: Delete.
843 * config/rs6000/xfpu.md: Delete.
844 * config/rs6000/xilinx.h: Delete.
845 * config/rs6000/xilinx.opt: Delete.
846 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
847 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
849 2018-05-04 Tom de Vries <tom@codesourcery.com>
852 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
855 2018-05-04 Richard Biener <rguenther@suse.de>
858 * tree-complex.c (update_complex_assignment): We are always in SSA form.
859 (expand_complex_div_wide): Likewise.
860 (expand_complex_operations_1): Likewise.
861 (expand_complex_libcall): Preserve EH info of the original stmt.
862 (tree_lower_complex): Handle removed blocks.
863 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
864 on complex multiplication and division libcall builtins.
866 2018-05-04 Richard Biener <rguenther@suse.de>
869 * fold-const.c (negate_expr_p): Restrict negation of operand
870 zero of a division to when we know that can happen without
872 (fold_negate_expr_1): Likewise.
874 2018-05-04 Jakub Jelinek <jakub@redhat.com>
877 * real.h (real_nextafter): Declare.
878 * real.c (real_nextafter): New function.
879 * fold-const-call.c (fold_const_nextafter): New function.
880 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
882 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
883 even when arg1_mode is different from arg0_mode.
885 2018-05-03 Nathan Sidwell <nathan@acm.org>
887 * doc/extend.texi (Deprecated Features): Remove
889 (Backwards Compatibility): Likewise.
890 * doc/invoke.texi (C++ Language Options): Likewise.
891 (C++ Dialect Options): Likewise.
893 2018-05-03 Jakub Jelinek <jakub@redhat.com>
896 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
897 _mm512_mask_mullox_epi64): New intrinsics.
899 2018-05-03 Tom de Vries <tom@codesourcery.com>
902 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
903 dump files): Add offload-tree.
905 2018-05-03 Richard Biener <rguenther@suse.de>
907 PR tree-optimization/85615
908 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
909 to loops not nested in BBs loop father to avoid creating multi-entry
912 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
914 PR tree-optimization/70291
915 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
916 arguments. Change return type to tree. Emit libcall as a new
917 statement rather than replacing existing one when inplace_p is true.
918 (expand_complex_multiplication_components): New function.
919 (expand_complex_multiplication): Expand floating-point complex
920 multiplication using the above.
921 (expand_complex_division): Rename inner_type parameter to type.
922 Update expand_complex_libcall call-site.
923 (expand_complex_operations_1): Update expand_complex_multiplication
924 and expand_complex_division call-sites.
926 2018-05-02 Jakub Jelinek <jakub@redhat.com>
929 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
930 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
931 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
932 the highest significant bit of the shift count mask is clear. In
933 check whether and[sq]i3 is needed verify that all significant bits
934 of the shift count other than the highest are set.
936 2018-05-02 Tom de Vries <tom@codesourcery.com>
939 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
940 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
941 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
942 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
943 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
944 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
945 __builtin_goacc_parlevel_size.
947 2018-05-02 Richard Biener <rguenther@suse.de>
949 PR tree-optimization/85597
950 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
951 do not use split vect_get_vec_defs call but call vect_get_slp_defs
954 2018-05-02 Tom de Vries <tom@codesourcery.com>
957 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
958 dump files): Add ltrans-tree.
960 2018-05-02 Tom de Vries <tom@codesourcery.com>
963 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
964 dump files): Add wpa-ipa.
966 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
968 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
969 powerpc*-*-linux*paired* target.
970 * config/rs6000/750cl.h: Delete.
971 * config/rs6000/paired.h: Delete.
972 * config/rs6000/paired.md: Delete.
973 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
975 * config/rs6000/rs6000-builtin.def: Remove paired float support.
976 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
977 comment. Remove paired float support.
978 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
979 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
981 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
982 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
984 * config/rs6000/rs6000.c: Remove paired float support.
985 (paired_expand_vector_init, paired_expand_vector_move,
986 paired_emit_vector_compare, paired_emit_vector_cond_expr,
987 (paired_expand_lv_builtin, paired_expand_stv_builtin,
988 paired_expand_builtin, paired_expand_predicate_builtin,
989 paired_init_builtins): Delete.
990 * config/rs6000/rs6000.h: Remove paired float support.
991 * config/rs6000/rs6000.md: Remove paired float support.
992 (move_from_CR_ov_bit): Delete.
993 * config/rs6000/rs6000.opt (mpaired): Delete.
994 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
995 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
997 2018-05-02 Richard Biener <rguenther@suse.de>
1000 * gimplify.c (gimplify_save_expr): When in SSA form allow
1001 SAVE_EXPRs to compute to SSA vars.
1003 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1006 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1007 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1008 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1009 clobber operands[2], instead use a new pseudo. Formatting fixes.
1011 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1013 PR tree-optimization/85586
1014 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1015 exit early for statements in the same group if the accesses are
1018 2018-05-02 Tom de Vries <tom@codesourcery.com>
1021 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1024 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1026 PR tree-optimization/85143
1027 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1029 2018-05-01 Tom de Vries <tom@codesourcery.com>
1032 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1033 not found" error message.
1035 2018-05-01 Tom de Vries <tom@codesourcery.com>
1038 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1039 * vec.c (test_ordered_remove_if): New function.
1040 (vec_c_tests): Call test_ordered_remove_if.
1041 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1042 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1043 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1044 VEC_ORDERED_REMOVE_IF.
1046 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1048 PR tree-optimization/82665
1049 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1050 pointer subtraction where arguments come from a memchr call.
1052 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1054 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1055 --push-state --as-needed and --pop-state instead of --as-needed and
1056 --no-as-needed if ld supports it.
1057 * configure: Regenerated.
1060 * doc/install.texi2html: Replace _002d with - and _002a with * in
1061 generated html files using sed.
1063 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1066 * gcc-rich-location.c (blank_line_before_p): New function.
1067 (use_new_line): New function.
1068 (gcc_rich_location::add_fixit_insert_formatted): New function.
1069 * gcc-rich-location.h
1070 (gcc_rich_location::add_fixit_insert_formatted): New function.
1072 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1074 * selftest.c (assert_streq): Rename "expected" and "actual" to
1075 "val1" and "val2". Extend NULL-handling to cover both inputs
1076 symmetrically, while still requiring both to be non-NULL for a pass.
1077 * selftest.h (assert_streq): Rename "expected" and "actual" to
1079 (ASSERT_EQ): Likewise.
1080 (ASSERT_EQ_AT): Likewise.
1081 (ASSERT_KNOWN_EQ): Likewise.
1082 (ASSERT_KNOWN_EQ_AT): Likewise.
1083 (ASSERT_NE): Likewise.
1084 (ASSERT_MAYBE_NE): Likewise.
1085 (ASSERT_MAYBE_NE_AT): Likewise.
1086 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1087 the assertion to pass.
1088 (ASSERT_STREQ_AT): Likewise.
1090 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1092 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1093 interaction with -pie.
1095 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1097 * selftest.h: Fix alphabetization of per-source-file selftest
1100 2018-04-30 Jason Merrill <jason@redhat.com>
1102 PR c++/61982 - dead stores to destroyed objects.
1103 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1106 2018-04-30 Jason Merrill <jason@redhat.com>
1108 * tree.c (build_clobber): New.
1109 * tree.h: Declare it.
1110 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1112 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1114 * diagnostic-show-locus.c (layout::layout): Update for
1115 location_get_source_line returning a char_span.
1116 (struct char_span): Move to input.h.
1117 (struct correction): Update for fields in char_span becoming
1119 (struct source_line): Update for location_get_source_line
1120 returning a char_span.
1121 (layout::print_line): Likewise.
1122 * edit-context.c (edited_file::print_content): Likewise.
1123 (edited_file::print_diff_hunk): Likewise.
1124 (edited_file::print_run_of_changed_lines): Likewise.
1125 (edited_file::get_num_lines): Likewise.
1126 (edited_line::edited_line): Likewise.
1127 * final.c (asm_show_source): Likewise.
1128 * input.c (location_get_source_line): Convert return type
1129 from const char * to char_span, losing the final "line_len"
1131 (dump_location_info): Update for the above.
1132 (get_substring_ranges_for_loc): Likewise. Use a char_span
1133 when handling the literal within the line.
1134 (test_reading_source_line): Update for location_get_source_line
1135 returning a char_span.
1136 * input.h (class char_span): Move here from
1137 diagnostic-show-locus.c, converting from a struct to a class.
1138 Make data members private.
1139 (char_span::operator bool): New.
1140 (char_span::length): New.
1141 (char_span::get_buffer): New.
1142 (char_span::operator[]): New.
1143 (char_span::subspan): Make const.
1144 (char_span::xstrdup): New.
1145 (location_get_source_line): Convert return type from const char *
1146 to char_span, losing the final "line_size" param.
1148 2018-04-30 Jan Hubicka <jh@suse.cz>
1150 * lto-wrapper.c (ltrans_priorities): New static var.
1151 (cmp_priority): New.
1152 (run_gcc): Read priorities and if doing parallel build order
1153 the Makefile by them.
1155 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1157 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1159 2018-04-30 Richard Biener <rguenther@suse.de>
1161 * tree-cfg.c (verify_address): Remove base argument, add
1162 flag whether to check TREE_ADDRESSABLE and do that.
1163 (verify_expr): Remove.
1164 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1165 (verify_gimple_assign_single): Likewise.
1166 (verify_gimple_switch): Likewise.
1167 (verify_expr_location_1): Dereference tp once. Add (disabled)
1168 piece from verify_expr.
1169 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1171 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1173 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1175 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1177 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1178 (small_data_pattern): Likewise.
1179 (arc_rewrite_small_data): Likewise.
1180 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1181 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1182 (get_symbol_alignment): New function.
1183 (legitimate_small_data_address_p): Likewise.
1184 (legitimate_scaled_address): Update, call
1185 legitimate_small_data_address_p.
1186 (output_sdata): New static variable.
1187 (arc_print_operand): Update how we handle small data operands.
1188 (arc_print_operand_address): Likewise.
1189 (arc_legitimate_address_p): Update, use
1190 legitimate_small_data_address_p.
1191 (arc_rewrite_small_data_p): Remove.
1192 (arc_rewrite_small_data_1): Likewise.
1193 (arc_rewrite_small_data): Likewise.
1194 (small_data_pattern): Likewise.
1195 (compact_sda_memory_operand): Update to use
1196 legitimate_small_data_address_p and get_symbol_alignment.
1197 (prepare_move_operands): Don't rewite sdata pattern.
1198 (prepare_extend_operands): Remove.
1199 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1201 (zero_extendqisi2): Likewise.
1202 (zero_extendhisi2): Likewise.
1203 (extendqihi2): Likewise.
1204 (extendqisi2): Likewise.
1205 (extendhisi2): Likewise.
1209 * config/arc/constraints.md (Usd): Change it to memory constraint.
1211 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1213 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1214 as source of std instructions.
1215 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1216 allow 6-bit constants as source for store instructions.
1217 (movdi_insn): Update instruction pattern to allow 6-bit constants
1218 as source for store instructions.
1220 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1222 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1224 2018-04-30 Nathan Sidwell <nathan@acm.org>
1225 Sandra Loosemore <sandra@codesourcery.com>
1227 * dumpfile.c (dump_open): Allow '-' for stdout.
1228 * doc/invoke.texi (Developer Options): Document dump filename
1229 determination early. Document stdin/stdout selection.
1231 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1233 Microblaze Target: PIC data text relative
1235 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1236 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1238 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1239 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1240 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1241 New addressing mode for data-text relative position indepenedent code.
1242 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1243 'ADDRESS_SYMBOLIC_TXT_REL'.
1244 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1245 (microblaze_legitimate_pic_operand): Exclude function calls from
1246 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1247 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1249 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1250 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1251 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1252 for 'address + offset'.
1253 (microblaze_expand_prologue): Add new function prologue call for
1255 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1256 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1257 table in case of TARGET_PIC_DATA_TEXT_REL.
1258 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1259 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1260 Add new macros 'UNSPEC_TEXT',
1261 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1262 + exclude function calls from 'UNSPEC_PLT' in case of data text
1264 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1265 new target hook for generating address diff vector tables in case of
1267 * doc/tm.texi : Regenerate.
1268 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1269 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1270 of addr diff vector generation.
1271 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1272 target hook definition.
1273 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1274 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1275 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1276 option for data text relative.
1278 2018-04-30 Richard Biener <rguenther@suse.de>
1280 * tree-chrec.h (evolution_function_is_constant_p): Remove
1282 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1284 2018-04-30 Richard Biener <rguenther@suse.de>
1287 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1289 2018-04-30 Richard Biener <rguenther@suse.de>
1291 PR tree-optimization/28364
1292 PR tree-optimization/85275
1293 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1294 copying first exit test.
1296 2018-04-28 Mark Wielaard <mark@klomp.org>
1298 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1300 (dwarf_AT): Handle DW_AT_addr_base.
1301 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1303 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1306 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1307 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1308 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1309 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1311 2018-04-28 Richard Biener <rguenther@suse.de>
1313 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1314 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1315 to reflect use. Only add interesting stmts.
1317 2018-04-27 Martin Jambor <mjambor@suse.cz>
1320 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1321 the jump function allows for passing through aggregate values.
1323 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1325 * input.h (in_system_header_at): Convert from macro to inline
1327 (from_macro_expansion_at): Likewise.
1328 (from_macro_definition_at): Likewise.
1330 2018-04-27 Jeff Law <law@redhat.com>
1332 * config.gcc: Mark tile* targets as deprecated/obsolete.
1334 2018-04-27 Richard Biener <rguenther@suse.de>
1336 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1339 2018-04-27 Richard Biener <rguenther@suse.de>
1341 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1343 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1345 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1346 with Yd constraint. Set "preferred_for_speed" attribute from
1347 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1349 (*movdi_internal): Ditto.
1350 (movti_interunit splitters): Remove
1351 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1352 (movdi_interunit splitters): Ditto.
1353 * config/i386/constraints.md (Ye): Remove.
1354 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1356 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1359 * config/aarch64/constraints.md (Usg): Limit to 31.
1362 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1364 PR tree-optimization/85529
1365 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1366 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1367 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1368 zero extension or masking of the MSB bit.
1369 (optimize_range_tests): Add FIRST_BB argument, pass it through
1370 to optimize_range_tests_var_bound.
1371 (maybe_optimize_range_tests, reassociate_bb): Adjust
1372 optimize_range_tests callers.
1374 2018-04-26 Richard Biener <rguenther@suse.de>
1375 Jakub Jelinek <jakub@redhat.com>
1377 * cgraph.h (symbol_table): Just declare debug method here.
1378 * symtab.c (symbol_table::debug): Define.
1380 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1382 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1384 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1386 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1387 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1388 (*movdi_internal): Substitute Yi and Yj constraint with x
1389 and Ym and Yn constraint with y constraint. Update "isa"
1390 attribute and set "preferred_for_speed" attribute from
1391 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1392 (*movsi_internal): Ditto.
1393 (*movdf_internal): Ditto.
1394 (*movsf_internal): Ditto.
1395 (*zero_extendsidi2): Ditto.
1396 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1397 (sse2_loadld): Ditto.
1398 (*vec_extract<ssevecmodelower>_0): Ditto.
1399 (*vec_extractv4si_0_zext_sse4): Ditto.
1400 (vec_concatv2di): Ditto.
1401 (*vec_dup<mode>): Ditto.
1402 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1403 * config/i386/constraints.md (Yi): Remove.
1408 2018-04-26 Nathan Sidwell <nathan@acm.org>
1410 * dumpfile.c (dump_open): New.
1411 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1412 (dump_finish): Detect stdio/stderr by value not name.
1414 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1416 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1418 2018-04-26 Tom de Vries <tom@codesourcery.com>
1421 * config/nvptx/nvptx.c (verify_neutering_jumps)
1422 (verify_neutering_labels): New function
1423 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1425 2018-04-26 Tom de Vries <tom@codesourcery.com>
1428 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1429 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1432 2018-04-26 Richard Biener <rguenther@suse.de>
1433 Tom de Vries <tom@codesourcery.com>
1436 * lto-streamer-out.c (output_function): Fixup loops if required to match
1437 discovery done in the reader.
1439 2018-04-26 Richard Biener <rguenther@suse.de>
1441 PR tree-optimization/85116
1442 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1443 have a loop exit from the single latch predecessor. Remove
1444 case of header with just condition.
1445 (ch_base::copy_headers): Exclude infinite loops from any
1447 (pass_ch::execute): Record exits.
1449 2018-04-26 Richard Biener <rguenther@suse.de>
1451 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1452 prologue cost vector and pass it to vect_get_load_cost.
1453 (vect_get_peeling_costs_all_drs): Likewise.
1454 (vect_peeling_hash_get_lowest_cost): Likewise.
1455 (vect_enhance_data_refs_alignment): Likewise.
1457 2018-04-26 Richard Biener <rguenther@suse.de>
1460 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1461 checking of integer<->pointer conversions.
1462 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1463 sign-/zero-extending pointer types.
1464 (expand_omp_for_static_chunk): Likewise.
1466 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1467 Jean Lee <xiaoyur347@gmail.com>
1469 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1470 (TARGET_ASAN_SHADOW_OFFSET): Define.
1471 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1472 true for -fsanitize=address.
1474 2018-04-25 Mark Wielaard <mark@klomp.org>
1476 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1479 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1481 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1482 than "alu", remove explicit "memory" and "imm_disp" attributes.
1483 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1486 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1487 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1488 gen_lowpart_no_emit.
1490 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1493 * config/i386/i386.c (ix86_expand_builtin): Change memory
1494 operand to XI, extend p0 to Pmode.
1495 * config/i386/i386.md: Change unspec volatile and operand
1496 1 mode to XI, change operand 0 mode to P.
1498 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1500 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1501 GET_MODE_MASK before any checking.
1502 (nds32_can_use_bset_p): Likewise.
1503 (nds32_can_use_btgl_p): Likewise.
1505 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1507 * config/nds32/nds32-doubleword.md: New define_split pattern for
1508 illegal register number.
1510 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1512 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1514 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1516 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1518 2018-04-25 Richard Biener <rguenther@suse.de>
1520 * lto-streamer.h (LTO_major_version): Bump to 8.
1522 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1524 * BASE-VER: Set to 9.0.0.
1526 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1528 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1529 in __abskf2 and __powikf2.
1531 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1534 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1535 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1536 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1537 Use the above on operand 2. Reindent.
1538 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1540 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1543 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1545 * config/i386/i386.opt (mcet): Removed.
1546 * doc/install.texi: Remove -mcet documentation.
1547 * doc/invoke.texi: Likewise.
1549 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1552 * doc/install.texi: Remove -mcet from bootstrap-cet.
1554 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1557 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1558 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1562 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1563 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1564 containing a CONST_VECTOR.
1566 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1568 * doc/install.texi: Update newlib dependency for nvptx.
1570 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1573 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1574 instead of INTVAL when shifting x left.
1576 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1578 PR tree-optimization/85478
1579 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1580 vect_grouped_store_supported for single element vectors.
1582 2018-04-24 Richard Biener <rguenther@suse.de>
1585 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1586 load cost increase to the case of non-constant step.
1588 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1591 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1592 destination if any_malformed_asm.
1594 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1597 * expr.c (store_field): In the bitfield case, if the value comes from
1598 a function call and is returned in registers by means of a PARALLEL,
1599 do not change the mode of the temporary unless BLKmode and VOIDmode.
1601 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1603 PR rtl-optimization/85423
1604 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1605 dependencies to debug insns when the previous insn is non-debug.
1607 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1609 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1610 enums into a single definition.
1611 (fls): Fix predicates and printing.
1614 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1616 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1617 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1618 and short u6 immediate.
1619 (check_if_valid_sleep_operand): Remove.
1620 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1622 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1624 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1625 flag_always_save_lp condition.
1626 * config/nds32/nds32.opt (malways-save-lp): New option.
1628 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1630 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1631 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1632 * config/nds32/nds32.h
1633 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1634 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1636 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1638 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1639 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1641 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1642 Chung-Ju Wu <jasonwucj@gmail.com>
1644 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1646 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1647 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1649 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1651 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1653 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1655 * config/nds32/nds32-protos.h (nds32_data_alignment,
1656 nds32_local_alignment): Declare.
1657 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1658 nds32_local_alignment): New functions.
1659 (TARGET_CONSTANT_ALIGNMENT): Define.
1660 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1662 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1664 * config/nds32/nds32.c
1665 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1666 (TARGET_MODES_TIEABLE_P): Likewise.
1668 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1670 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1673 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1674 Chung-Ju Wu <jasonwucj@gmail.com>
1676 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1677 for unaligned access.
1678 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1680 * config/nds32/nds32-intrinsic.md: Likewise.
1681 * config/nds32/nds32_intrinsic.h: Likewise.
1682 * config/nds32/nds32.h (nds32_builtins): Likewise.
1683 * config/nds32/nds32.opt (munaligned-access): New option.
1684 * config/nds32/nds32.c (nds32_asm_file_start): Display
1685 flag_unaligned_access status.
1687 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1689 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1690 -mno-relax is present.
1691 * config/riscv/linux.h (LINK_SPEC): Ditto.
1693 2018-04-20 Martin Sebor <msebor@redhat.com>
1696 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1698 (gimple_fold_builtin_stxcpy_chk): Same.
1699 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1701 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1704 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1705 __powikf2 when long double is IEEE 128-bit.
1707 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1709 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1710 step to make sure stack always aligned.
1712 2018-04-20 Carl Love <cel@us.ibm.com>
1715 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1716 size check for arg0.
1718 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1719 Tom de Vries <tom@codesourcery.com>
1722 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1723 Emit insns for calls too.
1724 (nvptx_find_par): Always look for worker-level predecessor insn.
1725 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1727 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1728 (nvptx_process_pars): Propagate frames for calls.
1730 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1733 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1735 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1736 (ix86_handle_option): Don't handle OPT_mibt.
1737 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1739 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1741 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1742 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1743 (ix86_target_macros): Define __CET__ with flag_cf_protection
1744 for -fcf-protection.
1745 * config/i386/i386.c (isa2_opts): Remove -mibt.
1746 * config/i386/i386.h (TARGET_IBT): Removed.
1747 (TARGET_IBT_P): Likewise.
1748 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1749 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1750 * config/i386/i386.opt (mcet): Update help message.
1753 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1754 -mcet as an alias for -mshstk.
1756 2018-04-20 Richard Biener <rguenther@suse.de>
1759 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1760 complexity by forcing a single use of the multiply operand.
1762 2018-04-20 Martin Jambor <mjambor@suse.cz>
1765 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1766 recursion dependency to only apply to non-clones.
1768 2018-04-20 Martin Jambor <mjambor@suse.cz>
1771 * ipa-cp.c (create_specialized_node): Check that clones of
1772 self-recursive edges exist during IPA-CP.
1774 2018-04-19 Toon Moene <toon@moene.org>
1776 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1779 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1781 PR tree-optimization/85467
1782 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1783 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1784 VECTOR_CST element to type.
1786 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1789 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1790 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1791 (builtin_longjmp): Likewise.
1792 (save_stack_nonlocal): New pattern.
1793 (restore_stack_nonlocal): Likewise.
1795 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1798 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1799 Replace ASM_OUTPUT_LABEL with fprintf.
1801 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1804 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1805 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1806 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1807 define __IBT__ and __SHSTK__ for -fcf-protection.
1808 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1810 (ix86_trampoline_init): Likewise.
1811 (x86_output_mi_thunk): Likewise.
1812 (ix86_notrack_prefixed_insn_p): Likewise.
1813 (ix86_option_override_internal): Don't disallow -fcf-protection.
1814 * config/i386/i386.md (rdssp<mode>): Also enable for
1816 (incssp<mode>): Likewise.
1817 (nop_endbr): Likewise.
1818 * config/i386/i386.opt (mcet): Change help message to built-in
1822 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1823 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1824 enable CET built-in functions.
1826 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1828 * common/config/i386/i386-common.c
1829 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1830 OPTION_MASK_ISA_MOVDIRI_UNSET,
1831 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1832 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1833 * config.gcc (movdirintrin.h): New header.
1834 * config/i386/cpuid.h (bit_MOVDIRI,
1835 bit_MOVDIR64B): New bits.
1836 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1838 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1839 (VOID, PVOID, PCVOID)): New function types.
1840 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1841 __builtin_ia32_directstoreu_u64,
1842 __builtin_ia32_movdir64b): New builtins.
1843 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1844 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1846 (ix86_valid_target_attribute_inner_p): Ditto.
1847 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1848 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1849 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1850 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1851 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1852 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1853 (movdiri<mode>, movdir64b_<mode>): New.
1854 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1855 * config/i386/immintrin.h: Include movdirintrin.h.
1856 * config/i386/movdirintrin.h: New file.
1857 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1859 2018-04-19 Richard Biener <rguenther@suse.de>
1862 * cfg.c (clear_bb_flags): When loop state says we have
1863 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1865 2018-04-19 Richard Biener <rguenther@suse.de>
1867 PR tree-optimization/84737
1868 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1869 copying restrict info.
1870 (vect_setup_realignment): Use it.
1871 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1872 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1873 the first DR to all generated stores.
1874 (vectorizable_load): Likewise for loads.
1876 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1878 PR tree-optimization/85446
1879 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1880 the integral and pointer types to have the same precision.
1882 * doc/install.texi: Document --disable-cet being the default and
1885 2018-04-18 Martin Liska <mliska@suse.cz>
1887 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1890 2018-04-18 Martin Liska <mliska@suse.cz>
1893 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1896 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1897 arguments if they are comparable.
1899 2018-04-18 Martin Liska <mliska@suse.cz>
1902 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1905 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1908 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1911 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1912 ENDBR after calling __morestack.
1914 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1917 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1918 by using gcc_base_ver to generate a gcc_driver_version, and use
1919 it when generating GCC_DRIVER_NAME.
1920 * configure: Regenerate.
1922 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1925 * config.gcc: Obsolete powerpc*-*-*spe*.
1927 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1930 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1931 (stabstr_D): Change type of unum from unsigned int to
1932 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1935 2018-04-17 Jim Wilson <jimw@sifive.com>
1938 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1939 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1940 Set arg_pointer_offset after using pretend_args_size.
1942 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1944 PR rtl-optimization/85431
1945 * dse.c (record_store): Ignore zero width stores.
1948 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1949 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1950 __builtin_stack_restore rather than after it.
1951 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1952 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1953 argument instead of virtual_dynamic_stack_rtx.
1955 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1957 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1959 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1960 Add note to error message to explain internal mapping of overloaded
1961 built-in function name to non-overloaded built-in function name.
1962 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1965 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1968 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1969 where the inputs overlap with the output.
1971 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1973 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1974 (=v, v) alternative and explicit "memory" attribute.
1975 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1976 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1978 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1979 "sselog1" type instead of "sselog".
1980 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1981 "sselog". Remove explicit "memory" attribute.
1982 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1983 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1985 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1986 "sselog1" type instead of "sselog". Remove explicit "memory"
1988 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1989 use "sselog1" type instead of "sselog". Remove explicit "memory"
1991 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1992 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1994 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1995 "sselog1" type instead of "sselog". Remove explicit "memory"
1997 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1998 use "sselog1" type instead of "sselog". Remove explicit "memory"
2002 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2005 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2008 2018-04-17 Martin Jambor <mjambor@suse.cz>
2011 * ipa-cp.c (create_specialized_node): Call
2012 expand_all_artificial_thunks if necessary.
2014 2018-04-17 Martin Liska <mliska@suse.cz>
2017 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2018 in message, remote space in between '_G' and '('.
2020 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2023 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2024 avx512f_vmcmp<mode>3<round_saeonly_name>,
2025 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2026 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2027 avx512f_rndscale<mode><round_saeonly_name>,
2028 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2029 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2030 Use %<iptr>2 instead of %2 for -masm=intel.
2031 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2032 avx512f_vcvttss2usi<round_saeonly_name>,
2033 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2035 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2036 avx512f_vcvttsd2usi<round_saeonly_name>,
2037 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2038 Use %q1 instead of %1 for -masm=intel.
2039 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2040 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2041 of %3 for -masm=intel.
2042 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2044 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2046 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2048 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2049 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2051 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2053 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2054 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2055 %g1 and one with %0 and %1.
2056 (avx512er_vmrcp28<mode><round_saeonly_name>,
2057 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2059 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2060 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2061 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2062 of %0 and %{%4%} for -masm=intel.
2063 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2064 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2065 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2066 order of %0 and %{%5%}%{z%} for -masm=intel.
2068 2018-04-17 Jan Hubicka <jh@suse.cz>
2071 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2073 2018-04-17 Martin Liska <mliska@suse.cz>
2076 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2077 for target_clone error message. Make default implementation
2078 clone to be a local declaration.
2079 (separate_attrs): Add new argument and check for an empty
2081 (expand_target_clones): Handle it.
2082 (ipa_target_clone): Make redirection just for target_clones
2085 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2086 Tom de Vries <tom@codesourcery.com>
2089 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2090 tiled basic blocks without omp continue statements.
2092 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2095 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2096 vec_extract expression as having side effects to make sure it gets
2099 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2102 * config/i386/i386.c (get_builtin_code_for_version): Check
2105 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2108 * gcc/config.gcc: Support "skylake".
2109 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2111 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2112 (processor_target_table): Add "skylake".
2113 (ix86_option_override_internal): Add "skylake".
2114 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2115 PROCESSOR_CANNONLAKE.
2116 (get_builtin_code_for_version): Fix priority for
2117 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2118 PROCESSOR_SKYLAKE-AVX512.
2119 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2120 (processor_type): Add PROCESSOR_SKYLAKE.
2122 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2123 Jason Merrill <jason@redhat.com>
2126 * convert.c (convert_to_integer_1): Use direct recursion for
2127 enumeral types and types with a precision less than the number
2128 of bits in their mode.
2130 2018-04-16 Julia Koval <julia.koval@intel.com>
2133 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2134 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2136 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2139 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2140 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2141 and -mno-direct-move.
2143 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2146 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2147 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2148 Ensure negative shifts result in {0}.
2150 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2152 PR rtl-optimization/79916
2153 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2154 regs (if any) to define how to gnerate SD moves when LRA is in
2157 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2159 PR rtl-optimization/85393
2160 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2161 * except.c (expand_dw2_landing_pad_for_region): Make static.
2162 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2163 a label and unconditional jump to old_bb, rather than
2164 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2167 PR rtl-optimization/85376
2168 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2169 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2170 instead of a specific value.
2172 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2173 Bin Cheng <bin.cheng@arm.com>
2175 PR tree-optimization/82965
2176 PR tree-optimization/83991
2177 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2178 by_profile_only parameter.
2179 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2180 information if the loop was predicted to iterate too many times.
2181 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2183 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2186 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2189 2018-04-13 Martin Liska <mliska@suse.cz>
2190 Jakub Jelinek <jakub@redhat.com>
2193 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2194 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2195 * builtins.c (expand_builtin_memory_copy_args): Use
2196 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2197 handle dest_addr == pc_rtx.
2199 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2202 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2203 asked to not generate direct moves.
2204 (fix_trunc<mode>si2_stfiwx): Similar.
2205 (fix_trunc<mode>si2_internal): Similar.
2207 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2210 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2211 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2212 lookup if dest in some wider mode is known to be const0_rtx and
2213 if so, record permanent equivalence for it to be ZERO_EXTEND of
2214 the narrower mode destination.
2216 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2218 * lto-streamer-out.c (output_function): Revert 259346.
2219 * omp-expand.c (expand_oacc_for): Likewise.
2221 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2223 PR rtl-optimization/85354
2224 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2225 * sel-sched.c (sel_global_init): ... here.
2227 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2230 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2231 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2232 mode for PE-COFF targets.
2233 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2234 (i386_pe_asm_lto_end): Likewise.
2235 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2236 (TARGET_ASM_LTO_END): Likewise.
2237 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2238 (i386_pe_asm_lto_start): New function.
2239 (i386_pe_asm_lto_end): Likewise.
2241 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2242 Richard Biener <rguenther@suse.de>
2245 * lto-streamer-out.c (output_function): Fix CFG loop state before
2247 * omp-expand.c (expand_oacc_for): Handle calls to internal
2248 functions like regular functions.
2250 2018-04-12 Richard Biener <rguenther@suse.de>
2253 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2254 for the early LTO debug to properly generate references to it
2255 during DIE emission. Do not re-use that for the skeleton for
2257 (dwarf2out_early_finish): Likewise.
2259 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2262 * config/i386/sse.md
2263 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2264 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2265 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2266 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2267 and output is a reg, avoid creating invalid lowpart subreg, but
2268 instead split into a 512-bit move. Don't split if not AVX512VL,
2269 input is xmm16+ reg and output is a mem.
2270 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2271 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2272 xmm16+ reg and output is a mem.
2274 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2276 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2277 also for flag_dwarf2_cfi_asm.
2279 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2281 PR rtl-optimization/85342
2282 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2283 a bool scalar var inside of the loop instead. Don't try to update
2284 recog_data.operand after failed apply_change_group.
2286 2018-04-12 Tom de Vries <tom@codesourcery.com>
2289 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2290 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2291 array with flexible array member as array without given dimension.
2292 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2293 argument for undefined param to true.
2295 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2298 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2299 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2300 from PowerPC section.
2301 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2302 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2303 help text that is too long.
2304 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2305 help text that is too long.
2306 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2307 help text that is too long.
2309 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2311 * config/alpha/alpha.md (stack_probe_internal): Rename
2312 from "probe_stack". Update all callers.
2314 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2316 PR rtl-optimization/84566
2317 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2318 sched_macro_fuse_insns.
2320 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2323 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2324 (compute_block_dependences): ... from here.
2326 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2328 PR tree-optimization/85331
2329 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2330 from int to HOST_WIDE_INT.
2332 2018-04-11 Martin Jambor <mjambor@suse.cz>
2335 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2336 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2337 not the same as the source val.
2338 (cgraph_edge_brings_value_p): New parameter.
2339 (gather_edges_for_value): Pass destination value to
2340 cgraph_edge_brings_value_p.
2341 (perhaps_add_new_callers): Likewise.
2342 (get_info_about_necessary_edges): Likewise and exclude values brought
2343 only by self-recursive edges.
2344 (create_specialized_node): Redirect only clones of self-calling edges.
2345 (+self_recursive_pass_through_p): New function.
2346 (find_more_scalar_values_for_callers_subset): Use it.
2347 (find_aggregate_values_for_callers_subset): Likewise.
2348 (known_aggs_to_agg_replacement_list): Removed.
2349 (decide_whether_version_node): Re-calculate known constants for all
2350 remaining context clones.
2352 2018-04-11 Richard Biener <rguenther@suse.de>
2355 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2356 from early DWARF output.
2357 (dwarf2out_early_finish): Output line info unconditionally into
2358 early DWARF and add reference to it.
2360 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2363 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2364 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2365 other than V2DFmode using iptr mode attribute.
2366 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2368 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2370 PR rtl-optimization/84659
2371 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2373 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2376 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2378 (output_loc_list): Pass address of a dummy size variable even in the
2379 locview handling loop.
2380 (index_location_lists): Add comment on why skip_loc_list_entry can't
2383 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2386 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2389 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2392 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2393 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2394 and -mstring-compare-inline-limit.
2396 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2399 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2400 for stack clash protection in a register whenever we need it to be in
2403 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2405 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2406 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2408 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2411 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2413 (mlong-double-): Ditto.
2414 * config/rs6000/sysv4.opt (msdata=): Ditto.
2415 (mtls-size=): Ditto.
2417 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2419 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2420 erroneous entries for
2421 "vector int vec_ldl (int, long int *)", and
2422 "vector unsigned int vec_ldl (int, unsigned long int *)".
2423 Add comments and entries for
2424 "vector bool char vec_ldl (int, bool char *)",
2425 "vector bool short vec_ldl (int, bool short *)",
2426 "vector bool int vec_ldl (int, bool int *)",
2427 "vector bool long long vec_ldl (int, bool long long *)",
2428 "vector pixel vec_ldl (int, pixel *)",
2429 "vector long long vec_ldl (int, long long *)",
2430 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2431 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2432 type tree bool_long_long_type_node and correct definition of
2433 bool_V2DI_type_node to make reference to this new type tree.
2434 (rs6000_mangle_type): Replace erroneous reference to
2435 bool_long_type_node with bool_long_long_type_node.
2436 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2437 comments to emphasize sign distinctions for char and int types and
2438 replace RS6000_BTI_bool_long constant with
2439 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2440 use of RS6000_BTI_pixel.
2441 (bool_long_type_node): Remove this macro definition.
2442 (bool_long_long_type_node): New macro definition
2444 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2446 PR rtl-optimization/85300
2447 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2448 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2449 simplify_unary_operation fails.
2451 2018-04-10 Martin Liska <mliska@suse.cz>
2453 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2454 cgraph_edge and ipa_ref.
2456 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2460 * config/i386/sse.md
2461 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2462 computation of the VEC_MERGE selector from mask.
2463 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2464 Fix decoding of the VEC_MERGE selector into mask.
2466 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2468 PR tree-optimization/85286
2469 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2471 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2473 * final.c (final_1): Set insn_last_address as well as
2474 insn_current_address.
2476 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2479 * explow.c (emit_stack_probe): Call validize_mem on memory location
2480 before passing it to gen_probe_stack. Create address operand and
2481 legitimize it for the probe_stack_address case.
2483 2018-04-09 Jan Hubicka <jh@suse.cz>
2486 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2487 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2488 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2490 (free_lang_data): Rebuild type inheritance graph.
2492 2018-04-09 Martin Sebor <msebor@redhat.com>
2494 * invoke.texi (-finline-small-functions): Mention other optimization
2496 (-findirect-inlining, -fpartial-inlining): Same.
2497 (-finline-functions-called-once): Same.
2498 (-freorder-blocks-and-partition): Same.
2500 2018-04-09 Jan Hubicka <jh@suse.cz>
2503 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2504 jumps; choose last target that matches the criteria (i.e.
2505 no partition changes for non-crossing jumps).
2506 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2507 support for redirecting crossing jumps to non-crossing.
2509 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2511 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2512 also for naked functions.
2514 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2516 * config/arc/arc.md (add_shift): New pattern.
2517 (add_shift2): Likewise.
2518 (sub_shift): Likewise.
2519 (sub_shift_cmp0_noout): Likewise.
2520 (compare_si_ashiftsi): Likewise.
2521 (xbfu_cmp0_noout): New combine pattern.
2522 (xbfu_cmp0"): Likewise.
2523 (movsi_set_cc_insn): Place the predicable variant first.
2524 (commutative_binary_cmp0_noout): Remove clobber.
2525 (commutative_binary_cmp0): New pattern.
2526 (noncommutative_binary_cmp0): Likewise.
2527 (noncommutative_binary_cmp0_noout): Likewise.
2528 (noncommutative_binary_comparison_result_used): Removed.
2529 (rsub_cmp0): New pattern.
2530 (rsub_cmp0_noout): Likewise.
2531 (extzvsi): Changed, keep only meaningful variants.
2532 (SQH, SEZ): New iterators.
2533 (SQH_postfix): New mode attribute.
2534 (SEZ_prefix): New code attribute.
2535 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2536 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2537 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2539 (noncommutative_operator): Check the availability of barrel
2542 2018-04-09 Richard Biener <rguenther@suse.de>
2544 PR tree-optimization/85284
2545 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2546 Only use the niter constraining form of simple_iv when the exit
2549 2018-04-09 Tom de Vries <tom@codesourcery.com>
2552 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2553 (define_expand "*memory_barrier"): New define_expand.
2554 (define_insn "memory_barrier"): New insn.
2556 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2558 PR rtl-optimization/80463
2559 PR rtl-optimization/83972
2560 PR rtl-optimization/83480
2562 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2563 correct producer for the insn.
2564 (tidy_control_flow): Fixup seqnos in case of debug insns.
2566 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2568 PR rtl-optimization/83913
2570 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2571 different sched-times when merging exprs.
2573 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2575 PR rtl-optimization/83962
2577 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2578 tidy_fallthru_edge and tidy_control_flow.
2580 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2582 PR rtl-optimization/83530
2584 * sel-sched.c (force_next_insn): New global variable.
2585 (remove_insn_for_debug): When force_next_insn is true, also leave only
2586 next insn in the ready list.
2587 (sel_sched_region): When the region wasn't scheduled, make another pass
2588 over it with force_next_insn set to 1.
2590 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2592 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2594 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2595 for interrupt control.
2596 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2597 functions for interrupt control.
2598 * config/nds32/nds32-intrinsic.md: Likewise.
2599 * config/nds32/nds32_intrinsic.h: Likewise.
2600 * config/nds32/nds32.h (nds32_builtins): Likewise.
2602 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2604 * config/nds32/nds32.c (nds32_init_machine_status,
2605 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2606 strict_aligned_p field.
2607 (nds32_expand_to_rtl_hook): New function.
2608 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2609 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2611 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2612 Chung-Ju Wu <jasonwucj@gmail.com>
2614 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2615 * config/nds32/nds32-n7.md: New file.
2616 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2617 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2619 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2620 * config/nds32/nds32.md (pipeline_model): Add n7.
2621 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2622 * config/nds32/pipelines.md: Include n7 settings.
2624 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2625 Chung-Ju Wu <jasonwucj@gmail.com>
2627 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2628 * config/nds32/nds32-e8.md: New file.
2629 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2630 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2632 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2633 * config/nds32/nds32.md (pipeline_model): Add e8.
2634 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2635 * config/nds32/pipelines.md: Include e8 settings.
2637 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2638 Chung-Ju Wu <jasonwucj@gmail.com>
2640 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2641 * config/nds32/nds32-n8.md: New file.
2642 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2643 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2645 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2646 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2647 * config/nds32/nds32.md (pipeline_model): Add n8.
2648 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2649 * config/nds32/pipelines.md: Include n8 settings.
2651 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2652 Chung-Ju Wu <jasonwucj@gmail.com>
2654 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2655 * config/nds32/nds32-n9-2r1w.md: New file.
2656 * config/nds32/nds32-n9-3r2w.md: New file.
2657 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2658 nds32_register_ports): New or modify for cpu n9.
2659 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2661 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2662 * config/nds32/nds32-utils.c: New file.
2663 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2664 TARGET_MUL_SLOW): Define.
2665 * config/nds32/nds32.md (pipeline_model): New attribute.
2666 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2667 New options that support cpu n9.
2668 * config/nds32/pipelines.md: Include n9 settings.
2669 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2671 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2673 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2674 information if necessary.
2675 (output_cond_branch_compare_zero): Likewise.
2676 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2677 (nds32_target_alignment): Refine for alignment.
2678 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2679 (FUNCTION_BOUNDARY): Modify.
2680 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2682 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2684 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2686 * config/nds32/constants.md (unspec_volatile_element): Add values for
2687 TLB operation and data prefetch.
2688 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2689 functions for TLB operation and data prefetch.
2690 * config/nds32/nds32-intrinsic.md: Likewise.
2691 * config/nds32/nds32_intrinsic.h: Likewise.
2692 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2693 (nds32_print_operand): Likewise.
2694 * config/nds32/nds32.h (nds32_builtins): Likewise.
2696 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2697 Andrew Pinski <pinsika@gcc.gnu.org>
2700 * match.pd: Use constant_boolean_node of correct type instead of
2701 boolean_true_node or boolean_false_node for simplifying
2702 pointer comparisons to zero.
2704 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2706 PR tree-optimization/80021
2707 * tree.c (verify_type_variant): Make error call in verify_variant_match
2708 translatable and remove final full stop.
2710 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2712 * config/nds32/constants.md (unspec_volatile_element): Add
2713 UNSPEC_VOLATILE_EH_RETURN.
2714 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2715 nds32_output_stack_pop): Support dwarf exception handling process.
2716 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2717 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2718 exception handling process.
2719 (nds32_compute_stack_frame): Likewise.
2720 (nds32_return_addr_rtx): Likewise.
2721 (nds32_initial_elimination_offset): Likewise.
2722 (nds32_expand_prologue): Likewise.
2723 (nds32_expand_epilogue): Likewise.
2724 (nds32_dynamic_chain_address): New function.
2725 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2727 (DYNAMIC_CHAIN_ADDRESS): Define.
2728 (EH_RETURN_DATA_REGNO): Define.
2729 (EH_RETURN_STACKADJ_RTX): Define.
2730 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2731 patterns for dwarf exception handling.
2733 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2735 * config/nds32/nds32.h: Clean up obsolete macros.
2737 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2739 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2740 Add enum values for particular instructions.
2741 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2742 particular intrinsic functions.
2743 * config/nds32/nds32-intrinsic.md: Likewise.
2744 * config/nds32/nds32_intrinsic.h: Likewise.
2745 * config/nds32/nds32.h (nds32_builtins): Likewise.
2746 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2747 (btst, ave): New patterns for particular instructions.
2749 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2751 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2752 Add enum values for atomic load/store and memory sync.
2753 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2755 * config/nds32/nds32-intrinsic.md: Likewise.
2756 * config/nds32/nds32_intrinsic.h: Likewise.
2757 * config/nds32/nds32.h (nds32_builtins): Likewise.
2759 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2761 PR tree-optimization/85257
2762 * fold-const.c (native_encode_vector): If not all elts could fit
2763 and off is -1, return 0 rather than offset.
2764 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2765 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2766 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2767 adjust buffer in native_interpret_expr call.
2769 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2771 * config/nds32/constants.md (unspec_volatile_element): Add cache
2772 control enum values.
2773 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2774 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2775 * config/nds32/nds32.c (nds32_cctl_names): New.
2776 (nds32_print_operand): Handle cache control register names.
2777 * config/nds32/nds32.h (nds32_builtins): New enum values.
2778 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2780 * config/nds32/nds32.md (type): Add mmu.
2781 * config/nds32/pipelines.md (simple_insn): Add mmu.
2783 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2785 * config/nds32/nds32.md (type): Remove call.
2786 * config/nds32/pipelines.md (simple_insn): Likewise.
2788 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2790 * config/nds32/constants.md (unspec_volatile_element): Add
2791 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2792 UNSPEC_VOLATILE_FMFCFG.
2793 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2794 description for fmfcfg and fmfcsr.
2795 (bdesc_1arg): Add fmtcsr.
2796 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2797 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2798 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2799 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2800 unspec_fmfcfg): New patterns.
2801 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2802 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2803 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2804 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2805 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2806 __nds32__fmfcfg): Define.
2808 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2810 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2811 intrinsic register names.
2812 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2813 intrinsic register enum values and macros.
2815 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2817 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2818 for load/store addressing form.
2819 (nds32_print_operand_address): Likewise.
2821 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2824 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2825 based on LABEL_REF. Remove useless assertion.
2826 (pic_address_needs_scratch): Fix formatting.
2827 (sparc_legitimize_pic_address): Minor tweaks.
2828 (sparc_delegitimize_address): Adjust assertion accordingly.
2829 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2830 into symbolic_operand.
2831 (movsi_high_pic_label_ref): Likewise.
2832 (movsi_lo_sum_pic_label_ref): Likewise.
2833 (movdi_pic_label_ref): Likewise.
2834 (movdi_high_pic_label_ref): Likewise.
2835 (movdi_lo_sum_pic_label_ref): Likewise.
2837 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2839 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2840 custom LIB_SPEC setup.
2842 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2843 Kito Cheng <kito.cheng@gmail.com>
2845 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2846 * config/riscv/freebsd.h: New.
2848 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2850 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2851 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2854 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2855 Kito Cheng <kito.cheng@gmail.com>
2857 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2858 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2859 * config/nds32/nds32-protos.h (nds32_output_call,
2860 nds32_output_return): Declare.
2861 * config/nds32/nds32.md: Refine all the call and return patterns.
2863 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2866 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2867 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2869 PR rtl-optimization/84872
2870 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2871 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2874 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2876 * expr.c (copy_blkmode_to_reg): Revert 254862.
2877 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2879 2018-04-06 Richard Biener <rguenther@suse.de>
2882 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2883 after seeing a component reference with an adjacent field. Treat
2884 refs to arrays at struct end of external decls similar to
2885 refs to unconstrained commons.
2887 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2890 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2891 look through SAVE_EXPRs with non-side-effects argument. Adjust
2893 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2896 2018-04-06 Richard Biener <rguenther@suse.de>
2899 * alias.c (find_base_term): New wrapper around find_base_term
2900 unwinding CSELIB_VAL_PTR changes.
2901 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2904 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2906 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2908 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2909 constant definitions.
2910 ("nop"): lr 0,0 -> nopr r0
2911 ("nop_lr0", "nop_lr1"): New insn definitions.
2913 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2915 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2916 NDS32_V3PUSH_AVAILABLE_P macro.
2918 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2919 Chung-Ju Wu <jasonwucj@gmail.com>
2921 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2922 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2923 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2924 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2925 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2926 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2927 * config/nds32/constraints.md: New constraints and checking for hard
2928 float configuration.
2929 * config/nds32/iterators.md: New mode iterator and attribute for hard
2930 float configuration.
2931 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2933 * config/nds32/nds32-fpu.md: New file.
2934 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2935 deal with hard float code generation.
2936 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2938 (abi_type, float_reg_number): New enum type.
2939 * config/nds32/nds32-predicates.c: New predicates for hard float.
2940 * config/nds32/nds32-protos.h: Declare functions for hard float.
2941 * config/nds32/nds32.c: Implementation for hard float configuration.
2942 * config/nds32/nds32.h: Definitions for hard float configuration.
2943 * config/nds32/nds32.md: Include hard float machine description and
2944 modify patterns for hard float configuration.
2945 * config/nds32/nds32.opt: New options for hard float configuration.
2946 * config/nds32/predicates.md: New predicates for hard float
2949 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2951 * common/config/nds32/nds32-common.c
2952 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2954 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2957 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2958 CONSTRUCTOR_ELT (ctor, ...)->value.
2960 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2963 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2965 2018-04-05 Tom de Vries <tom@codesourcery.com>
2968 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2971 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2972 Kito Cheng <kito.cheng@gmail.com>
2974 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2975 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2976 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2977 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2979 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2980 Kito Cheng <kito.cheng@gmail.com>
2982 * config/nds32/constraints.md (Ufe): New memory constraint.
2983 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2984 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2985 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2987 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2988 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2990 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2992 * config/nds32/nds32.md: Use optimize_size in the condition for
2993 alu-shift instructions.
2995 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2997 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2999 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3001 * config/nds32/nds32.md (negsi2): Refine pattern.
3003 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3004 Chung-Ju Wu <jasonwucj@gmail.com>
3006 * config/nds32/iterators.md (shift_rotate): New code iterator.
3007 (shift): New code attribute.
3008 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3009 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3010 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3011 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3012 bit-wise operations.
3013 (andsi3, *andsi3): Ditto.
3014 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3015 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3016 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3017 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3018 nds32_ior_operand, nds32_xor_operand): New predicates.
3020 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3022 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3023 (addsi3, subsi3): ... this.
3025 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3027 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3029 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3031 * config/nds32/nds32.md: Adjust indention.
3033 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3035 * config/nds32/nds32.md (feature): New attribute.
3037 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3039 * config/nds32/nds32.md (subtype): New attribute.
3041 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3044 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3045 expansion to perform a bitwise AND of the argument followed by a
3046 boolean negation of the result.
3048 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3050 PR rtl-optimization/84878
3051 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3052 the basic block. Assert the use reference is not artificial and that
3053 it has an associated insn.
3055 2018-04-04 Michael Matz <matz@suse.de>
3057 * builtins.c (compute_objsize): Pass correct operand
3058 to array_at_struct_end_p.
3060 2018-04-04 Richard Biener <rguenther@suse.de>
3063 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3064 from contexts for DINFO_LEVEL_TERSE and below.
3066 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3068 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3069 resiter_operand condition.
3070 * config/nds32/nds32.md (*move<mode>): Ditto.
3072 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3073 Monk Chiang <sh.chiang04@gmail.com>
3075 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3077 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3079 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3081 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3082 Kito Cheng <kito.cheng@gmail.com>
3084 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3085 nds32_cond_code_str, output_cond_branch,
3086 output_cond_branch_compare_zero, nds32_expand_cbranch,
3087 nds32_expand_cstore, nds32_expand_movcc,
3088 nds32_output_cbranchsi4_equality_zero,
3089 nds32_output_cbranchsi4_equality_reg,
3090 nds32_output_cbranchsi4_equality_reg_or_const_int,
3091 nds32_output_cbranchsi4_greater_less_zero: New functions.
3092 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3093 nds32_expand_cstore, nds32_expand_movcc,
3094 nds32_output_cbranchsi4_equality_zero,
3095 nds32_output_cbranchsi4_equality_reg,
3096 nds32_output_cbranchsi4_equality_reg_or_const_int,
3097 nds32_output_cbranchsi4_greater_less_zero): Declare.
3098 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3099 nds32_rimm11s_operand): New predicates.
3100 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3101 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3104 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3106 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3107 * config/nds32/nds32.md: Ditto.
3108 * config/nds32/pipelines.md: Ditto.
3110 2018-04-04 Richard Biener <rguenther@suse.de>
3112 PR tree-optimization/85168
3113 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3114 propagating abnormals.
3116 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3118 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3120 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3121 Kito Cheng <kito.cheng@gmail.com>
3123 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3124 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3125 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3126 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3127 * config/nds32/nds32.md (sibcall_internal): New.
3128 (sibcall_register): Remove.
3129 (sibcall_immediate): Remove.
3130 (sibcall_value_internal): New.
3131 (sibcall_value_register): Remove.
3132 (sibcall_value_immediate): Remove.
3133 * config/nds32/predicates.md (nds32_general_register_operand): New.
3134 (nds32_call_address_operand): New.
3136 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3138 PR rtl-optimization/85167
3139 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3140 bb_defs if *split_p, instead preinitialize it to NULL.
3142 PR tree-optimization/85156
3143 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3144 evaluating the argument multiple times.
3146 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3148 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3150 (_mm_cvtpd_ps): Likewise.
3151 (_mm_cvttpd_epi32): Likewise.
3152 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3153 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3154 vector, pixel, and bool following altivec.h include.
3156 2018-04-03 Martin Sebor <msebor@redhat.com>
3158 * doc/extend.texi (Common Function Attributes): Clarify.
3159 (const attribute): Likewise.
3160 (pure attribute): Likewise.
3162 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3165 * config/i386/i386.c (ix86_expand_vector_set): Use
3166 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3168 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3170 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3171 instructions when changing rounding bits to preserve precision bits
3172 in the x87 control word.
3174 2018-04-03 Martin Liska <mliska@suse.cz>
3176 PR tree-optimization/82491
3177 * rtl.h (strip_offset_and_add): Replace += suboffset with
3178 poly_uint64 () + suboffset.
3180 2018-03-29 Martin Liska <mliska@suse.cz>
3181 Martin Jambor <mjambor@suse.cz>
3184 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3185 param_type is not an integral or pointer type.
3187 2018-04-03 Richard Biener <rguenther@suse.de>
3189 * sese.h (recompute_all_dominators): Remove.
3191 2018-04-02 Martin Sebor <msebor@redhat.com>
3193 * doc/invoke.texi (-Wrestrict): Fix typos.
3195 2018-04-02 Jim Wilson <jimw@sifive.com>
3197 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3198 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3199 (<optab>di3, <optab>si3_extend): Likewise.
3200 (<optab>si3_mask, <optab>si3_mask_1): New.
3201 (<optab>di3_mask, <optab>di3_mask_1): New.
3202 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3203 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3204 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3206 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3208 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3211 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3213 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3214 (nds32_canonicalize_comparison): New function.
3216 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3217 Kito Cheng <kito.cheng@gmail.com>
3218 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3220 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3221 * config/nds32/constants.md (unspec_volatile_element): Add
3222 UNSPEC_VOLATILE_RELAX_GROUP.
3223 * config/nds32/nds32-relax-opt.c: New file.
3224 * config/nds32/nds32-predicates.c
3225 (nds32_symbol_load_store_p): New function.
3226 * config/nds32/nds32-protos.h
3227 (nds32_symbol_load_store_p): Declare function.
3228 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3229 * config/nds32/nds32.c
3230 (nds32_register_pass): New function to register pass.
3231 (nds32_register_passes): New function to register passes.
3232 * config/nds32/nds32.md (relax_group): New pattern.
3233 * config/nds32/nds32.opt (mrelax-hint): New option.
3234 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3236 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3238 * config/nds32/t-nds32: Modify files dependency.
3240 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3242 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3243 (PROFILE_HOOK): Define its implementation.
3245 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3247 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3248 type and 32-bit size.
3250 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3253 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3254 (V_128_256): New mode iterator.
3255 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3256 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3257 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3259 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3260 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3262 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3265 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3266 NaN inputs correctly.
3268 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3271 * config/rs6000/vsx.md (??r): New mode attribute.
3272 (*vsx_mov<mode>_64bit): Use it.
3273 (*vsx_mov<mode>_32bit): Likewise.
3275 2018-03-30 Martin Sebor <msebor@redhat.com>
3277 PR tree-optimization/84818
3278 * builtins.c (check_access): Use warning_n.
3280 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3283 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3285 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3288 2018-03-30 Julia Koval <julia.koval@intel.com>
3291 * x86-tune.def (movx, partial_reg_dependency): Enable for
3294 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3297 * lra-constraints.c (process_alt_operands): Move setting
3298 this_alternative_matches below.
3300 2018-03-29 Martin Liska <mliska@suse.cz>
3303 * doc/invoke.texi: Document how LTO works with debug info.
3304 Describe auto-load support of binutils. Mention 'x86-64'
3305 as valid option value of -march option.
3307 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3309 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3312 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3313 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3314 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3317 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3320 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3321 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3322 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3323 for RS6000_BTM_POWERPC64.
3324 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3325 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3326 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3331 2018-03-28 Carl Love <cel@us.ibm.com>
3334 2017-09-27 Carl Love <cel@us.ibm.com>
3336 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3337 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3338 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3341 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3343 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3344 instead of __vector bool.
3345 (_mm_max_pu8): Likewise.
3346 (_mm_min_pi16): Likewise.
3348 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3351 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3352 (DIVWEUO): Likewise.
3354 (DIVDEUO): Likewise.
3355 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3356 DIVWEUO and DIVDEUO.
3357 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3358 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3359 (div_extend): Likewise.
3360 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3362 (__builtin_divweuo): Likewise.
3363 (__builtin_divdeo): Likewise.
3364 (__builtin_divdeuo): Likewise.
3366 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3369 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3370 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3372 PR tree-optimization/82004
3373 * gimple-match-head.c (optimize_pow_to_exp): New function.
3374 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3375 Don't fold to exp if optimize_pow_to_exp is false.
3377 2018-03-28 Martin Liska <mliska@suse.cz>
3380 * calls.c (initialize_argument_information): Fix trailing space.
3381 * common.opt: Fix typo and provide better explanation for
3382 -fsanitize-coverage option.
3383 * config/i386/i386.opt: Fix typo.
3385 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3386 Martin Liska <mliska@suse.cz>
3389 * gimplify.c (asan_poison_variable): Don't do the check for
3390 gimplify_omp_ctxp here.
3391 (gimplify_decl_expr): Do it here.
3392 (gimplify_target_expr): Likewise.
3394 2018-03-28 Martin Liska <mliska@suse.cz>
3397 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3398 chkp_type_bounds_count if MPX is not enabled.
3400 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3402 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3404 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3407 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3408 function to create the function decl for complex long double
3409 multiply and divide for -mabi=ieeelongdouble.
3410 (init_float128_ieee): Call it.
3412 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3415 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3416 -fcf-protection=branch -mibt.
3417 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3419 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3422 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3424 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3427 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3428 extern array declarations.
3430 2018-03-27 Richard Biener <rguenther@suse.de>
3433 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3434 explicit single_use checks.
3436 2018-03-27 Richard Biener <rguenther@suse.de>
3438 PR tree-optimization/85082
3439 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3442 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3444 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3445 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3446 Turn on fasynchronous-unwind-tables and funwind-tables.
3448 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3451 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3452 (*bmi_blsr_<mode>_ccz): Ditto.
3454 2018-03-26 Tom de Vries <tom@codesourcery.com>
3456 PR tree-optimization/85063
3457 * omp-general.c (offloading_function_p): New function. Factor out
3459 * omp-offload.c (pass_omp_target_link::gate): ... here.
3460 * omp-general.h (offloading_function_p): Declare.
3461 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3462 with attribute omp declare target for offloading functions.
3464 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3466 PR tree-optimization/84005
3467 * tree-data-ref.h (get_base_for_alignment): Declare.
3468 * tree-data-ref.c (get_base_for_alignment_1): New function.
3469 (get_base_for_alignment): Likewise.
3470 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3471 get_base_for_alignment to find a suitable base object, instead
3472 of always using drb->base_address.
3474 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3477 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3478 known size by default.
3480 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3483 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3484 and non BLKmode operands.
3486 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3489 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3490 Clean up attributes.
3492 2018-03-23 Richard Biener <rguenther@suse.de>
3495 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3496 we are going to emit early debug for LTO.
3498 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3501 * function.c (match_asm_constraints_1): Don't optimize if input
3502 doesn't satisfy general_operand predicate for output's mode.
3505 * alias.c (write_dependence_p): Don't require for x_canonicalized
3506 non-VOIDmode if x has VOIDmode.
3509 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3510 just don't try to optimize it rather than assert it never happens.
3512 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3514 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3515 macro expansions for definition of ST_INTERNAL_<mode> and
3516 LD_INTERNAL_<mode> builtins.
3517 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3519 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3521 (altivec_expand_st_builtin): Likewise.
3522 (altivec_expand_builtin): Remove calls to deleted functions.
3523 (rs6000_address_for_altivec): Delete this function.
3524 * config/rs6000/vector.md: Remove expands for
3525 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3527 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3530 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3531 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3532 re-computing once computed.
3533 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3534 (arm_init_machine_status): Initialize
3535 machine->static_chain_stack_bytes.
3537 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3540 * doc/extend.texi: Add four new prototypes for vec_ld.
3541 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3542 definitions for more logical presentation.
3543 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3544 entries for V1TI variants of __builtin_altivec_ld builtin.
3545 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3546 handling of V1TI variant of LVX icode pattern.
3547 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3548 (rs6000_gimple_fold_builtin): Likewise.
3549 (altivec_init_builtins): Add code to define
3550 __builtin_altivec_lvx_v1ti function.
3552 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3555 * function.c (match_asm_constraints_1): Don't do the optimization
3556 if input isn't a REG, SUBREG, MEM or constant.
3558 2018-03-22 Tom de Vries <tom@codesourcery.com>
3560 PR tree-optimization/84956
3561 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3562 bb_has_abnormal_pred.
3564 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3567 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3568 DECL_INITIAL (decl) to decl at the end.
3569 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3572 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3574 * doc/extend.texi (__builtin_tgmath): Document when complex
3575 integer types are treated as _Complex _Float64.
3577 2018-03-21 Tom de Vries <tom@codesourcery.com>
3579 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3581 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3583 PR tree-optimization/84960
3584 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3585 if it is ENTRY block, move them into single succ of ENTRY in that case.
3587 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3589 PR tree-optimization/84811
3590 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3591 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3592 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3593 (known_subrange_p): Update accordingly. Cast each value involved
3594 in the size comparison, rather than casting the result of the
3597 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3599 PR tree-optimization/84982
3600 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3601 by flipping the least significant bit rather than all bits from
3602 bitpos to bitpos + bitsize - 1.
3604 2018-03-21 Nathan Sidwell <nathan@acm.org>
3606 * doc/extend.texi (Deprecated Features): Remove mention of
3607 long-deleted deprecations.
3609 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3612 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3613 * configure: Regenerate.
3615 2018-03-21 Tom de Vries <tom@codesourcery.com>
3617 PR tree-optimization/83126
3618 * tree-parloops.c (num_phis): New function.
3619 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3621 2018-03-21 Nathan Sidwell <nathan@acm.org>
3623 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3624 mention anon-struct/union members and trailing attributes.
3626 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3628 PR tree-optimization/84969
3629 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3630 builtin memset partitions if they set different rhs values.
3632 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3634 PR rtl-optimization/84989
3635 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3636 VEC_DUPLICATE with scalar result mode.
3638 2018-03-21 Martin Liska <mliska@suse.cz>
3641 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3642 not intended return statement.
3644 2018-03-21 Martin Liska <mliska@suse.cz>
3647 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3648 (chkp_find_bound_slots_1): Limit number of iterations.
3650 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3653 * Minor grammar fixes for x86 options.
3655 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3658 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3659 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3661 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3664 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3665 (altivec_lvx_<mode>_1op): Likewise.
3666 (altivec_stvx_<mode>_2op): Likewise.
3667 (altivec_stvx_<mode>_1op): Likewise.
3668 (altivec_lvx_<VM2:mode>): New define_expand.
3669 (altivec_stvx_<VM2:mode>): Likewise.
3670 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3671 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3672 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3673 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3674 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3675 (rs6000_gen_lvx): Likewise.
3676 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3677 (altivec_expand_stv_builtin): Likewise.
3678 (altivec_expand_builtin): Likewise.
3679 * config/rs6000/vector.md: Likewise.
3681 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3684 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3687 2018-03-20 Richard Biener <rguenther@suse.de>
3690 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3691 sign-conversions as zero, fall back to standard scalar_stmt
3694 2018-03-20 Martin Liska <mliska@suse.cz>
3697 * predict.c (rebuild_frequencies): Handle case when we have
3698 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3700 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3703 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3704 flag_section_anchors.
3705 * varasm.c (use_blocks_for_decl_p): Remove hack for
3706 dw2_force_const_mem.
3709 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3711 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3712 be created, use lowpart_subreg of operands[0] rather than operands[0]
3714 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3715 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3716 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3717 and n constraint instead of aarch64_shift_imm_di and Usd.
3718 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3719 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3721 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3724 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3725 to favor GPR over NEON registers.
3726 (<shift>di3_neon): Likewise.
3728 2018-03-20 Tom de Vries <tom@codesourcery.com>
3731 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3732 (nvptx_process_pars): Emit bar.sync asap and alap.
3734 2018-03-20 Tom de Vries <tom@codesourcery.com>
3737 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3738 seen_label if seen_label is already set.
3740 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3743 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3744 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3745 Use 1U instead of 1. Formatting fixes.
3748 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3749 instead of TREE_TYPE (s1) for the return value.
3751 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3753 PR tree-optimization/84946
3754 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3755 bitsize + bitsize in poly_uint64 rather than poly_int64.
3758 * dwarf2asm.c: Include fold-const.c.
3759 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3760 of decl rather than decl itself.
3762 PR rtl-optimization/84643
3763 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3765 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3768 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3769 calling assemble_variable.
3771 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3774 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3775 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3777 2018-03-19 Jim Wilson <jimw@sifive.com>
3780 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3781 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3782 (riscv_first_stack_step): Likewise.
3783 (riscv_option_override): Use STACK_BOUNDARY instead of
3785 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3787 (BIGGEST_ALIGNMENT): Set to 128.
3788 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3789 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3792 2018-03-19 Richard Biener <rguenther@suse.de>
3794 PR tree-optimization/84933
3795 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3796 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3798 2018-03-19 Richard Biener <rguenther@suse.de>
3800 PR tree-optimization/84859
3801 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3802 (cond_if_else_store_replacement): Perform sinking operation on
3803 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3804 Generalize what a BB with a single eligible store is.
3806 2018-03-19 Richard Biener <rguenther@suse.de>
3808 PR tree-optimization/84929
3809 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3810 chrec_is_positive against non-chrec arg.
3812 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3815 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3817 2018-03-18 Martin Liska <mliska@suse.cz>
3819 PR rtl-optimization/84635
3820 * regrename.c (build_def_use): Use matches_mode only when
3823 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3825 PR tree-optimization/84913
3826 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3827 vectorize chains of COND_EXPRs.
3829 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3831 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3833 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3835 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3837 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3839 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3841 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3842 Kito Cheng <kito.cheng@gmail.com>
3844 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3845 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3846 (nds32_adjust_reg_alloc_order): New function.
3847 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3849 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3851 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3852 nds32_print_operand, nds32_print_operand_address): Use
3853 HOST_WIDE_INT_PRINT_DEC instead.
3855 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3857 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3859 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3862 * config/i386/i386.c (initial_ix86_tune_features,
3863 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3865 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3866 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3867 rather than 1u << ix86_tune. Formatting fix.
3868 (ix86_option_override_internal): Change ix86_arch_mask from
3869 unsigned int to unsigned HOST_WIDE_INT, initialize to
3870 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3871 (ix86_function_specific_restore): Likewise.
3873 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3876 * postreload.c (reload_combine_recognize_pattern): Perform
3877 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3878 truncate_int_for_mode the result for the destination's mode.
3881 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3882 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3884 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3887 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3890 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3891 regno_allocno_class_array and sorted_pseudos.
3892 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3893 insns where regno is used.
3895 2018-03-16 Martin Liska <mliska@suse.cz>
3898 * multiple_target.c (create_dispatcher_calls): Redirect
3899 reference in the symbol table.
3901 2018-03-16 Martin Liska <mliska@suse.cz>
3904 * multiple_target.c (create_dispatcher_calls): Redirect also
3907 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3911 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3912 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3913 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3915 2018-03-16 Julia Koval <julia.koval@intel.com>
3917 * doc/invoke.texi (Skylake Server): Add CLWB.
3918 Cannonlake): Remove CLWB.
3920 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3922 PR tree-optimization/84841
3923 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3925 (FLOAT_ONE_CONST_TYPE): Define.
3926 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3927 (sort_by_operand_rank): Put entries with higher constant_type last
3928 rather than first to match comments.
3930 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3932 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3935 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3938 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3939 check and use address of target always.
3941 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3944 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3945 (indirect_thunk_bnd_needed): Likewise.
3946 (indirect_thunks_used): Likewise.
3947 (indirect_thunks_bnd_used): Likewise.
3948 (indirect_return_needed): New.
3949 (indirect_return_bnd_needed): Likewise.
3950 (output_indirect_thunk_function): Add a bool argument for
3952 (output_indirect_thunk_function): Don't generate alias for
3953 function return thunk.
3954 (ix86_code_end): Call output_indirect_thunk_function to generate
3955 function return thunks.
3956 (ix86_output_function_return): Set indirect_return_bnd_needed
3957 and indirect_return_needed instead of indirect_thunk_bnd_needed
3958 and indirect_thunk_needed.
3960 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3962 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3963 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3964 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3966 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3967 Paul Hua <paul.hua.gm@gmail.com>
3970 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3972 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3974 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3975 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3978 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3981 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3982 instead of GET_MODE_SIZE when comparing Units.
3984 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3987 * varasm.c (hash_section): Return an unchangeble hash value
3988 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3989 Return !aarch64_can_use_per_function_literal_pools_p ().
3991 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3994 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3995 rather than address of cmode as last argument to prepare_cmp_insn.
3997 2018-03-15 Julia Koval <julia.koval@intel.com>
3999 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4000 F_AVX512VNNI, F_AVX512BITALG): New.
4002 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4005 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4006 insn for floating-point loads and stores.
4008 2018-03-14 Carl Love <cel@us.ibm.com>
4010 * config/rs6000/rs6000-c.c: Add macro definitions for
4011 ALTIVEC_BUILTIN_VEC_PERMXOR.
4012 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4013 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4014 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4015 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4017 * config/doc/extend.texi: Add prototypes for vec_permxor.
4019 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4022 * diagnostic-show-locus.c (class layout_point): Convert m_line
4023 from int to linenum_type.
4024 (line_span::comparator): Use linenum "compare" function when
4025 comparing line numbers.
4026 (test_line_span): New function.
4027 (layout_range::contains_point): Convert param "row" from int to
4029 (layout_range::intersects_line_p): Likewise.
4030 (layout::will_show_line_p): Likewise.
4031 (layout::print_source_line): Likewise.
4032 (layout::should_print_annotation_line_p): Likewise.
4033 (layout::print_annotation_line): Likewise.
4034 (layout::print_leading_fixits): Likewise.
4035 (layout::annotation_line_showed_range_p): Likewise.
4036 (struct line_corrections): Likewise for field m_row.
4037 (line_corrections::line_corrections): Likewise for param "row".
4038 (layout::print_trailing_fixits): Likewise.
4039 (layout::get_state_at_point): Likewise.
4040 (layout::get_x_bound_for_row): Likewise.
4041 (layout::print_line): Likewise.
4042 (diagnostic_show_locus): Likewise for locals "last_line" and
4044 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4045 * input.c (selftest::test_linenum_comparisons): New function.
4046 (selftest::input_c_tests): Call it.
4047 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4048 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4049 * selftest.h (ASSERT_GT): New macro.
4050 (ASSERT_GT_AT): New macro.
4051 (ASSERT_LT): New macro.
4052 (ASSERT_LT_AT): New macro.
4054 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4056 PR rtl-optimization/84780
4057 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4059 2018-03-14 Martin Liska <mliska@suse.cz>
4061 * tree.c (record_node_allocation_statistics): Use
4062 get_stats_node_kind.
4063 (get_stats_node_kind): New function extracted from
4064 record_node_allocation_statistics.
4065 (free_node): Use get_stats_node_kind.
4067 2018-03-14 Richard Biener <rguenther@suse.de>
4069 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4070 that the value-set of ANTIC_IN doesn't grow.
4073 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4075 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4076 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4078 2018-03-14 Julia Koval <julia.koval@intel.com>
4080 * config.gcc (icelake-client, icelake-server): New.
4082 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4083 (initial_ix86_arch_features): Ditto.
4084 (PTA_SKYLAKE): Add SGX.
4085 (PTA_ICELAKE): Remove.
4086 (PTA_ICELAKE_CLIENT): New.
4087 (PTA_ICELAKE_SERVER): New.
4088 (ix86_option_override_internal): Split up icelake on icelake client and
4090 (get_builtin_code_for_version): Ditto.
4091 (fold_builtin_cpu): Ditto.
4092 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4093 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4094 * config/i386/i386.h (processor_type): Ditto.
4095 * doc/invoke.texi: Ditto.
4097 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4100 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4101 INTEGER_CST offset, add it together with bitpos / 8 and
4102 sign extend based on POINTER_SIZE.
4106 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4109 * config/i386/constraints.md (Yc): New register constraint.
4110 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4111 Use Yc constraint for alternative 2 of operand 0. Remove
4112 preferred_for_speed attribute.
4114 2018-03-14 Richard Biener <rguenther@suse.de>
4116 PR tree-optimization/84830
4117 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4118 with the old one to avoid oscillations.
4120 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4123 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4125 (assign_by_spills): Return a flag of reload assignment failure.
4126 Do not process the reload assignment failures. Do not spill other
4127 reload pseudos if they has the same reg class. Update n if
4129 (lra_assign): Add a return arg. Set up from the result of
4130 assign_by_spills call.
4131 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4132 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4133 usage_insns if it is not NULL.
4134 (spill_hard_reg_in_range): New function.
4135 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4136 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4137 function prototypes.
4138 (lra_assign): Change prototype.
4139 * lra.c (lra): Add code to deal with fails by splitting hard reg
4142 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4144 * config/riscv/riscv.opt (mrelax): New option.
4145 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4146 "norelax" when riscv_mrelax is disabled.
4147 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4149 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4152 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4153 reassociation for int modes.
4155 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4157 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4158 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4160 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4161 * config/aarch64/aarch64-sve.md
4162 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4163 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4164 (*extend<mode><Vwide>2): Rename to...
4165 (aarch64_sve_extend<mode><Vwide>2): ...this.
4166 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4167 renaming the old pattern to...
4168 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4170 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4171 define_expand, renaming the old pattern to...
4172 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4173 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4174 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4175 account when deciding which SVE instruction the optab should use.
4176 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4178 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4180 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4181 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4182 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4183 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4184 (tlsdesc_small_<mode>): Turn a define_expand and use
4185 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4186 (tlsdesc_small_advsimd_<mode>): ...this.
4187 (tlsdesc_small_sve_<mode>): New pattern.
4189 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4191 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4192 (UNSPEC_UMUL_HIGHPART): New constants.
4193 (MUL_HIGHPART): New int iteraor.
4194 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4195 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4197 (*<su>mul<mode>3_highpart): New define_insn.
4199 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4202 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4205 2018-03-13 Martin Liska <mliska@suse.cz>
4208 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4210 (sem_item_optimizer::~sem_item_optimizer): Release it.
4211 (sem_item_optimizer::merge_classes): Register variable aliases.
4212 (sem_item_optimizer::fixup_pt_set): New function.
4213 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4214 * ipa-icf.h: Declare new variables and functions.
4216 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4219 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4220 integer_pow2p@2 and test integer_pow2p in condition.
4221 (A < 0 ? C : 0): Similarly for @1.
4224 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4225 characters starting at p contain '\0' character, don't look beyond
4229 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4230 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4233 * reg-stack.c (change_stack): Change update_end var from int to
4234 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4235 also call set_block_for_insn on the newly added insns and rescan.
4238 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4239 on the last operand.
4242 * tree.c (stabilize_reference_1): Return save_expr (e) for
4243 STATEMENT_LIST even if it doesn't have side-effects.
4245 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4247 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4249 2018-03-12 Renlin Li <renlin.li@arm.com>
4251 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4252 aarch64_output_scalar_simd_mov_immediate.
4254 2018-03-12 Martin Sebor <msebor@redhat.com>
4256 PR tree-optimization/83456
4257 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4258 for perfectly overlapping calls to memcpy.
4259 (gimple_fold_builtin_memory_chk): Same.
4260 (gimple_fold_builtin_strcpy): Handle no-warning.
4261 (gimple_fold_builtin_stxcpy_chk): Same.
4262 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4264 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4266 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4267 parameter. Use it for SFmode.
4268 (rs6000_function_arg_advance_1): Adjust.
4269 (rs6000_function_arg): Adjust.
4270 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4272 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4274 PR rtl-optimization/84169
4275 PR rtl-optimization/84780
4276 * combine.c (can_combine_p): Check for a 2-insn combination whether
4277 the destination register is used between the two insns, too.
4279 2018-03-12 Richard Biener <rguenther@suse.de>
4281 PR tree-optimization/84803
4282 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4283 for refs DR analysis didn't process.
4285 2018-03-12 Richard Biener <rguenther@suse.de>
4287 PR tree-optimization/84777
4288 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4289 force-vectorize loops ignore whether we are optimizing for size.
4291 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4293 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4294 (TARGET_MD_ASM_ADJUST): Define.
4296 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4297 Kito Cheng <kito.cheng@gmail.com>
4298 Chung-Ju Wu <jasonwucj@gmail.com>
4300 * config/nds32/nds32.c (nds32_compute_stack_frame,
4301 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4302 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4303 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4304 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4305 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4306 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4307 * config/nds32/nds32.md (prologue, epilogue): Use macro
4308 NDS32_V3PUSH_AVAILABLE_P to do checking.
4310 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4313 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4314 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4315 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4316 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4317 addition of most attributes on !orig_type_die or the attribute not
4318 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4320 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4321 Chung-Ju Wu <jasonwucj@gmail.com>
4323 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4325 * config/nds32/nds32.opt (mvh): New option.
4327 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4328 Chung-Ju Wu <jasonwucj@gmail.com>
4330 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4332 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4333 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4336 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4337 Chung-Ju Wu <jasonwucj@gmail.com>
4339 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4341 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4342 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4344 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4345 Kito Cheng <kito.cheng@gmail.com>
4346 Chung-Ju Wu <jasonwucj@gmail.com>
4348 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4349 UNSPEC_FFMISM and UNSPEC_FLMISM.
4350 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4351 for ffb, ffmism and flmism.
4352 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4353 (unspec_ffmism): Ditto.
4354 (unspec_flmism): Ditto.
4355 (nds32_expand_builtin_impl): Check if string extension is available.
4356 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4357 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4359 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4362 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4365 * lra-assigns.c (assign_by_spills): Return a flag of reload
4366 assignment failure. Do not process the reload assignment
4367 failures. Do not spill other reload pseudos if they has the same
4369 (lra_assign): Add a return arg. Set up from the result of
4370 assign_by_spills call.
4371 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4372 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4373 usage_insns if it is not NULL.
4374 (spill_hard_reg_in_range): New function.
4375 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4376 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4377 function prototypes.
4378 (lra_assign): Change prototype.
4379 * lra.c (lra): Add code to deal with fails by splitting hard reg
4382 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4385 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4387 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4390 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4391 (dw_val_node): Add val_symbolic_view.
4392 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4393 (symview_upper_bound): New.
4394 (new_line_info_table): Initialize symviews_since_reset.
4395 (dwarf2out_source_line): Count symviews_since_reset and set
4396 symview_upper_bound.
4397 (dw_val_equal_p): Handle symview.
4398 (add_AT_symview): New.
4399 (print_dw_val): Handle symview.
4400 (attr_checksum, attr_checksum_ordered): Likewise.
4401 (same_dw_val_p, size_of_die): Likewise.
4402 (value_format, output_die): Likewise.
4403 (add_high_low_attributes): Use add_AT_symview for entry_view.
4404 (dwarf2out_finish): Reset symview_upper_bound, clear
4407 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4410 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4411 Add strict argument and use it.
4412 (rs6000_split_multireg_move): Update for new strict argument.
4413 (mem_operand_gpr): Disallow all non-offsettable addresses.
4414 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4416 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4419 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4420 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4421 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4424 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4425 decl, use remap_type if we want to use the type.
4427 2018-03-09 Martin Sebor <msebor@redhat.com>
4429 PR tree-optimization/84526
4430 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4432 (builtin_access::generic_overlap): Be prepared to handle non-array
4435 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4437 PR rtl-optimization/84682
4438 * lra-constraints.c (process_address_1): Check is_address flag
4439 for address constraints.
4440 (process_alt_operands): Likewise.
4441 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4442 preprocess_constraints.
4443 * recog.h (preprocess_constraints): Add oploc parameter.
4445 * recog.c (preprocess_constraints): Test address_operand for
4446 CT_ADDRESS constraints.
4448 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4451 * lra-assigns.c (assign_by_spills): Return a flag of reload
4452 assignment failure. Do not process the reload assignment
4453 failures. Do not spill other reload pseudos if they has the same
4455 (lra_assign): Add a return arg. Set up from the result of
4456 assign_by_spills call.
4457 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4458 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4459 usage_insns if it is not NULL.
4460 (spill_hard_reg_in_range): New function.
4461 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4462 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4463 function prototypes.
4464 (lra_assign): Change prototype.
4465 * lra.c (lra): Add code to deal with fails by splitting hard reg
4468 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4471 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4472 Accept complain bool parameter. Only emit errors if it is true.
4473 (arm_parse_cpu_option_name): Likewise.
4474 (arm_target_thumb_only): Adjust callers of the above.
4475 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4476 prototype to take a default true bool parameter.
4477 (arm_parse_arch_option_name): Likewise.
4479 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4480 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4484 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4485 * configure: Regenerate.
4486 * configure.ac ("linker --version-script option"): New.
4487 ("linker soname option"): New.
4489 2018-03-09 Richard Biener <rguenther@suse.de>
4491 PR tree-optimization/84775
4492 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4493 immediate uses of predicate stmts and mark them modified.
4496 PR tree-optimization/84178
4497 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4499 (version_loop_for_if_conversion): Delay update_ssa call.
4500 (tree_if_conversion): Delay update_ssa until after predicate
4503 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4506 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4507 when the function accesses prior frames.
4509 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4512 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4513 gen_llsym, otherwise call maybe_gen_llsym.
4516 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4517 has ',' character inside of it.
4519 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4522 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4523 as clobbering CC_REGNUM.
4525 2018-03-08 Richard Biener <rguenther@suse.de>
4528 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4529 (follow_copies_to_constant): Do not follow SSA names registered
4532 2018-03-08 Richard Biener <rguenther@suse.de>
4534 PR tree-optimization/84178
4535 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4537 (version_loop_for_if_conversion): Delay update_ssa call.
4538 (tree_if_conversion): Delay update_ssa until after predicate
4541 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4543 PR tree-optimization/84178
4544 * tree-if-conv.c (release_bb_predicate): Remove the
4545 the assertion that the stmts have NULL use_ops.
4546 Discard the statements, asserting that they haven't
4547 yet been added to a BB.
4549 2018-03-08 Richard Biener <rguenther@suse.de>
4551 PR tree-optimization/84746
4552 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4553 (phi_translate): Pass in destination ANTIC_OUT set.
4554 (phi_translate_1): Likewise. For a simplified result lookup
4555 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4556 (phi_translate_set): Adjust.
4557 (do_pre_regular_insertion): Likewise.
4558 (do_pre_partial_partial_insertion): Likewise.
4560 2018-03-08 Martin Liska <mliska@suse.cz>
4562 PR gcov-profile/84735
4563 * doc/gcov.texi: Document usage of profile files.
4564 * gcov-io.h: Document changes in the format.
4566 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4570 * dwarf2out.c (struct dw_line_info_table): Update comments for
4572 (FORCE_RESET_NEXT_VIEW): New.
4573 (FORCE_RESETTING_VIEW_P): New.
4574 (RESETTING_VIEW_P): Check for -1 too.
4575 (ZERO_VIEW_P): Likewise.
4576 (new_line_info_table): Force-reset next view.
4577 (dwarf2out_begin_function): Likewise.
4578 (dwarf2out_source_line): Simplify zero_view_p initialization.
4579 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4580 view directly. Omit view when omitting .loc at line 0.
4582 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4584 PR tree-optimization/84740
4585 * tree-switch-conversion.c (process_switch): Call build_constructors
4586 only if info.phi_count is non-zero.
4588 PR tree-optimization/84739
4589 * tree-tailcall.c (find_tail_calls): Check call arguments against
4590 DECL_ARGUMENTS (current_function_decl) rather than
4591 DECL_ARGUMENTS (func) when checking for tail recursion.
4593 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4595 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4596 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4597 Volker Reichelt's entry and add entries for people that perform
4598 GCC fuzzy testing and report numerous bugs.
4600 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4603 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4604 readonly data in sdata, if that is disabled.
4605 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4606 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4607 -mreadonly-in-sdata option.
4609 2018-03-07 Martin Sebor <msebor@redhat.com>
4611 PR tree-optimization/84468
4612 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4613 basic block when looking for nul assignment.
4615 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4618 * except.h (output_function_exception_table): Adjust prototype.
4619 * except.c (output_function_exception_table): Remove FNNAME parameter
4620 and add SECTION parameter. Ouput one part of the table at a time.
4621 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4622 the first part of the exception table and emit unwind directives.
4623 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4624 (i386_pe_seh_cold_init): Likewise.
4625 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4626 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4627 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4628 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4629 * config/i386/winnt.c: Include except.h.
4630 (struct seh_frame_state): Add reg_offset, after_prologue and
4631 in_cold_section fields.
4632 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4633 (i386_pe_seh_cold_init): New function.
4634 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4635 to seh->in_cold_section.
4636 (seh_emit_push): Record the offset of the push.
4637 (seh_emit_save): Record the offet of the save.
4638 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4639 Test seh->after_prologue to disregard the epilogue.
4640 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4641 (i386_pe_end_cold_function): New function.
4643 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4646 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4647 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4650 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4652 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4653 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4656 * multiple_target.c: Include tree-inline.h and intl.h.
4657 (expand_target_clones): Diagnose and fail if node->definition and
4658 !tree_versionable_function_p (node->decl).
4660 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4662 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4664 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4665 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4666 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4668 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4671 * combine.c (try_combine): Use reg_or_subregno instead of handling
4672 just paradoxical SUBREGs and REGs.
4674 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4676 * config/arc/arc.c (arc_finalize_pic): Remove function.
4677 (arc_must_save_register): We use single base PIC register, remove
4678 checks to save/restore the PIC register.
4679 (arc_expand_prologue): Likewise.
4680 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4682 (arc_verify_short): Likewise.
4683 (arc_attr_type): Likewise.
4684 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4685 (walk_stores): Likewise.
4686 (arc_address_cost): Make it static.
4687 (arc_verify_short): Likewise.
4688 (branch_dest): Likewise.
4689 (arc_attr_type): Likewise.
4690 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4691 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4692 (arc_final_prescan_insn): Remove inserting the nops due to
4693 hardware hazards. It is done in reorg step.
4694 (insn_length_variant_t): Remove.
4695 (insn_length_parameters_t): Likewise.
4696 (arc_insn_length_parameters): Likewise.
4697 (arc_get_insn_variants): Likewise.
4698 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4700 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4703 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4706 PR tree-optimization/84687
4707 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4709 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4711 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4713 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4714 Rename to ppc_speculation_barrier.
4715 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4716 __builtin_ppc_speculation_barrier.
4718 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4721 * combine.c (combine_simplify_rtx): Don't try to simplify if
4722 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4725 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4727 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4728 to 32 bytes when compiling for POWER9.
4730 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4733 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4734 regparm >= 3 with no arg reg available also for calls with
4735 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4738 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4740 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4742 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4745 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4747 2018-03-05 Richard Biener <rguenther@suse.de>
4749 PR tree-optimization/84486
4750 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4751 When inserting a __builtin_assume_aligned call set the LHS
4752 SSA name alignment info accordingly.
4754 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4756 PR tree-optimization/84114
4757 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4758 Avoid reassociation of FLOAT_MODE addition.
4760 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4762 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4763 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4764 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4765 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4766 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4767 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4768 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4770 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4771 __builtin_ia32_wbinvd): New builtins.
4772 (SPECIAL_ARGS2): New.
4773 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4774 (SPECIAL_ARGS2): New.
4775 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4776 (ix86_valid_target_attribute_inner_p): Ditto.
4777 (ix86_init_mmx_sse_builtins): Add special_args2.
4778 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4779 TARGET_WBNOINVD_P): New.
4780 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4781 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4782 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4783 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4784 * config/i386/pconfigintrin.h: New file.
4785 * config/i386/wbnoinvdintrin.h: Ditto.
4786 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4787 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4789 2018-03-05 Richard Biener <rguenther@suse.de>
4791 PR tree-optimization/84670
4792 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4794 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4795 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4796 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4797 doesn't grow if all successors (recursively) were visited at least
4800 2018-03-05 Richard Biener <rguenther@suse.de>
4802 PR tree-optimization/84650
4803 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4804 if executed in the loop pipeline.
4806 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4808 * doc/configfiles.texi (Configuration Files): Move info about
4809 conditionalizing $target-protos.h to...
4810 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4811 differs from $target-protos.h.
4813 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4814 Chung-Ju Wu <jasonwucj@gmail.com>
4816 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4817 * config/nds32/nds32-multiple.md (setmemsi): Define.
4818 * config/nds32/nds32-memory-manipulation.c
4819 (nds32_gen_dup_4_byte_to_word_value): New.
4820 (emit_setmem_word_loop): New.
4821 (emit_setmem_byte_loop): New.
4822 (nds32_expand_setmem_loop): New.
4823 (nds32_expand_setmem_loop_v3m): New.
4824 (nds32_expand_setmem_unroll): New.
4825 (nds32_expand_setmem): New.
4827 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4828 Chung-Ju Wu <jasonwucj@gmail.com>
4830 * config/nds32/nds32-memory-manipulation.c
4831 (nds32_emit_load_store): New.
4832 (nds32_emit_post_inc_load_store): New.
4833 (nds32_emit_mem_move): New.
4834 (nds32_emit_mem_move_block): New.
4835 (nds32_expand_movmemsi_loop_unknown_size): New.
4836 (nds32_expand_movmemsi_loop_known_size): New.
4837 (nds32_expand_movmemsi_loop): New.
4838 (nds32_expand_movmemsi_unroll): New.
4839 (nds32_expand_movmemqi): Rename ...
4840 (nds32_expand_movmemsi): ... to this.
4841 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4842 (movmemsi): ... to this.
4843 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4844 (nds32_expand_movmemsi): ... to this.
4846 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4847 Monk Chiang <sh.chiang04@gmail.com>
4848 Chung-Ju Wu <jasonwucj@gmail.com>
4850 * config/nds32/nds32-protos.h
4851 (nds32_expand_load_multiple): New arguments.
4852 (nds32_expand_store_multiple): Ditto.
4853 (nds32_valid_multiple_load_store): Rename ...
4854 (nds32_valid_multiple_load_store_p): ... to this.
4855 * config/nds32/nds32-memory-manipulation.c
4856 (nds32_expand_load_multiple): Refine implementation.
4857 (nds32_expand_store_multiple): Ditto.
4858 * config/nds32/nds32-multiple.md
4859 (load_multiple): Update nds32_expand_load_multiple interface.
4860 (store_multiple): Update nds32_expand_store_multiple interface.
4861 * config/nds32/nds32-predicates.c
4862 (nds32_valid_multiple_load_store): Rename ...
4863 (nds32_valid_multiple_load_store_p): ... to this and refine
4865 * config/nds32/predicates.md
4866 (nds32_load_multiple_and_update_address_operation): New predicate.
4867 (nds32_store_multiple_and_update_address_operation): New predicate.
4869 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4870 Chung-Ju Wu <jasonwucj@gmail.com>
4872 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4873 (combo): New attribute.
4874 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4876 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4878 * config/nds32/nds32.opt: Change -mcmodel= default value.
4880 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4881 Monk Chiang <sh.chiang04@gmail.com>
4882 Chung-Ju Wu <jasonwucj@gmail.com>
4884 * config/nds32/constants.md (unspec_element): New enum.
4885 * config/nds32/constraints.md (Umw): New constraint.
4886 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4887 * config/nds32/nds32-intrinsic.md: Likewise.
4888 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4889 (nds32_valid_smw_lwm_base_p): New.
4890 (nds32_output_smw_single_word): New.
4891 (nds32_output_lmw_single_word): New.
4892 (nds32_expand_unaligned_load): New.
4893 (nds32_expand_unaligned_store): New.
4894 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4895 (nds32_output_smw_single_word): Declare.
4896 (nds32_output_lmw_single_word): Declare.
4897 (nds32_expand_unaligned_load): Declare.
4898 (nds32_expand_unaligned_store): Declare.
4899 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4900 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4901 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4902 NDS32_BUILTIN_UASTORE_DW.
4903 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4906 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4907 Kito Cheng <kito.cheng@gmail.com>
4908 Chung-Ju Wu <jasonwucj@gmail.com>
4910 * config/nds32/nds32-intrinsic.c
4911 (nds32_expand_builtin_null_ftype_reg): Delete.
4912 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4913 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4914 (nds32_read_argument): New.
4915 (nds32_legitimize_target): Ditto.
4916 (nds32_legitimize_argument): Ditto.
4917 (nds32_check_constant_argument): Ditto.
4918 (nds32_expand_unop_builtin): Ditto.
4919 (nds32_expand_unopimm_builtin): Ditto.
4920 (nds32_expand_binop_builtin): Ditto.
4921 (nds32_builtin_decl_impl): Ditto.
4922 (builtin_description): Ditto.
4923 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4924 (nds32_init_builtins_impl): Ditto.
4925 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4926 (nds32_builtin_decl): New.
4927 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4928 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4930 2018-03-02 Jeff Law <law@redhat.com>
4932 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4933 (redundant_insn, fill_simple_delay_slots): Likewise.
4934 (fill_slots_from_thread): Likewise.
4935 * resource.c (mark_referenced_resources): Likewise.
4936 (mark_set_resources, find_dead_or_set_registers): Likewise.
4938 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4940 * substring-locations.h (format_warning_va): Formatting fix for
4942 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4944 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4945 * substring-locations.c: Include intl.h.
4946 (format_warning_va): Turned into small wrapper around
4947 format_warning_n_va, renamed to ...
4948 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4949 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4951 (format_warning_at_substring_n): New function.
4952 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4953 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4954 format_warning_at_substring with just a shorter name instead of
4955 const function pointer.
4956 (fmtwarn_n): New function.
4957 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4958 appropriate, get rid of all the fmtstr temporaries, move conditionals
4959 with G_() wrapped string literals directly into fmtwarn arguments,
4960 cast dir.len to (int), formatting fixes.
4962 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4964 * doc/invoke.texi: Remove "Cilk Plus" references.
4966 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4967 Richard Biener <rguenther@suse.de>
4970 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4971 for error or warning attributes if CALL_FROM_THUNK_P is set.
4974 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4977 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4978 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4981 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4982 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4984 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4985 instead of -1U in last predictors element's probability member.
4987 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4990 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4991 arguments if they are comparable.
4993 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4995 PR tree-optimization/84634
4996 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4997 masks and masked_loop_p with a single loop_masks, making sure it's
4998 null for bb vectorization.
5000 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5002 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5003 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5004 loop->force_vectorize to check whether there is no alias.
5006 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5009 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5011 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5013 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5014 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5015 instead of a loop around prev_real_insn.
5016 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5020 * config/i386/i386.c (ix86_print_operand): Use conditional
5021 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5024 2018-03-02 Richard Biener <rguenther@suse.de>
5026 PR tree-optimization/84427
5027 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5028 (bitmap_set_subtract_values): Rewrite to handle multiple
5031 (prune_clobbered_mems): Likewise.
5032 (phi_translate): Take edge instead of pred/phiblock.
5033 (phi_translate_1): Likewise.
5034 (phi_translate_set): Likewise. Insert all translated
5035 exprs for a value into the set, keeping possibly multiple
5036 expressions per value.
5037 (compute_antic_aux): Adjust for phi_translate changes.
5038 When intersecting union the expressions and prune those
5039 not in the final value set, keeping possibly multiple
5040 expressions per value. Do not use value-insertion
5041 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5042 all expressions. Add verification that the value-sets
5043 only shrink during iteration.
5044 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5045 (do_pre_regular_insertion): Likewise.
5046 (do_pre_partial_partial_insertion): Likewise.
5048 2018-03-02 Richard Biener <rguenther@suse.de>
5051 * config/darwin.c (saved_debug_info_level): New static global.
5052 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5053 (darwin_asm_lto_end): Restore debug info generation settings.
5055 2018-03-01 Martin Liska <mliska@suse.cz>
5058 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5061 2018-03-01 Richard Biener <rguenther@suse.de>
5064 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5065 type annotation with LTO when debug was disabled at compile-time.
5067 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5069 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5071 (mips_final_postscan_insn): Likewise.
5073 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5075 PR rtl-optimization/84528
5076 * alias.c (init_alias_target): Add commentary.
5077 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5078 a unique base value if the frame pointer is not eliminated
5079 to the stack pointer.
5081 2018-03-01 Tom de Vries <tom@codesourcery.com>
5083 PR rtl-optimization/83327
5084 * lra-int.h (hard_regs_spilled_into): Declare.
5085 * lra.c (hard_regs_spilled_into): Define.
5086 (init_reg_info): Init hard_regs_spilled_into.
5087 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5088 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5089 (process_bb_lives): Handle hard_regs_spilled_into.
5090 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5091 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5093 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5095 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5096 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5097 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5098 * config/rs6000/aix72.h: New file.
5100 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5102 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5103 instead of warning_at with conditional singular and plural messages
5107 * stor-layout.c (update_alignment_for_field): For
5108 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5109 && !DECL_PACKED (field), do the alignment update, just use
5110 only desired_align instead of MAX (type_align, desired_align)
5112 (place_field): Don't do known_align < desired_align handling
5113 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5114 is non-NULL, instead do it after rli->prev_field handling and
5115 only if not within a bitfield word. For DECL_PACKED (field)
5116 use type_align of BITS_PER_UNIT.
5118 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5120 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5121 superfluous parentheses and trailing spaces.
5123 2018-02-28 Richard Biener <rguenther@suse.de>
5125 PR tree-optimization/84584
5126 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5127 SCoPs with fake exit edge.
5129 2018-02-28 Martin Liska <mliska@suse.cz>
5132 * timevar.c (timer::print): Fix format to properly print 100%
5135 2018-02-28 Richard Biener <rguenther@suse.de>
5138 * genmatch.c (capture_info::walk_match): Do not mark
5139 captured expressions without operands as expr_p given
5140 they act more like predicates and should be subject to
5141 "lost tail" side-effect preserving.
5143 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5145 PR rtl-optimization/81611
5146 * auto-inc-dec.c (attempt_change): Move dead note from
5147 mem_insn if it's the next use of regno
5148 (find_address): Take address use of reg holding
5149 non-incremented value. Add parm to limit search to the named
5151 (merge_in_block): Attempt to use a mem insn that is the next
5152 use of the original regno.
5154 2018-02-27 Martin Sebor <msebor@redhat.com>
5157 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5158 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5160 2018-02-27 Martin Sebor <msebor@redhat.com>
5162 PR translation/84207
5163 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5164 n argument to unsigned HOST_WIDE_INT.
5165 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5166 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5167 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5168 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5170 2018-02-27 Richard Biener <rguenther@suse.de>
5172 PR tree-optimization/84512
5173 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5174 Do not use the estimate returned from record_stmt_cost for
5175 the scalar iteration cost but sum properly using add_stmt_cost.
5177 2018-02-27 Richard Biener <rguenther@suse.de>
5179 PR tree-optimization/84466
5180 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5181 Adjust last change to less strictly validate use operands.
5183 2018-02-27 Martin Liska <mliska@suse.cz>
5185 PR gcov-profile/84548
5186 * gcov.c (process_file): Allow partial overlap and consider it
5187 also as group functions.
5188 (output_lines): Properly calculate range of lines for a group.
5190 2018-02-27 Martin Liska <mliska@suse.cz>
5192 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5193 'ggc' suffixes. Change first column width.
5194 (timer::print): Fix formatting of the column.
5196 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5198 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5199 preserve inline entry blocks for the sake of debug inline
5200 entry point markers alone.
5201 (remove_unused_locals): Suggest in comments a better place to
5202 force the preservation of inline entry blocks that are
5203 otherwise unused, but do not preserve them.
5205 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5207 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5209 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5212 * config/i386/constraints.md (Bs): Replace
5213 ix86_indirect_branch_register with
5214 TARGET_INDIRECT_BRANCH_REGISTER.
5216 * config/i386/i386.md (indirect_jump): Likewise.
5217 (tablejump): Likewise.
5218 (*sibcall_memory): Likewise.
5219 (*sibcall_value_memory): Likewise.
5220 Peepholes of indirect call and jump via memory: Likewise.
5221 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5222 (*sibcall_value_GOT_32): Likewise.
5223 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5224 (GOT_memory_operand): Likewise.
5225 (call_insn_operand): Likewise.
5226 (sibcall_insn_operand): Likewise.
5227 (GOT32_symbol_operand): Likewise.
5228 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5230 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5232 PR rtl-optimization/83496
5233 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5234 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5235 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5236 redundant insn, if any.
5237 (relax_delay_slots): Likewise.
5238 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5240 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5242 PR tree-optimization/83965
5243 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5244 that grouped statements are part of a reduction chain. Return
5245 true if the statement is not marked as a reduction itself but
5247 (vect_recog_dot_prod_pattern): Don't check whether the statement
5248 is part of a group here.
5249 (vect_recog_sad_pattern): Likewise.
5250 (vect_recog_widen_sum_pattern): Likewise.
5252 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5255 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5257 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5260 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5262 (ix86_output_indirect_function_return): New prototype.
5263 (ix86_split_simple_return_pop_internal): Likewise.
5264 * config/i386/i386.c (indirect_return_via_cx): New.
5265 (indirect_return_via_cx_bnd): Likewise.
5266 (indirect_thunk_name): Handle return va CX_REG.
5267 (output_indirect_thunk_function): Create alias for
5268 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5269 (ix86_output_indirect_jmp): Remove the bool argument.
5270 (ix86_output_indirect_function_return): New function.
5271 (ix86_split_simple_return_pop_internal): Likewise.
5272 * config/i386/i386.md (*indirect_jump): Don't pass false
5273 to ix86_output_indirect_jmp.
5274 (*tablejump_1): Likewise.
5275 (simple_return_pop_internal): Change it to define_insn_and_split.
5276 Call ix86_split_simple_return_pop_internal to split it for
5278 (simple_return_indirect_internal): Call
5279 ix86_output_indirect_function_return instead of
5280 ix86_output_indirect_jmp.
5282 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5285 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5286 memset and value initialization afterwards.
5288 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5290 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5292 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5295 * common/config/aarch64/aarch64-common.c
5296 (aarch_option_optimization_table[]): Switch
5297 off fomit-frame-pointer
5299 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5300 Chung-Ju Wu <jasonwucj@gmail.com>
5302 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5304 (store_multiple): Ditto.
5306 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5308 * config.gcc: Add --with-cpu support for nds32 target.
5309 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5310 * config/nds32/nds32.opt: Add -mcpu= option.
5312 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5314 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5315 isel=yes): Warn for these deprecated options.
5317 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5319 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5320 ISA_2_5_MASKS_EMBEDDED.
5322 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5324 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5325 p->max as pointers rather than using iterative_hash_expr.
5327 2018-02-23 Carl Love <cel@us.ibm.com>
5329 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5330 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5332 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5333 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5334 P8V_BUILTIN_VEC_VUNSIGNED2.
5336 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5339 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5340 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5342 * lra-constraints.c (curr_insn_transform): Set up
5343 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5345 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5346 * lra-eliminations.c (spill_pseudos): Ditto.
5347 (process_insn_for_elimination): Ditto.
5348 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5349 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5350 LRA_NON_CLOBBERED_ALT.
5352 2018-02-22 Martin Sebor <msebor@redhat.com>
5354 PR tree-optimization/84480
5355 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5356 to maybe_diag_stxncpy_trunc. Call it.
5357 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5358 from gimple_fold_builtin_strcpy. Print inlining stack.
5359 (handle_builtin_stxncpy): Print inlining stack.
5360 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5362 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5365 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5366 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5367 and -fcheck-pointer-bounds are used together.
5368 (indirect_thunk_prefix): New enum.
5369 (indirect_thunk_need_prefix): New function.
5370 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5371 "_nt" instead of "_bnd" for NOTRACK prefix.
5372 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5373 (output_indirect_thunk_function): Likewise.
5375 (ix86_code_end): Update output_indirect_thunk_function calls.
5376 (ix86_output_indirect_branch_via_reg): Replace
5377 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5378 (ix86_output_indirect_branch_via_push): Likewise.
5379 (ix86_output_function_return): Likewise.
5380 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5381 incompatible with -fcf-protection=branch and
5382 -fcheck-pointer-bounds.
5384 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5387 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5388 Change gcc_assert call to output_operand_lossage.
5390 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5392 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5394 2018-02-22 DJ Delorie <dj@redhat.com>
5395 Sebastian Perta <sebastian.perta@renesas.com>
5396 Oleg Endo <olegendo@gcc.gnu.org>
5398 * config/rx/rx.c (rx_rtx_costs): New function.
5399 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5401 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5403 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5405 2018-02-22 Martin Liska <mliska@suse.cz>
5408 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5409 Add "native" as a possible value.
5411 2018-02-22 Martin Liska <mliska@suse.cz>
5414 * config/i386/i386.c (ix86_option_override_internal):
5415 Add "native" as a possible value for -march and -mtune.
5417 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5420 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5421 to all type variants.
5423 PR tree-optimization/84503
5424 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5425 width as info->bitpos + info->bitsize - start.
5426 (merged_store_group::merge_overlapping): Simplify width computation.
5427 (check_no_overlap): New function.
5428 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5429 start + width and last_order of the group, fail if check_no_overlap
5431 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5432 to group if check_no_overlap fails.
5434 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5436 * config/rs6000/altivec.md: Delete contraint arguments to
5437 define_expand, define_split, and define_peephole2, and in
5438 define_insn_and_split if always unused.
5439 * config/rs6000/darwin.md: Ditto.
5440 * config/rs6000/dfp.md: Ditto.
5441 * config/rs6000/rs6000.md: Ditto.
5442 * config/rs6000/sync.md: Ditto.
5443 * config/rs6000/vector.md: Ditto.
5444 * config/rs6000/vsx.md: Ditto.
5446 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5448 * config/rs6000/altivec.md: Write output control strings as braced
5449 blocks instead of double-quoted strings.
5450 * config/rs6000/darwin.md: Ditto.
5451 * config/rs6000/rs6000.md: Ditto.
5452 * config/rs6000/vector.md: Ditto.
5453 * config/rs6000/vsx.md: Ditto.
5455 2018-02-21 Jason Merrill <jason@redhat.com>
5457 PR c++/84314 - ICE with templates and fastcall attribute.
5458 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5460 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5462 * ipa-cp.c (determine_versionability): Fix comment typos.
5464 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5467 * ipa-cp.c (determine_versionability): Do not version functions caling
5470 2018-02-21 Martin Liska <mliska@suse.cz>
5473 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5474 Add "native" as a possible value.
5475 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5476 the macro when native cpu detection is available.
5478 2018-02-21 Martin Liska <mliska@suse.cz>
5481 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5482 Add "native" as a possible value.
5483 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5484 when native cpu detection is available.
5486 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5487 Martin Sebor <msebor@redhat.com>
5489 PR tree-optimization/84478
5490 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5492 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5493 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5494 support which is conservatively correct, for 2 only stay conservative
5495 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5496 argument to the 2 argument get_range_strlen, adjust 6 arg
5497 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5499 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5500 (gimple_fold_builtin_strlen): Pass true as last argument to
5503 2018-02-20 Martin Sebor <msebor@redhat.com>
5506 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5507 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5508 (builtin_memref::builtin_memref): Factor out parts into
5509 set_base_and_offset and call it.
5511 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5514 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5515 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5516 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5517 search at the associated MODE_INT.
5519 2018-02-20 Jeff Law <law@redhat.com>
5522 PR tree-optimization/81592
5524 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5525 for range data rather than using global data.
5526 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5527 range data rather than using global data.
5528 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5529 pass it to children as needed.
5530 (struct directive::fmtresult): Similarly.
5531 (struct directive::set_width): Similarly.
5532 (struct directive::set_precision): Similarly.
5533 (format_integer, format_directive, parse_directive): Similarly.
5534 (format_none): Accept unnamed vr_values parameter.
5535 (format_percent, format_floating, format_character): Similarly.
5536 (format_string, format_plain): Similarly.
5537 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5538 the EVRP range analyzer for range data rather than using global data.
5539 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5540 gimple-ssa-evrp-analyze.h
5541 (class sprintf_dom_walker): Add after_dom_children member function.
5542 Add evrp_range_analyzer member.
5543 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5544 range analyzer as needed.
5545 (sprintf_dom_walker::after_dom_children): New member function.
5546 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5548 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5549 (evrp_range_analyzer::pop_to_marker): Likewise.
5551 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5553 PR tree-optimization/84419
5554 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5555 with the required type if its current type is compatible but
5558 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5561 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5562 after vectorization.
5564 2018-02-20 Martin Liska <mliska@suse.cz>
5567 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5568 possible values if we don't have a hint.
5570 2018-02-20 Martin Liska <mliska@suse.cz>
5574 * final.c (shorten_branches): Build align_tab array with one
5576 * opts.c (finish_options): Add alignment option limit check.
5577 (MAX_CODE_ALIGN): Likewise.
5578 (MAX_CODE_ALIGN_VALUE): Likewise.
5579 * doc/invoke.texi: Document maximum allowed option value for
5580 all -falign-* options.
5582 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5585 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5586 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5587 * var-tracking.c (emit_note_insn_var_location): Remove all references
5588 to NOTE_INSN_CALL_ARG_LOCATION.
5589 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5590 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5591 Use copy_rtx_if_shared.
5592 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5593 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5594 (dwarf2out_var_location): Remove handling of
5595 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5597 * final.c (final_scan_insn): Remove all references to
5598 NOTE_INSN_CALL_ARG_LOCATION.
5599 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5600 before dumping final insns.
5601 * except.c (emit_note_eh_region_end): Remove all references to
5602 NOTE_INSN_CALL_ARG_LOCATION.
5603 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5604 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5605 * config/arc/arc.c (hwloop_optimize): Likewise.
5606 * config/arm/arm.c (create_fix_barrier): Likewise.
5607 * config/s390/s390.c (s390_chunkify_start): Likewise.
5608 * config/sh/sh.c (find_barrier): Likewise.
5609 * config/i386/i386.c (rest_of_insert_endbranch,
5610 ix86_seh_fixup_eh_fallthru): Likewise.
5611 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5612 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5613 * config/frv/frv.c (frv_function_prologue): Likewise.
5614 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5616 (note_outside_basic_block_p): Remove all references to
5617 NOTE_INSN_CALL_ARG_LOCATION.
5618 * gengtype.c (adjust_field_rtx_def): Likewise.
5619 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5621 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5622 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5625 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5628 PR tree-optimization/84452
5629 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5630 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5633 2018-02-19 Martin Liska <mliska@suse.cz>
5636 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5638 2018-02-19 Martin Liska <mliska@suse.cz>
5639 Richard Sandiford <richard.sandiford@linaro.org>
5641 PR tree-optimization/82491
5642 * gimple-fold.c (get_base_constructor): Make earlier bail out
5645 2018-02-19 Carl Love <cel@us.ibm.com>
5647 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5648 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5650 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5651 P8V_BUILTIN_VEC_NEG.
5653 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5655 * config/rl78/rl78.md (movdf): New define expand.
5657 2018-02-19 Martin Liska <mliska@suse.cz>
5660 * doc/invoke.texi: Fix typo.
5661 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5663 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5665 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5666 handle rs6000_single_float and rs6000_double_float specially for
5669 2018-02-16 Jeff Law <law@redhat.com>
5671 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5674 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5677 * ipa-inline.c (inline_small_functions): Fix a typo.
5679 2018-02-16 Nathan Sidwell <nathan@acm.org>
5681 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5683 2018-02-16 Carl Love <cel@us.ibm.com>
5685 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5686 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5687 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5688 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5689 expansion to P8V_BUILTIN_VEC_FLOAT2.
5691 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5693 PR rtl-optimization/70023
5694 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5695 src_regno into account.
5697 2018-02-16 Carl Love <cel@us.ibm.com>
5699 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5700 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5701 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5702 * config/rs6000/rs6000.c: Remove case statements for
5703 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5704 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5705 and P9V_BUILTIN_VEC_VINSERT4B.
5706 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5707 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5708 * config/rs6000/vsx.md:
5709 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5712 2018-02-16 Carl Love <cel@us.ibm.com>
5714 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5716 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5718 * config/rs6000/rs6000-c.c: Add the definitions for
5719 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5720 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5721 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5722 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5723 definition for insert4b and define insn *insert3b_internal.
5724 * doc/extend.texi: Add documentation for vec_extract4b.
5726 2018-02-16 Nathan Sidwell <nathan@acm.org>
5728 * doc/extend.texi (Backwards Compatibility): Mention friend
5729 injection. Note for-scope is deprecated.
5730 * doc/invoke.texi (-ffriend-injection): Deprecate.
5732 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5734 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5735 that moved to I2, also allow destinations that are a paradoxical
5736 subreg (instead of a normal reg).
5738 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5741 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5744 2018-02-16 Richard Biener <rguenther@suse.de>
5746 PR tree-optimization/84037
5747 PR tree-optimization/84016
5749 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5750 Adjust vec_construct for the fact we need additional higher latency
5751 128bit inserts for AVX256 and AVX512 vector builds.
5752 (ix86_add_stmt_cost): Scale vector construction cost for
5755 2018-02-16 Richard Biener <rguenther@suse.de>
5757 PR tree-optimization/84417
5758 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5759 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5760 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5762 2018-02-16 Martin Liska <mliska@suse.cz>
5765 * internal-fn.def (ASAN_CHECK): Set proper flags.
5766 (ASAN_MARK): Likewise.
5768 2018-02-16 Julia Koval <julia.koval@intel.com>
5770 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5771 from PTA_CANNONLAKE.
5773 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5776 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5777 Use ++iter rather than iter++ for std::list iterators.
5778 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5779 defer deleting them until all nodes in the forest are processed. Do
5780 free even leaf nodes. Change to_process into auto_vec.
5783 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5784 * vec.h (vec_default_construct): Use memset instead of placement new
5785 if BROKEN_VALUE_INITIALIZATION is defined.
5786 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5787 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5790 PR rtl-optimization/83723
5791 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5792 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5793 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5795 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5797 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5799 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5801 PR rtl-optimization/81443
5802 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5803 from inner REGs to paradoxical SUBREGs.
5805 2018-02-16 Richard Biener <rguenther@suse.de>
5807 PR tree-optimization/84399
5808 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5809 For operands we can analyze at their definition make sure we can
5810 analyze them at each use as well.
5812 2018-02-16 Richard Biener <rguenther@suse.de>
5814 PR tree-optimization/84190
5815 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5816 volatile accesses if the decl isn't volatile.
5818 2018-02-15 Jason Merrill <jason@redhat.com>
5820 PR c++/84314 - ICE with templates and fastcall attribute.
5821 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5822 TYPE_CANONICAL on an existing type.
5824 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5826 PR tree-optimization/84383
5827 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5828 dstoff nor call operand_equal_p if dstbase is NULL.
5830 PR tree-optimization/84334
5831 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5832 also a CONSTANT_CLASS_P, punt.
5834 2018-02-14 Jim Wilson <jimw@sifive.com>
5836 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5837 first SMALL_OPERAND check. New local min_second_step. Move assert
5838 to where locals are set. Add TARGET_RVC support.
5839 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5841 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5843 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5845 2018-02-14 Martin Sebor <msebor@redhat.com>
5847 PR tree-optimization/83698
5848 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5849 arrays constrain the offset range to their bounds.
5850 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5851 (builtin_access::overlap): Avoid setting the size of overlap if it's
5853 (maybe_diag_overlap): Also consider arrays when deciding what values
5854 of offsets to include in diagnostics.
5856 2018-02-14 Martin Sebor <msebor@redhat.com>
5859 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5860 that correspond to the kind of a declaration.
5862 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5865 * config/pa/pa.md: Load address of PIC label using the linkage table
5866 if the label is nonlocal.
5868 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5870 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5871 warning message if user requests -maltivec=be.
5872 * doc/invoke.texi: Document deprecation of -maltivec=be.
5874 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5877 * config/rs6000/rs6000-c.c: Update definitions for
5878 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5879 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5881 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5884 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5885 add _get_ssp intrinsics. Remove argument from
5886 __builtin_ia32_rdssp[d|q].
5887 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5888 * config/i386/i386-builtin.def: Remove argument from
5889 __builtin_ia32_rdssp[d|q].
5890 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5891 ix86_expand_special_args_builtin for _rdssp[d|q].
5892 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5893 Clear register before usage.
5894 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5895 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5897 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5899 PR tree-optimization/84357
5900 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5901 operand 1 of an ARRAY_REF too.
5903 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5906 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5907 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5909 (set_of_reg): New struct.
5910 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5911 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5912 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5914 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5915 Split into bitclr, bitset, bitinvert patterns if appropriate.
5916 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5917 use rx_fuse_in_memory_bitop.
5918 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5919 to named insn, correct maximum insn length.
5921 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5924 * machmode.def: Define a complex mode for PARTIAL_INT.
5925 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5927 * doc/rtl.texi: Document CSPImode.
5928 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5930 (msp430_hard_regno_nregs_with_padding): Likewise.
5932 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5935 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5937 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5939 PR rtl-optimization/84169
5940 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5941 we generated a parallel as new i3 and we split that to new i2 and i3
5942 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5943 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5944 those to i2, not i1. Partially rewrite this scan code.
5946 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5949 * stor-layout.c (place_field): For variable length fields, adjust
5950 offset_align afterwards not just based on the field's alignment,
5951 but also on the size.
5954 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5955 of exps and logs in the use_exp2 case.
5957 2018-02-13 Jeff Law <law@redhat.com>
5959 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5962 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5965 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5969 * common.opt (gas-loc-support, gas-locview-support): New.
5970 (ginline-points, ginternal-reset-location-views): New.
5971 * doc/invoke.texi: Document them. Use @itemx where intended.
5972 (gvariable-location-views): Adjust.
5973 * target.def (reset_location_view): New.
5974 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5975 (TARGET_RESET_LOCATION_VIEW): New.
5976 * doc/tm.texi: Rebuilt.
5977 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5978 (dwarf2out_default_as_locview_support): New.
5979 (output_asm_line_debug_info): Use option variables.
5980 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5981 (output_loc_list): Likewise.
5982 (add_high_low_attributes): Check option variables.
5983 Don't output entry view attribute in strict mode.
5984 (gen_inlined_subroutine_die): Check option variables.
5985 (dwarf2out_inline_entry): Likewise.
5986 (init_sections_and_labels): Likewise.
5987 (dwarf2out_early_finish): Likewise.
5988 (maybe_reset_location_view): New, from...
5989 (dwarf2out_var_location): ... here. Call it.
5990 * debug.h (dwarf2out_default_as_loc_support): Declare.
5991 (dwarf2out_default_as_locview_support): Declare.
5992 * hooks.c (hook_int_rtx_insn_0): New.
5993 * hooks.h (hook_int_rtx_insn_0): Declare.
5994 * toplev.c (process_options): Take -gas-loc-support and
5995 -gas-locview-support from dwarf2out. Enable
5996 -gvariable-location-views by default only with locview
5997 assembler support. Enable -ginternal-reset-location-views by
5998 default only if the target defines the corresponding hook.
5999 Enable -ginline-points by default if location views are
6000 enabled; force it disabled if statement frontiers are
6002 * tree-inline.c (expand_call_inline): Check option variables.
6003 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6005 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6007 PR tree-optimization/84321
6008 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6009 handling. Also check whether the anti-range contains any values
6010 that satisfy the mask; switch to a VR_RANGE if not.
6012 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6015 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6017 2018-02-13 Martin Jambor <mjambor@suse.cz>
6020 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6021 of call statements, also set location of a load to a temporary.
6023 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6025 * config/rl78/rl78.c (add_vector_labels): New function.
6026 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6027 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6028 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6029 which checks that no arguments are passed.
6030 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6031 * doc/extend.texi: Documentation for the new attribute.
6033 2018-02-13 Andreas Schwab <schwab@suse.de>
6035 * config/riscv/linux.h (CPP_SPEC): Define.
6037 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6040 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6041 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6042 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6043 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6044 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6045 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6046 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6047 temporarily for AES and PCLMUL builtins.
6049 PR tree-optimization/84339
6050 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6051 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6055 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6056 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6057 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6059 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6061 * omp-simd-clone.h: New file.
6062 * omp-simd-clone.c: Include omp-simd-clone.h.
6063 (expand_simd_clones): No longer static.
6064 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6065 cgraph.h and omp-simd-clone.h.
6066 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6067 (vect_recog_widen_shift_pattern): Formatting fix.
6068 (vect_pattern_recog_1): Don't check optab for calls.
6071 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6072 operands[2] into a REG before using gen_lowpart on it.
6074 2018-02-12 Jeff Law <law@redhat.com>
6077 * config/sh/sh.c (find_barrier): Consider a sibling call
6080 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6081 successfully back substituting a reg.
6083 2018-02-12 Richard Biener <rguenther@suse.de>
6085 PR tree-optimization/84037
6086 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6087 parameter, move visited init to caller.
6088 (vect_slp_analyze_operations): Separate cost from validity
6089 check, initialize visited once for all instances.
6090 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6092 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6094 (vect_model_promotion_demotion_cost): Likewise.
6095 (vectorizable_bswap): Guard cost modeling with !slp_node
6096 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6098 (vectorizable_call): Likewise.
6099 (vectorizable_conversion): Likewise.
6100 (vectorizable_assignment): Likewise.
6101 (vectorizable_shift): Likewise.
6102 (vectorizable_operation): Likewise.
6103 (vectorizable_store): Likewise.
6104 (vectorizable_load): Likewise.
6105 (vectorizable_condition): Likewise.
6106 (vectorizable_comparison): Likewise.
6108 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6111 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6112 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6113 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6115 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6118 * params.def (inline-min-speedup): Increase from 8 to 15.
6119 (max-inline-insns-auto): Decrease from 40 to 30.
6120 * ipa-split.c (consider_split): Add some buffer for function to
6121 be considered inlining candidate.
6122 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6125 2018-02-12 Richard Biener <rguenther@suse.de>
6127 PR tree-optimization/84037
6128 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6129 matched stmts if we cannot swap the non-matched ones.
6131 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6133 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6134 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6135 _mm_maskz_scalef_round_ss): New intrinsics.
6136 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6137 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6138 __builtin_ia32_scalefss_round): Remove.
6139 (__builtin_ia32_scalefsd_mask_round,
6140 __builtin_ia32_scalefss_mask_round): New intrinsics.
6141 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6142 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6143 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6144 "<round_constraint>")): Changed to ...
6145 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6146 "<round_scalar_constraint>")): ... this.
6147 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6148 %0, %1, %2<round_op3>}"): Changed to ...
6149 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6150 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6151 %2<round_scalar_mask_op3>}"): ... this.
6152 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6154 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6156 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6157 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6158 (_mm_maskz_sqrt_round_ss): New intrinsics.
6159 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6160 (__builtin_ia32_sqrtsd_mask_round)
6161 (__builtin_ia32_sqrtss_mask_round): New builtins.
6162 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6163 (__builtin_ia32_sqrtss_round): Remove.
6164 (__builtin_ia32_sqrtsd_mask_round)
6165 (__builtin_ia32_sqrtss_mask_round): New builtins.
6166 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6167 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6168 ((match_operand:VF_128 1 "vector_operand"
6169 "xBm,<round_constraint>")): Changed to ...
6170 ((match_operand:VF_128 1 "vector_operand"
6171 "xBm,<round_scalar_constraint>")): ... this.
6172 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6173 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6174 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6175 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6176 %<iptr>1<round_scalar_mask_op3>}): ... this.
6177 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6178 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6180 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6183 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6184 Cast vec_cmpeq result to correct type.
6185 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6186 Cast vec_cmpgt result to correct type.
6188 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6190 * final.c (final_scan_insn_1): Renamed from...
6191 (final_scan_insn): ... this. New wrapper, to recover
6192 seen from the outermost call in recursive ones.
6193 * config/sparc/sparc.c (output_return): Drop seen from call.
6194 (output_sibcall): Likewise.
6195 * config/visium/visium.c (output_branch): Likewise.
6197 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6199 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6202 2018-02-10 Alan Modra <amodra@gmail.com>
6205 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6206 Specify LR as an input.
6208 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6211 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6212 remove_member_access_dummy_vars): New functions.
6213 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6214 lower_omp_1, execute_lower_omp): Use them.
6216 PR rtl-optimization/84308
6217 * shrink-wrap.c (spread_components): Release todo vector.
6219 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6221 PR rtl-optimization/57193
6222 * ira-color.c (struct allocno_color_data): Add member
6223 conflict_allocno_hard_prefs.
6224 (update_conflict_allocno_hard_prefs): New.
6225 (bucket_allocno_compare_func): Add a preference based on
6226 conflict_allocno_hard_prefs.
6227 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6228 (color_allocnos): Remove a dead code. Initiate
6229 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6231 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6234 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6235 constraint from =wa to wa. Avoid a subreg on the output operand,
6236 instead use a pseudo and subreg it in a move.
6237 (p9_xxbrd_<mode>): Changed to ...
6238 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6239 (p9_xxbrd_v2df): New expander.
6240 (p9_xxbrw_<mode>): Changed to ...
6241 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6242 (p9_xxbrw_v4sf): New expander.
6244 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6246 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6248 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6251 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6252 multiply in 32-bit mode.
6253 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6254 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6257 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6259 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6260 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6261 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6262 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6264 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6267 * dwarf2out.c (is_trivial_indirect_ref): New function.
6268 (dwarf2out_late_global_decl): Do not generate a location
6269 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6270 and that are not defined in the current unit.
6272 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6274 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6275 instead of a libcall for UNORDERED.
6277 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6280 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6281 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6283 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6286 * config/s390/s390.c (s390_set_current_function): Invoke
6287 s390_indirect_branch_settings also if fndecl didn't change.
6289 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6291 * config/rs6000/rs6000.md (blockage): Set length to zero.
6293 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6295 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6297 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6300 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6301 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6305 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6306 PARALLEL incoming that failed vt_get_decl_and_offset check.
6309 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6310 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6312 (get_variable_section): For decls in named .bss* sections pass true as
6313 second argument to bss_initializer_p.
6315 2018-02-09 Marek Polacek <polacek@redhat.com>
6316 Jakub Jelinek <jakub@redhat.com>
6319 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6320 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6321 Sync some changes from cxx_fold_indirect_ref.
6323 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6325 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6327 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6328 (BLOCK_INLINE_ENTRY_LABEL): New.
6329 (dwarf2out_var_location): Disregard inline entry markers.
6330 (inline_entry_data): New struct.
6331 (inline_entry_data_hasher): New hashtable type.
6332 (inline_entry_data_hasher::hash): New.
6333 (inline_entry_data_hasher::equal): New.
6334 (inline_entry_data_table): New variable.
6335 (add_high_low_attributes): Add DW_AT_entry_pc and
6336 DW_AT_GNU_entry_view attributes if a pending entry is found
6337 in inline_entry_data_table. Add old entry_pc attribute only
6338 if debug nonbinding markers are disabled.
6339 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6340 markers are enabled.
6341 (block_within_block_p, dwarf2out_inline_entry): New.
6342 (dwarf2out_finish): Check that no entries remained in
6343 inline_entry_data_table.
6344 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6345 (final_scan_insn, notice_source_line): Likewise.
6346 (rest_of_clean_state): Skip inline entry markers.
6347 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6349 * gimple.c (gimple_build_debug_inline_entry): New.
6350 * gimple.h (enum gimple_debug_subcode): Add
6351 GIMPLE_DEBUG_INLINE_ENTRY.
6352 (gimple_build_debug_inline_entry): Declare.
6353 (gimple_debug_inline_entry_p): New.
6354 (gimple_debug_nonbind_marker_p): Adjust.
6355 * insn-notes.def (INLINE_ENTRY): New.
6356 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6357 inline entry marker notes.
6358 (print_insn): Likewise.
6359 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6360 (INSN_DEBUG_MARKER_KIND): Likewise.
6361 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6362 * tree-inline.c (expand_call_inline): Build and insert
6363 debug_inline_entry stmt.
6364 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6365 inline entry blocks early, if nonbind markers are enabled.
6366 (dump_scope_block): Dump fragment info.
6367 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6368 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6369 (gimple_build_debug_inline_entry): New.
6370 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6371 Enable/disable inline entry points too.
6372 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6373 (DEBUG_INSN): Describe inline entry markers.
6375 * common.opt (gvariable-location-views): New.
6376 (gvariable-location-views=incompat5): New.
6377 * config.in: Rebuilt.
6378 * configure: Rebuilt.
6379 * configure.ac: Test assembler for view support.
6380 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6381 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6382 * dwarf2out.c (var_loc_view): New typedef.
6383 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6384 (dwarf2out_locviews_in_attribute): New.
6385 (dwarf2out_locviews_in_loclist): New.
6386 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6387 (enum dw_line_info_opcode): Add LI_adv_address.
6388 (struct dw_line_info_table): Add view.
6389 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6390 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6391 (zero_view_p): New variable.
6392 (ZERO_VIEW_P): New macro.
6393 (output_asm_line_debug_info): New.
6394 (struct var_loc_node): Add view.
6395 (add_AT_view_list, AT_loc_list): New.
6396 (add_var_loc_to_decl): Add view param. Test it against last.
6397 (new_loc_list): Add view params. Record them.
6398 (AT_loc_list_ptr): Handle loc and view lists.
6399 (view_list_to_loc_list_val_node): New.
6400 (print_dw_val): Handle dw_val_class_view_list.
6401 (size_of_die): Likewise.
6402 (value_format): Likewise.
6403 (loc_list_has_views): New.
6404 (gen_llsym): Set vl_symbol too.
6405 (maybe_gen_llsym, skip_loc_list_entry): New.
6406 (dwarf2out_maybe_output_loclist_view_pair): New.
6407 (output_loc_list): Output view list or entries too.
6408 (output_view_list_offset): New.
6409 (output_die): Handle dw_val_class_view_list.
6410 (output_dwarf_version): New.
6411 (output_compilation_unit_header): Use it.
6412 (output_skeleton_debug_sections): Likewise.
6413 (output_rnglists, output_line_info): Likewise.
6414 (output_pubnames, output_aranges): Update version comments.
6415 (output_one_line_info_table): Output view numbers in asm comments.
6416 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6417 Call maybe_gen_llsym.
6418 (loc_list_from_tree_1): Adjust.
6419 (add_AT_location_description): Create view list attribute if
6420 needed, check it's absent otherwise.
6421 (convert_cfa_to_fb_loc_list): Adjust.
6422 (maybe_emit_file): Call output_asm_line_debug_info for test.
6423 (dwarf2out_var_location): Reset views as needed. Precompute
6424 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6425 attribute. Set view.
6426 (new_line_info_table): Reset next view.
6427 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6428 (dwarf2out_source_line): Likewise. Output view resets and labels to
6429 the assembler, or select appropriate line info opcodes.
6430 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6431 (optimize_string_length): Catch it. Adjust.
6432 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6433 dw_val_class_view_list, and remove it if no longer needed.
6434 (hash_loc_list): Hash view numbers.
6435 (loc_list_hasher::equal): Compare them.
6436 (optimize_location_lists): Check whether a view list symbol is
6437 needed, and whether the locview attribute is present, and
6438 whether they match. Remove the locview attribute if no longer
6440 (index_location_lists): Call skip_loc_list_entry for test.
6441 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6442 Use output_dwarf_version.
6443 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6444 (struct dw_val_node): Add val_view_list.
6445 * final.c (SEEN_NEXT_VIEW): New.
6446 (set_next_view_needed): New.
6447 (clear_next_view_needed): New.
6448 (maybe_output_next_view): New.
6449 (final_start_function): Rename to...
6450 (final_start_function_1): ... this. Take pointer to FIRST,
6451 add SEEN parameter. Emit param bindings in the initial view.
6452 (final_start_function): Reintroduce SEEN-less interface.
6453 (final): Rename to...
6454 (final_1): ... this. Take SEEN parameter. Output final pending
6455 next view at the end.
6456 (final): Reintroduce seen-less interface.
6457 (final_scan_insn): Output pending next view before switching
6458 sections or ending a block. Mark the next view as needed when
6459 outputting variable locations. Notify debug backend of section
6460 changes, and of location view changes.
6461 (rest_of_handle_final): Adjust.
6462 * toplev.c (process_options): Autodetect value for debug variable
6463 location views option. Warn on incompat5 without -gdwarf-5.
6464 * doc/invoke.texi (gvariable-location-views): New.
6465 (gvariable-location-views=incompat5): New.
6466 (gno-variable-location-views): New.
6468 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6470 PR tree-optimization/84136
6471 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6472 that the result of find_edge is non-NULL.
6474 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6477 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6478 storing integer register in SImode. Fix cost of 256 and 512
6479 byte aligned SSE register store.
6481 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6483 * config/i386/i386.c (ix86_multiplication_cost): Fix
6484 multiplication cost for TARGET_AVX512DQ.
6486 2018-02-08 Marek Polacek <polacek@redhat.com>
6488 PR tree-optimization/84238
6489 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6492 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6494 PR tree-optimization/84265
6495 * tree-vect-stmts.c (vectorizable_store): Don't treat
6496 VMAT_CONTIGUOUS accesses as grouped.
6497 (vectorizable_load): Likewise.
6499 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6501 PR tree-optimization/81635
6502 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6503 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6504 (test_round_for_mask): New functions.
6505 (wide_int_cc_tests): Call test_round_for_mask.
6506 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6507 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6508 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6509 range returned by get_range_info.
6511 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6514 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6515 * symtab.c: Include builtins.h
6516 (symtab_node::output_to_lto_symbol_table_p): Move here
6517 from lto-streamer-out.c:output_symbol_p.
6518 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6519 (output_symbol_p): Move all logic to symtab.c
6520 (produce_symtab): Update.
6522 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6524 * config/s390/s390-opts.h (enum indirect_branch): Define.
6525 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6526 (s390_indirect_branch_via_thunk)
6527 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6528 (enum s390_indirect_branch_type): Define.
6529 * config/s390/s390.c (struct s390_frame_layout, struct
6530 machine_function): Remove.
6531 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6532 (indirect_branch_table_label_no, indirect_branch_table_name):
6534 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6535 (enum s390_indirect_branch_option): Define.
6536 (s390_return_addr_from_memory): New function.
6537 (s390_handle_string_attribute): New function.
6538 (s390_attribute_table): Add new attribute handler.
6539 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6540 (s390_indirect_branch_via_thunk): New function.
6541 (s390_indirect_branch_via_inline_thunk): New function.
6542 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6543 sibling call optimization for non z10 compiles.
6544 (s390_emit_call): Force indirect branch target to be a single
6545 register. Add r1 clobber for non-z10 compiles.
6546 (s390_emit_epilogue): Emit return jump via return_use expander.
6547 (s390_reorg): Handle JUMP_INSNs as execute targets.
6548 (s390_option_override_internal): Perform validity checks for the
6549 new command line options.
6550 (s390_indirect_branch_attrvalue): New function.
6551 (s390_indirect_branch_settings): New function.
6552 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6553 (s390_output_indirect_thunk_function): New function.
6554 (s390_code_end): Implement target hook.
6555 (s390_case_values_threshold): Implement target hook.
6556 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6558 * config/s390/s390.h (struct s390_frame_layout)
6559 (struct machine_function): Move here from s390.c.
6560 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6561 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6562 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6563 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6564 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6565 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6566 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6567 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6568 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6569 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6570 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6571 (mnemonic attribute): Add values which aren't recognized
6573 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6574 pattern for branch conversion. Fix mnemonic attribute.
6575 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6576 indirect branch via thunk if requested.
6577 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6578 ("*indirect_jump"): Disable for branch conversion using out of
6580 ("indirect_jump_via_thunk<mode>_z10")
6581 ("indirect_jump_via_thunk<mode>")
6582 ("indirect_jump_via_inlinethunk<mode>_z10")
6583 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6584 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6585 ("casesi_jump_via_inlinethunk<mode>_z10")
6586 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6587 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6588 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6589 ("*indirect2_jump"): Disable for branch conversion.
6590 ("casesi_jump"): Turn into expander and expand patterns for branch
6592 ("return_use"): New expander.
6593 ("*return"): Emit return via thunk and rename it to ...
6594 ("*return<mode>"): ... this one.
6595 * config/s390/s390.opt: Add new options and and enum for the
6598 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6600 * lra-constraints.c (match_reload): Unconditionally use
6601 gen_lowpart_SUBREG, rather than selecting between that
6602 and equivalent gen_rtx_SUBREG code.
6604 2018-02-08 Richard Biener <rguenther@suse.de>
6606 PR tree-optimization/84233
6607 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6608 changed flag instead of boguously re-using phi_inserted.
6610 2018-02-08 Martin Jambor <mjambor@suse.cz>
6612 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6613 static local variables.
6615 2018-02-08 Richard Biener <rguenther@suse.de>
6617 PR tree-optimization/84278
6618 * tree-vect-stmts.c (vectorizable_store): When looking for
6619 smaller vector types to perform grouped strided loads/stores
6620 make sure the mode is supported by the target.
6621 (vectorizable_load): Likewise.
6623 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6625 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6626 Increase LDP/STP opportunities by adding adjacent callee-saves.
6628 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6630 PR rtl-optimization/84068
6631 PR rtl-optimization/83459
6632 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6634 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6636 PR tree-optimization/84224
6637 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6638 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6641 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6644 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6645 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6646 count, remove one USE.
6648 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6650 * doc/install.texi (Configuration): Document the
6651 --with-long-double-format={ibm,ieee} PowerPC configuration
6655 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6656 Convert from define_expand to be define_insn_and_split. Rework
6657 float/double/_Float128 conversions to QI/HI/SImode to work with
6658 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6659 conversions to QI/HImode types did a store and then a load to
6660 truncate the value. For conversions to VSX registers, don't split
6661 the insn, instead emit the code directly. Use the code iterator
6662 any_fix to combine signed and unsigned conversions.
6663 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6664 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6665 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6666 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6667 (fix_<mode>di2_hw): Likewise.
6668 (fixuns_<mode>di2_hw): Likewise.
6669 (fix_<mode>si2_hw): Likewise.
6670 (fixuns_<mode>si2_hw): Likewise.
6671 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6672 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6673 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6674 fix<uns>_trunc<SFDF:mode>si2_p8.
6675 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6677 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6678 (fix<uns>_<mode>_mem): Likewise.
6679 (fctiw<u>z_<mode>_mem): Likewise.
6680 (fix<uns>_<mode>_mem): Likewise.
6681 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6682 the register allocator from doing a direct move to the GPRs to do
6683 a store, and instead use the ISA 3.0 store byte/half-word from
6684 vector register instruction. For IEEE 128-bit floating point,
6685 also optimize stores of 32-bit ints.
6686 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6688 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6690 * genextract.c (push_pathstr_operand): New function to support
6692 (walk_rtx): Call push_pathstr_operand.
6693 (print_path): Support [a-zA-Z].
6695 2018-02-07 Richard Biener <rguenther@suse.de>
6697 PR tree-optimization/84037
6698 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6699 (cse_and_gimplify_to_preheader): Declare.
6700 (vect_get_place_in_interleaving_chain): Likewise.
6701 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6703 (_loop_vec_info::~_loop_vec_info): Delete it.
6704 (cse_and_gimplify_to_preheader): New function.
6705 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6706 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6707 (vectorizable_load): Likewise. For grouped stores always base
6708 the IV on the first element.
6709 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6710 condition before gimplifying.
6712 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6714 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6715 *DIV_EXPR and *MOD_EXPR.
6717 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6720 * config/i386/i386.c (ix86_option_override_internal): Mask out
6721 the CF_SET bit when checking -fcf-protection.
6723 2018-02-07 Tom de Vries <tom@codesourcery.com>
6726 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6729 2018-02-07 Richard Biener <rguenther@suse.de>
6731 PR tree-optimization/84204
6732 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6735 PR tree-optimization/84205
6736 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6737 special-case isl_ast_op_zdiv_r.
6739 PR tree-optimization/84223
6740 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6741 Only add conditions from within the region.
6742 (gather_bbs::after_dom_children): Adjust.
6744 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6747 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6748 * config/avr/avr.md: Only post-reload split REG-REG moves if
6749 either register is GENERAL_REG_P.
6751 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6753 PR tree-optimization/84235
6754 * tree-ssa-scopedtables.c
6755 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6756 if the subtraction is performed in floating point type where NaNs are
6757 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6758 build 1. Formatting fix.
6760 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6763 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6764 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6765 and skip it regardless of bb boundaries. Use CALL_P macro,
6766 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6767 unnecessarily, formatting fix.
6769 2018-02-06 Michael Collison <michael.collison@arm.com>
6771 * config/arm/thumb2.md:
6772 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6773 (*thumb_mov_notscc): Ditto.
6775 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6778 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6779 unsigned_fix, not "s".
6781 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6783 * configure.ac (gcc_fn_eh_frame_ro): New function.
6784 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6785 correct .eh_frame permissions.
6786 * configure: Regenerate.
6788 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6790 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6793 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6795 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6796 Display warning message for -mno-speculate-indirect-jumps.
6798 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6800 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6802 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6804 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6806 PR tree-optimization/84225
6807 * tree-eh.c (find_trapping_overflow): Only call
6808 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6810 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6813 * config/i386/i386.c: Reimplement the check of possible options
6814 -mibt/-mshstk conbination. Change error messages.
6815 * doc/invoke.texi: Fix a typo: remove extra '='.
6817 2018-02-06 Marek Polacek <polacek@redhat.com>
6819 PR tree-optimization/84228
6820 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6822 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6825 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6826 emitted arch directives.
6827 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6828 __ARM_FEATURE_COPROC before changing architectures.
6830 2018-02-06 Richard Biener <rguenther@suse.de>
6832 * config/i386/i386.c (print_reg): Fix typo.
6833 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6835 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6837 * configure: Regenerate.
6839 2018-02-05 Martin Sebor <msebor@redhat.com>
6841 PR tree-optimization/83369
6842 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6845 2018-02-05 Martin Liska <mliska@suse.cz>
6847 * doc/invoke.texi: Cherry-pick upstream r323995.
6849 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6851 * ira.c (ira_init_register_move_cost): Adjust comment.
6853 2018-02-05 Martin Liska <mliska@suse.cz>
6855 PR gcov-profile/84137
6856 * doc/gcov.texi: Fix typo in documentation.
6858 2018-02-05 Martin Liska <mliska@suse.cz>
6860 PR gcov-profile/83879
6861 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6862 using dlopen functionality.
6864 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6866 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6867 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6868 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6869 _mm_maskz_range_round_ss): New intrinsics.
6870 (__builtin_ia32_rangesd128_round)
6871 (__builtin_ia32_rangess128_round): Remove.
6872 (__builtin_ia32_rangesd128_mask_round,
6873 __builtin_ia32_rangess128_mask_round): New builtins.
6874 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6875 __builtin_ia32_rangess128_round): Remove.
6876 (__builtin_ia32_rangesd128_mask_round,
6877 __builtin_ia32_rangess128_mask_round): New builtins.
6878 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6879 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6880 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6881 "<round_saeonly_constraint>")): Changed to ...
6882 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6883 "<round_saeonly_scalar_constraint>")): ... this.
6884 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6885 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6886 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6887 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6888 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6890 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6892 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6894 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6895 Remove all values except native, 8540 and 8548.
6897 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6899 * config/i386/i386.c (ix86_output_function_return): Pass
6900 INVALID_REGNUM, instead of -1, as invalid register number to
6901 indirect_thunk_name and output_indirect_thunk.
6903 2018-02-02 Julia Koval <julia.koval@intel.com>
6905 * config.gcc: Add -march=icelake.
6906 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6907 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6908 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6909 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6910 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6911 (processor_target_table): Add icelake.
6912 (ix86_option_override_internal): Handle new PTAs.
6913 (get_builtin_code_for_version): Handle icelake.
6914 (M_INTEL_COREI7_ICELAKE): New.
6915 (fold_builtin_cpu): Handle icelake.
6916 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6917 * doc/invoke.texi: Add -march=icelake.
6919 2018-02-02 Julia Koval <julia.koval@intel.com>
6921 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6922 to wide_int_bitmask.
6923 * wide-int-bitmask.h: New.
6925 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6928 * config/i386/i386.md: Replace Pmode with word_mode in
6929 builtin_setjmp_setup and builtin_longjmp to support x32.
6931 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6935 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6937 (rs6000_supported_cpu_names): New static variable.
6938 (linux_cpu_translation_table): Likewise.
6939 (elf_platform) <cpu>: Define new static variable and use it.
6940 Translate kernel AT_PLATFORM name to canonical name if needed.
6941 Error if platform name is unknown.
6943 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6946 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6948 2018-02-01 Jeff Law <law@redhat.com>
6951 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6952 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6953 the scratch if RELEASE_VIA_POP is false.
6954 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6955 If we have to save a temporary register, decrement SIZE appropriately.
6956 Pass new arguments to release_scratch_register_on_entry.
6957 (ix86_adjust_stack_and_probe): Likewise.
6958 (ix86_emit_probe_stack_range): Pass new arguments to
6959 release_scratch_register_on_entry.
6961 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6963 PR rtl-optimization/84157
6964 * combine.c (change_zero_ext): Use REG_P predicate in
6965 front of HARD_REGISTER_P predicate.
6967 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6969 * config/avr/avr.c (avr_option_override): Move disabling of
6970 -fdelete-null-pointer-checks to...
6971 * common/config/avr/avr-common.c (avr_option_optimization_table):
6974 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6976 PR tree-optimization/81635
6977 * tree-data-ref.c (split_constant_offset_1): For types that
6978 wrap on overflow, try to use range info to prove that wrapping
6981 2018-02-01 Renlin Li <renlin.li@arm.com>
6984 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6986 (aarch64_register_move_cost): Likewise.
6987 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6989 (REG_CLASS_NAMES): Likewise.
6990 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6991 TAILCALL_ADDR_REGS. Remove IP registers.
6992 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6994 2018-02-01 Richard Biener <rguenther@suse.de>
6996 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6997 for specifying RPO order and allow NULL for that.
6998 * domwalk.c (dom_walker::dom_walker): Likewise.
6999 (dom_walker::walk): Handle NULL RPO order.
7000 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7002 (rewrite_update_dom_walker): Likewise.
7003 (mark_def_dom_walker): Likewise.
7005 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7007 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7008 (aarch64_maybe_expand_sve_subreg_move): Declare.
7009 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7010 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7012 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7013 that are semantically a reverse operation.
7014 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7015 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7016 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7018 (aarch64_can_change_mode_class): For big-endian, forbid changes
7019 between two SVE modes if they have different element sizes.
7021 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7023 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7024 the TImode handling for big-endian targets.
7026 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7028 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7029 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7031 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7032 Remove BSWAP handing for big-endian targets and use the form of
7033 LD1RQ appropariate for the mode.
7035 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7037 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7038 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7041 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7044 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7045 check for operands that need to go through aarch64_sve_reload_be.
7047 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7049 PR tree-optimization/81661
7050 PR tree-optimization/84117
7051 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7052 * tree-eh.c: Include gimplify.h.
7053 (find_trapping_overflow, replace_trapping_overflow,
7054 rewrite_to_non_trapping_overflow): New functions.
7055 * tree-vect-loop.c: Include tree-eh.h.
7056 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7057 * tree-data-ref.c: Include tree-eh.h.
7058 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7060 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7062 PR rtl-optimization/84123
7063 * combine.c (change_zero_ext): Check if hard register satisfies
7064 can_change_dest_mode before calling gen_lowpart_SUBREG.
7066 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7069 * ira.c (ira_init_register_move_cost): Remove assert.
7071 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7073 PR rtl-optimization/84071
7074 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7075 * doc/tm.texi: Regenerate.
7077 2018-01-31 Richard Biener <rguenther@suse.de>
7079 PR tree-optimization/84132
7080 * tree-data-ref.c (analyze_miv_subscript): Properly
7081 check whether evolution_function_is_affine_multivariate_p
7082 before calling gcd_of_steps_may_divide_p.
7084 2018-01-31 Julia Koval <julia.koval@intel.com>
7087 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7088 * config/i386/i386.md (rdpid_rex64) New.
7089 (rdpid): Make 32bit only.
7091 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7094 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7095 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7097 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7100 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7102 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7104 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7106 PR rtl-optimization/84071
7107 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7108 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7110 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7112 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7113 (arc_attribute_table): Add 'aux' attribute.
7114 (arc_in_small_data_p): Consider aux like variables.
7115 (arc_is_aux_reg_p): New function.
7116 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7117 (arc_get_aux_arg): New function.
7118 (prepare_move_operands): Handle aux-register access.
7119 (arc_handle_aux_attribute): New function.
7120 * doc/extend.texi (ARC Variable attributes): Add subsection.
7122 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7124 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7125 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7126 (arc_attribute_table): Add 'uncached' attribute.
7127 (arc_print_operand): Print '.di' flag for uncached memory
7129 (arc_in_small_data_p): Do not consider for small data the uncached
7131 (arc_is_uncached_mem_p): New function.
7132 * config/arc/predicates.md (compact_store_memory_operand): Check
7133 for uncached memory accesses.
7134 (nonvol_nonimm_operand): Likewise.
7135 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7137 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7140 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7141 falign-loops=): Add Optimization flag.
7143 2018-01-30 Jeff Law <law@redhat.com>
7146 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7147 INT_REGISTERS_SAVED. Check it prior to calling
7148 get_scratch_register_on_entry.
7149 (ix86_adjust_stack_and_probe): Similarly.
7150 (ix86_emit_probe_stack_range): Similarly.
7151 (ix86_expand_prologue): Corresponding changes.
7153 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7156 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7157 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7159 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7162 * lra-constraints.c (curr_insn_transform): Process AND in the
7165 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7167 PR rtl-optimization/83986
7168 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7169 dependence against last_pending_memory_flush in addition to
7172 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7174 PR tree-optimization/81611
7175 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7178 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7181 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7184 2018-01-30 Richard Biener <rguenther@suse.de>
7185 Jakub Jelinek <jakub@redhat.com>
7187 PR tree-optimization/84111
7188 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7189 inner loops added during recursion, as they don't have up-to-date
7192 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7195 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7196 (can_inline_edge_by_limits_p): ... here.
7197 (can_early_inline_edge_p, check_callers,
7198 update_caller_keys, update_callee_keys, recursive_inlining,
7199 add_new_edges_to_heap, speculation_useful_p,
7200 inline_small_functions,
7201 inline_small_functions, flatten_function,
7202 inline_to_all_callers_1): Update.
7204 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7206 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7209 2018-01-30 Richard Biener <rguenther@suse.de>
7211 PR tree-optimization/83008
7212 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7213 invariant and constant vector uses in stmts when they need
7216 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7219 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7220 * configure: Regenerate.
7222 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7224 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7226 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7227 Use gen_rtx_REG rather than gen_lowpart.
7229 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7231 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7232 rather than 0 when creating partial subregs.
7234 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7236 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7239 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7242 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7243 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7244 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7245 flags. This restores the settings used before the 2017-07-24.
7246 Turning off pre increment/decrement/modify allows IVOPTS to
7247 optimize DF/SF loops where the index is an int.
7249 2018-01-29 Richard Biener <rguenther@suse.de>
7250 Kelvin Nilsen <kelvin@gcc.gnu.org>
7253 * tree-vect-stmts.c (vectorizable_call): Don't call
7254 targetm.vectorize_builtin_md_vectorized_function if callee is
7257 2018-01-22 Carl Love <cel@us.ibm.com>
7259 * doc/extend.tex: Fix typo in second arg in
7260 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7262 2018-01-29 Richard Biener <rguenther@suse.de>
7264 PR tree-optimization/84086
7265 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7266 (flush_ssaname_freelist): When SSA names were released reset
7267 the SCEV hash table.
7269 2018-01-29 Richard Biener <rguenther@suse.de>
7271 PR tree-optimization/84057
7272 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7273 removed paths when removing edges.
7275 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7277 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7278 -mfunction-return=@var{choice}.
7280 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7283 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7285 (layout::print_source_line): Likewise.
7286 (test_get_line_width_without_trailing_whitespace): Add test cases.
7288 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7291 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7294 2018-01-26 Jim Wilson <jimw@sifive.com>
7296 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7298 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7301 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7303 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7304 and CMP + SUB-immediate -> SUBS.
7306 2018-01-26 Martin Sebor <msebor@redhat.com>
7308 PR tree-optimization/83896
7309 * tree-ssa-strlen.c (get_string_len): Rename...
7310 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7311 Avoid assuming length is constant.
7312 (handle_char_store): Use HOST_WIDE_INT for string length.
7314 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7317 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7318 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7320 2018-01-26 Richard Biener <rguenther@suse.de>
7322 PR rtl-optimization/84003
7323 * dse.c (record_store): Only record redundant stores when
7324 the earlier store aliases at least all accesses the later one does.
7326 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7328 PR rtl-optimization/83985
7329 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7330 REG_CFA_RESTORE insns.
7331 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7332 REG_CFA_RESTORE insns here.
7335 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7336 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7338 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7340 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7341 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7342 (arc_init): Likewise.
7343 (arc_override_options): Likewise.
7344 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7346 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7348 * config/arc/arc.h (TARGET_DBNZ): Define.
7349 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7350 properly set the tune attribute.
7351 (dbnz): Use TARGET_DBNZ guard.
7352 * config/arc/arc.opt (mtune): Add core3 option.
7354 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7356 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7357 recognize new pic like addresses.
7358 (arc_delegitimize_address): Clean up.
7360 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7362 * config/arc/arc-arches.def: Option mrf16 valid for all
7364 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7365 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7366 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7367 * config/arc/arc-tables.opt: Regenerate.
7368 * config/arc/arc.c (arc_conditional_register_usage): Handle
7369 reduced register file case.
7370 (arc_file_start): Set must have build attributes.
7371 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7373 * config/arc/arc.opt (mrf16): Add new option.
7374 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7375 * config/arc/genmultilib.awk: Handle new mrf16 option.
7376 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7377 * config/arc/t-multilib: Regenerate.
7378 * doc/invoke.texi (ARC Options): Document mrf16 option.
7380 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7382 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7383 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7384 (arc_attribute_table): Add 'secure_call' attribute.
7385 (arc_print_operand): Print secure call operand.
7386 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7388 (arc_is_secure_call_p): New function. * config/arc/arc.md
7389 (call_i): Add support for sjli instruction.
7390 (call_value_i): Likewise.
7391 * config/arc/constraints.md (Csc): New constraint.
7393 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7394 John Eric Martin <John.Martin@emmicro-us.com>
7396 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7397 * config/arc/arc.c (_arc_jli_section): New struct.
7398 (arc_jli_section): New type.
7399 (rc_jli_sections): New static variable.
7400 (arc_handle_jli_attribute): New function.
7401 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7402 (arc_file_end): New function.
7403 (TARGET_ASM_FILE_END): Define.
7404 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7405 (arc_add_jli_section): New function.
7406 (jli_call_scan): Likewise.
7407 (arc_reorg): Call jli_call_scan.
7408 (arc_output_addsi): Remove 'S' from printing asm operand.
7409 (arc_is_jli_call_p): New function.
7410 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7412 (movhi_insn): Likewise.
7413 (movsi_insn): Likewise.
7414 (movsi_set_cc_insn): Likewise.
7415 (loadqi_update): Likewise.
7416 (load_zeroextendqisi_update): Likewise.
7417 (load_signextendqisi_update): Likewise.
7418 (loadhi_update): Likewise.
7419 (load_zeroextendhisi_update): Likewise.
7420 (load_signextendhisi_update): Likewise.
7421 (loadsi_update): Likewise.
7422 (loadsf_update): Likewise.
7423 (movsicc_insn): Likewise.
7424 (bset_insn): Likewise.
7425 (bxor_insn): Likewise.
7426 (bclr_insn): Likewise.
7427 (bmsk_insn): Likewise.
7428 (bicsi3_insn): Likewise.
7429 (cmpsi_cc_c_insn): Likewise.
7430 (movsi_ne): Likewise.
7431 (movsi_cond_exec): Likewise.
7432 (clrsbsi2): Likewise.
7443 (call_i): Remove 'S' asm letter, add jli instruction.
7444 (call_value_i): Likewise.
7445 * config/arc/arc.op (mjli-always): New option.
7446 * config/arc/constraints.md (Cji): New constraint.
7447 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7449 (subsf3_fpx): Likewise.
7450 (mulsf3_fpx): Likewise.
7451 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7453 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7455 * doc/invoke.texi (ARC): Document mjli-always option.
7457 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7459 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7460 avoid addition with 0 and use incw and decw where possible.
7462 2018-01-26 Richard Biener <rguenther@suse.de>
7464 PR tree-optimization/81082
7465 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7466 association if it requires casting to unsigned.
7467 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7468 from fold_plusminus_mult_expr to catch important cases late when
7469 range info is available.
7471 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7473 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7474 * configure.ac (hidden_linkonce): New test.
7475 * configure: Regenerate.
7476 * config.in: Regenerate.
7478 2018-01-26 Julia Koval <julia.koval@intel.com>
7480 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7481 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7482 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7483 _mm_mask_bitshuffle_epi64_mask): Fix type.
7484 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7485 USI_FTYPE_V4DI_V4DI_USI): Remove.
7486 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7487 __builtin_ia32_vpshufbitqmb256_mask,
7488 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7489 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7490 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7492 2018-01-26 Alan Modra <amodra@gmail.com>
7495 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7496 UNSPEC_VBPERMQ. Sort other unspecs.
7498 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7500 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7502 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7505 * predict.c (drop_profile): Do not push/pop cfun; update also
7507 (handle_missing_profiles): Fix logic looking for zero profiles.
7509 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7512 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7513 on functions with #pragma omp declare simd or functions with simd
7515 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7516 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7517 Remove trailing \n from warning_at calls.
7519 2018-01-25 Tom de Vries <tom@codesourcery.com>
7522 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7523 for neutered workers.
7525 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7528 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7529 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7531 2018-01-24 Jeff Law <law@redhat.com>
7534 * i386.c (get_probe_interval): Move to earlier point.
7535 (ix86_compute_frame_layout): If -fstack-clash-protection and
7536 the frame is larger than the probe interval, then use pushes
7537 to save registers rather than reg->mem moves.
7538 (ix86_expand_prologue): Remove conditional for int_registers_saved
7541 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7544 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7545 min/max for never referenced object.
7547 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7550 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7552 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7553 attributes from DECL_ATTRIBUTES (decl) without affecting
7554 DECL_ATTRIBUTES (current_function_decl).
7555 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7556 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7558 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7560 PR tree-optimization/83979
7561 * fold-const.c (fold_comparison): Use constant_boolean_node
7562 instead of boolean_{true,false}_node.
7564 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7566 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7569 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7571 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7572 Simplify the clause that sets the length attribute.
7573 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7574 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7575 clause that sets the length attribute.
7576 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7578 2018-01-24 Tom de Vries <tom@codesourcery.com>
7581 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7582 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7583 Add strict parameter.
7584 (prevent_branch_around_nothing): Insert dummy insn between branch to
7585 label and label with no ptx insn inbetween.
7586 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7588 2018-01-24 Tom de Vries <tom@codesourcery.com>
7591 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7592 for neutered threads in warp.
7593 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7595 2018-01-24 Richard Biener <rguenther@suse.de>
7597 PR tree-optimization/83176
7598 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7601 2018-01-24 Richard Biener <rguenther@suse.de>
7603 PR tree-optimization/82819
7604 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7605 code generating pluses that are no-ops in the target precision.
7607 2018-01-24 Richard Biener <rguenther@suse.de>
7610 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7612 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7614 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7615 to merge probabilities.
7616 * predict.c (probably_never_executed): Also mark as cold functions
7617 with global 0 profile and guessed local profile.
7618 * profile-count.c (profile_probability::combine_with_count): New
7620 * profile-count.h (profile_probability::operator*,
7621 profile_probability::operator*=, profile_probability::operator/,
7622 profile_probability::operator/=): Reduce precision to adjusted
7623 and set value to guessed on contradictory divisions.
7624 (profile_probability::combine_with_freq): Remove.
7625 (profile_probability::combine_wiht_count): Declare.
7626 (profile_count::force_nonzero):: Set to adjusted.
7627 (profile_count::probability_in):: Set quality to adjusted.
7628 * tree-ssa-tail-merge.c (replace_block_by): Use
7631 2018-01-23 Andrew Waterman <andrew@sifive.com>
7632 Jim Wilson <jimw@sifive.com>
7634 * config/riscv/riscv.c (riscv_stack_boundary): New.
7635 (riscv_option_override): Set riscv_stack_boundary. Handle
7636 riscv_preferred_stack_boundary_arg.
7637 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7638 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7639 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7640 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7641 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7642 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7644 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7647 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7648 of struct ix86_frame.
7649 (ix86_expand_epilogue): Likewise. Add a local variable for
7650 the reg_save_offset field in struct ix86_frame.
7652 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7654 PR tree-optimization/82604
7655 * tree-loop-distribution.c (enum partition_kind): New enum item
7656 PKIND_PARTIAL_MEMSET.
7657 (partition_builtin_p): Support above new enum item.
7658 (generate_code_for_partition): Ditto.
7659 (compute_access_range): Differentiate cases that equality can be
7660 proven at all loops, the innermost loops or no loops.
7661 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7662 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7663 (finalize_partitions, distribute_loop): Don't fuse partition of
7664 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7665 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7668 2018-01-23 Martin Liska <mliska@suse.cz>
7670 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7671 order to ignore the predictor.
7672 (PRED_POLYMORPHIC_CALL): Likewise.
7673 (PRED_RECURSIVE_CALL): Likewise.
7675 2018-01-23 Martin Liska <mliska@suse.cz>
7677 * tree-profile.c (tree_profiling): Print function header to
7678 aware reader which function we are working on.
7679 * value-prof.c (gimple_find_values_to_profile): Do not print
7680 not interesting value histograms.
7682 2018-01-23 Martin Liska <mliska@suse.cz>
7684 * profile-count.h (enum profile_quality): Add
7685 profile_uninitialized as the first value. Do not number values
7686 as they are zero based.
7687 (profile_count::verify): Update sanity check.
7688 (profile_probability::verify): Likewise.
7690 2018-01-23 Nathan Sidwell <nathan@acm.org>
7692 * doc/invoke.texi (ffor-scope): Deprecate.
7694 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7696 PR tree-optimization/83510
7697 * domwalk.c (set_all_edges_as_executable): New function.
7698 (dom_walker::dom_walker): Convert bool param
7699 "skip_unreachable_blocks" to enum reachability. Move setup of
7700 edge flags to set_all_edges_as_executable and only do it when
7701 reachability is REACHABLE_BLOCKS.
7702 * domwalk.h (enum dom_walker::reachability): New enum.
7703 (dom_walker::dom_walker): Convert bool param
7704 "skip_unreachable_blocks" to enum reachability.
7705 (set_all_edges_as_executable): New decl.
7706 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7707 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7709 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7710 but converting true to REACHABLE_BLOCKS.
7711 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7713 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7714 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7715 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7717 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7718 if check_all_array_refs will be called.
7720 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7722 * tree.c (selftest::test_location_wrappers): Add more test
7725 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7727 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7728 (selftest::test_bit_in_range): Likewise.
7730 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7733 * doc/sourcebuild.texi (vect_float): Say that the selector
7734 only describes the situation when -funsafe-math-optimizations is on.
7735 (vect_float_strict): Document.
7737 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7739 PR tree-optimization/83965
7740 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7741 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7742 instead of checking only for a reduction.
7743 (vect_recog_widen_sum_pattern): Likewise.
7745 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7747 * predict.c (probably_never_executed): Only use precise profile info.
7748 (compute_function_frequency): Skip after inlining hack since we now
7749 have quality checking.
7751 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7753 * profile-count.h (profile_probability::very_unlikely,
7754 profile_probability::unlikely, profile_probability::even): Set
7755 precision to guessed.
7757 2018-01-23 Richard Biener <rguenther@suse.de>
7759 PR tree-optimization/83963
7760 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7761 Properly terminate dominator walk when crossing the exit edge not
7762 when visiting its source block.
7764 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7767 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7768 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7770 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7772 PR tree-optimization/83957
7773 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7774 semicolon after for body surrounded by braces.
7776 PR tree-optimization/83081
7777 * profile-count.h (profile_probability::split): New method.
7778 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7779 Use profile_probability::split.
7780 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7781 when splitting a single conditional jump into 2.
7783 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7785 PR tree-optimization/69452
7786 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7789 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7791 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7792 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7793 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7795 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7797 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7798 * config/rl78/rl78.md (movdi): New define_expand.
7799 * config/rl78/rl78.c (rl78_split_movdi): New function.
7801 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7804 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7806 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7807 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7808 128-bit to produce an UNSPEC move to get the double word with the
7809 signbit and then a shift directly to do signbit.
7810 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7811 implementation with a new version that just does either a direct
7812 move or a regular move. Move memory interface to separate insns.
7813 Move insns so they are next to the expander.
7814 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7815 with signbit move. Split big and little endian case.
7816 (signbit<mode>2_dm_mem_le): Likewise.
7817 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7818 (signbit<mode>2_dm2): Likewise.
7820 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7822 * config/rl78/rl78.md (anddi3): New define_expand.
7824 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7826 * config/rl78/rl78.md (umindi3): New define_expand.
7828 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7830 * config/rl78/rl78.md (smindi3): New define_expand.
7832 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7834 * config/rl78/rl78.md (smaxdi3): New define_expand.
7836 2018-01-22 Carl Love <cel@us.ibm.com>
7838 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7839 LVX_V1TI): Add macro expansion.
7840 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7841 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7842 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7843 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7844 Change check to determine if the instruction is a byte reversing
7845 entry. Fix typo in comment.
7846 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7847 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7848 Add def_builtin calls for new builtins.
7849 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7850 Add define_insn expansion.
7852 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7854 * config/rl78/rl78.md (umaxdi3): New define_expand.
7856 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7858 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7859 for non-QImode registers.
7861 2018-01-22 Richard Biener <rguenther@suse.de>
7863 PR tree-optimization/83963
7864 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7865 including the loop exit block.
7866 (scop_detection::merge_sese): Likewise.
7867 (scop_detection::add_scop): Do it here instead.
7869 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7871 * doc/sourcebuild.texi (arm_softfloat): Document.
7873 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7876 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7877 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7878 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7880 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7881 David Edelsohn <dje.gcc@gmail.com>
7884 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7885 Change "crset eq" to "crset 2".
7886 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7887 (*call_indirect_aix<mode>_nospec): Likewise.
7888 (*call_value_indirect_aix<mode>_nospec): Likewise.
7889 (*call_indirect_elfv2<mode>_nospec): Likewise.
7890 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7891 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7892 change assembly output from . to $.
7893 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7894 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7895 (*tablejump<mode>_internal1_nospec): Likewise.
7897 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7900 * config/sh/sh_optimize_sett_clrt.cc:
7901 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7903 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7905 PR tree-optimization/83940
7906 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7907 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7908 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7909 use it to pass back the definition type.
7910 (vect_check_store_rhs): Likewise rhs_dt_out.
7911 (vect_build_gather_load_calls): Add a mask_dt argument and use
7912 it instead of a call to vect_is_simple_use.
7913 (vectorizable_store): Update calls to vect_check_load_store_mask
7914 and vect_check_store_rhs. Use the dt returned by the latter instead
7915 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7916 instead of calls to vect_is_simple_use. Pass the scalar rather
7917 than the vector operand to vect_is_simple_use when handling
7918 second and subsequent copies of an rhs value.
7919 (vectorizable_load): Update calls to vect_check_load_store_mask
7920 and vect_build_gather_load_calls. Use the cached mask_dt and
7921 gs_info.offset_dt instead of calls to vect_is_simple_use.
7923 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7926 * tree-emutls.c: Include gimplify.h.
7927 (lower_emutls_2): New function.
7928 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7929 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7930 it before further processing.
7933 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7934 UINTVAL (trueop1) instead of INTVAL (op1).
7936 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7940 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7941 INCOMING_FRAME_SP_OFFSET if not defined.
7942 (scan_trace): Add ENTRY argument. If true and
7943 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7944 emit a note to adjust the CFA offset.
7945 (create_cfi_notes): Adjust scan_trace callers.
7946 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7947 INCOMING_FRAME_SP_OFFSET in the CIE.
7948 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7949 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7951 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7952 * doc/tm.texi: Regenerated.
7954 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7956 PR rtl-optimization/83147
7957 * lra-constraints.c (remove_inheritance_pseudos): Use
7958 lra_substitute_pseudo_within_insn.
7960 2018-01-19 Tom de Vries <tom@codesourcery.com>
7961 Cesar Philippidis <cesar@codesourcery.com>
7964 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7966 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7969 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7970 spaces for function labels.
7972 2018-01-19 Martin Liska <mliska@suse.cz>
7974 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7975 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7976 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7977 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7978 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7979 (PRED_CONST_RETURN): Change from 69 to 65.
7980 (PRED_NULL_RETURN): Change from 91 to 71.
7981 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7982 (PRED_LOOP_GUARD): Change from 66 to 73.
7984 2018-01-19 Martin Liska <mliska@suse.cz>
7986 * predict.c (predict_insn_def): Add new assert.
7987 (struct branch_predictor): Change type to signed integer.
7988 (test_prediction_value_range): Amend test to cover
7990 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7991 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7992 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7993 (PRED_LOOP_IV_COMPARE): Likewise.
7994 * predict.h (PROB_UNINITIALIZED): Define new constant.
7996 2018-01-19 Martin Liska <mliska@suse.cz>
7998 * predict.c (dump_prediction): Add new format for
7999 analyze_brprob.py script which is enabled with -details
8001 * profile-count.h (precise_p): New function.
8003 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8005 PR tree-optimization/83922
8006 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8007 there are no statements that need masking.
8008 (vect_active_double_reduction_p): New function.
8009 (vect_analyze_loop_operations): Use it when handling phis that
8010 are not in the loop header.
8012 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8014 PR tree-optimization/83914
8015 * tree-vect-loop.c (vectorizable_induction): Don't convert
8016 init_expr or apply the peeling adjustment for inductions
8017 that are nested within the vectorized loop.
8019 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8021 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8024 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8028 * function.h (gimplify_parameters): Add gimple_seq * argument.
8029 * function.c: Include gimple.h and options.h.
8030 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8031 for the added local temporaries if needed.
8032 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8033 if there are any parameter cleanups, wrap whole body into a
8034 try/finally with the cleanups.
8036 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8039 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8040 Use GET_MODE_CLASS for scalar floating point.
8042 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8046 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8047 Fix call of call_cgraph_insertion_hooks.
8049 2018-01-18 Martin Sebor <msebor@redhat.com>
8051 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8053 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8056 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8059 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8062 * common.opt: (-ffile-prefix-map): New option.
8063 * opts.c (common_handle_option): Defer it.
8064 * opts-global.c (handle_common_deferred_options): Handle it.
8065 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8066 * file-prefix-map.h: New file.
8067 (remap_debug_filename, add_debug_prefix_map): ...here.
8068 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8069 * final.c (debug_prefix_map, add_debug_prefix_map
8070 remap_debug_filename): Move to...
8071 * file-prefix-map.c: New file.
8072 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8073 generalize, get rid of alloca(), use strrchr() instead of strchr().
8074 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8075 Implement in terms of add_prefix_map().
8076 (remap_macro_filename, remap_debug_filename): Implement in term of
8078 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8079 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8080 * dbxout.c: Include file-prefix-map.h.
8081 * varasm.c: Likewise.
8082 * vmsdbgout.c: Likewise.
8083 * xcoffout.c: Likewise.
8084 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8085 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8086 * doc/invoke.texi (-ffile-prefix-map): Document.
8087 (-fdebug-prefix-map): Update description.
8089 2018-01-18 Martin Liska <mliska@suse.cz>
8091 * config/i386/i386.c (indirect_thunk_name): Document that also
8093 (output_indirect_thunk): Document why both instructions
8094 (pause and lfence) are generated.
8096 2018-01-18 Richard Biener <rguenther@suse.de>
8098 PR tree-optimization/83887
8099 * graphite-scop-detection.c
8100 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8101 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8102 (scop_detection::merge_sese): Re-implement with a flood-fill
8103 algorithm that properly finds a SESE region if it exists.
8105 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8108 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8109 pointer_diff optimizations use view_convert instead of convert.
8111 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8113 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8114 Generate different code for -mno-speculate-indirect-jumps.
8115 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8116 (*call_indirect_aix<mode>): Disable for
8117 -mno-speculate-indirect-jumps.
8118 (*call_indirect_aix<mode>_nospec): New define_insn.
8119 (*call_value_indirect_aix<mode>): Disable for
8120 -mno-speculate-indirect-jumps.
8121 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8122 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8123 -mno-speculate-indirect-jumps.
8124 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8126 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8128 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8129 long double type, set the flags for noting the default long double
8130 type, even if we don't pass or return a long double type.
8132 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8135 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8138 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8140 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8141 support for merge[hl].
8142 (fold_mergehl_helper): New helper function.
8143 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8144 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8145 (altivec_vmrglw_direct): Add xxmrglw insn.
8147 2018-01-17 Andrew Waterman <andrew@sifive.com>
8149 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8150 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8152 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8155 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8156 call the lto_location_cache before reading the
8157 DECL_SOURCE_LOCATION of the types.
8159 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8160 Richard Sandiford <richard.sandiford@linaro.org>
8162 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8163 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8164 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8165 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8166 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8168 * config/aarch64/constraints.md (aarch64_movti_operand):
8170 * config/aarch64/predicates.md (Uti): Add new constraint.
8172 2018-01-17 Carl Love <cel@us.ibm.com>
8174 * config/rs6000/vsx.md (define_expand xl_len_r,
8175 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8176 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8178 (define_expand, define_insn): Move the shift left from the
8179 define_insn to the define_expand for lxvl and stxvl instructions.
8180 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8181 and XL_LEN_R definitions to PURE.
8183 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8185 * config/i386/i386.c (indirect_thunk_name): Declare regno
8186 as unsigned int. Compare regno with INVALID_REGNUM.
8187 (output_indirect_thunk): Ditto.
8188 (output_indirect_thunk_function): Ditto.
8189 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8190 in the call to output_indirect_thunk_function.
8192 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8195 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8196 rather than the size of inner_type to determine the stack slot size
8197 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8199 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8202 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8205 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8207 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8208 endian Linux systems to optionally enable multilibs for selecting
8209 the long double type if the user configured an explicit type.
8210 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8211 have no long double multilibs if not defined.
8212 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8213 warn if the user used -mabi={ieee,ibm}longdouble and we built
8214 multilibs for long double.
8215 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8216 appropriate multilib option.
8217 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8219 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8220 for building long double multilibs.
8221 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8223 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8225 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8228 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8230 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8233 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8236 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8239 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8241 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8242 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8244 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8246 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8247 different rtl trees depending on TARGET_64BIT.
8248 (rs6000_gen_lvx): Likewise.
8250 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8252 * config/visium/visium.md (nop): Tweak comment.
8253 (hazard_nop): Likewise.
8255 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8257 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8258 -mspeculate-indirect-jumps.
8259 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8260 for -mno-speculate-indirect-jumps.
8261 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8262 (*call_value_indirect_elfv2<mode>): Disable for
8263 -mno-speculate-indirect-jumps.
8264 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8265 (indirect_jump): Emit different RTL for
8266 -mno-speculate-indirect-jumps.
8267 (*indirect_jump<mode>): Disable for
8268 -mno-speculate-indirect-jumps.
8269 (*indirect_jump<mode>_nospec): New define_insn.
8270 (tablejump): Emit different RTL for
8271 -mno-speculate-indirect-jumps.
8272 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8273 (tablejumpsi_nospec): New define_expand.
8274 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8275 (tablejumpdi_nospec): New define_expand.
8276 (*tablejump<mode>_internal1): Disable for
8277 -mno-speculate-indirect-jumps.
8278 (*tablejump<mode>_internal1_nospec): New define_insn.
8279 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8282 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8284 * caller-save.c (insert_save): Drop unnecessary parameter. All
8287 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8288 Richard Biener <rguenth@suse.de>
8291 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8292 return early, inline manually is_gimple_sizepos. Make sure if we
8293 call gimplify_expr we don't end up with a gimple constant.
8294 * tree.c (variably_modified_type_p): Don't return true for
8295 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8296 * gimplify.h (is_gimple_sizepos): Remove.
8298 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8300 PR tree-optimization/83857
8301 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8302 vectorizable_live_operation for pure SLP statements.
8303 (vectorizable_live_operation): Handle PHIs.
8305 2018-01-16 Richard Biener <rguenther@suse.de>
8307 PR tree-optimization/83867
8308 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8309 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8311 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8314 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8315 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8316 If off is not INTEGER_CST, issue a may not be aligned warning
8317 rather than isn't aligned. Use isn%'t rather than isn't.
8318 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8320 <case MULT_EXPR>: Improve the case when bottom and one of the
8321 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8322 operand, in that case check if the other operand is multiple of
8323 bottom divided by the INTEGER_CST operand.
8325 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8328 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8329 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8330 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8331 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8332 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8333 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8334 (pa_function_arg_size): New function.
8336 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8338 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8339 in a separate statement.
8341 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8343 PR tree-optimization/83847
8344 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8345 group gathers and scatters.
8347 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8349 PR rtl-optimization/86620
8350 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8352 PR rtl-optimization/83213
8353 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8354 to last if both are JUMP_INSNs.
8356 PR tree-optimization/83843
8357 * gimple-ssa-store-merging.c
8358 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8359 store_immediate_info for bswap/nop orig_stores.
8361 2018-01-15 Andrew Waterman <andrew@sifive.com>
8363 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8365 <UDIV>: Increase cost if !TARGET_DIV.
8367 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8369 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8370 (define_attr "cr_logical_3op"): New.
8371 (cceq_ior_compare): Adjust.
8372 (cceq_ior_compare_complement): Adjust.
8373 (*cceq_rev_compare): Adjust.
8374 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8375 (is_cracked_insn): Adjust.
8376 (insn_must_be_first_in_group): Adjust.
8377 * config/rs6000/40x.md: Adjust.
8378 * config/rs6000/440.md: Adjust.
8379 * config/rs6000/476.md: Adjust.
8380 * config/rs6000/601.md: Adjust.
8381 * config/rs6000/603.md: Adjust.
8382 * config/rs6000/6xx.md: Adjust.
8383 * config/rs6000/7450.md: Adjust.
8384 * config/rs6000/7xx.md: Adjust.
8385 * config/rs6000/8540.md: Adjust.
8386 * config/rs6000/cell.md: Adjust.
8387 * config/rs6000/e300c2c3.md: Adjust.
8388 * config/rs6000/e500mc.md: Adjust.
8389 * config/rs6000/e500mc64.md: Adjust.
8390 * config/rs6000/e5500.md: Adjust.
8391 * config/rs6000/e6500.md: Adjust.
8392 * config/rs6000/mpc.md: Adjust.
8393 * config/rs6000/power4.md: Adjust.
8394 * config/rs6000/power5.md: Adjust.
8395 * config/rs6000/power6.md: Adjust.
8396 * config/rs6000/power7.md: Adjust.
8397 * config/rs6000/power8.md: Adjust.
8398 * config/rs6000/power9.md: Adjust.
8399 * config/rs6000/rs64.md: Adjust.
8400 * config/rs6000/titan.md: Adjust.
8402 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8404 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8405 ix86_indirect_branch_register logic.
8407 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8409 * config/i386/constraints.md (Bs): Update
8410 ix86_indirect_branch_register check. Don't check
8411 ix86_indirect_branch_register with GOT_memory_operand.
8413 * config/i386/predicates.md (GOT_memory_operand): Don't check
8414 ix86_indirect_branch_register here.
8415 (GOT32_symbol_operand): Likewise.
8417 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8419 * config/i386/predicates.md (constant_call_address_operand):
8420 Rewrite ix86_indirect_branch_register logic.
8421 (sibcall_insn_operand): Likewise.
8423 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8425 * config/i386/constraints.md (Bs): Replace
8426 ix86_indirect_branch_thunk_register with
8427 ix86_indirect_branch_register.
8429 * config/i386/i386.md (indirect_jump): Likewise.
8430 (tablejump): Likewise.
8431 (*sibcall_memory): Likewise.
8432 (*sibcall_value_memory): Likewise.
8433 Peepholes of indirect call and jump via memory: Likewise.
8434 * config/i386/i386.opt: Likewise.
8435 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8436 (GOT_memory_operand): Likewise.
8437 (call_insn_operand): Likewise.
8438 (sibcall_insn_operand): Likewise.
8439 (GOT32_symbol_operand): Likewise.
8441 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8444 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8445 type rather than type addr's type points to.
8446 (expand_omp_atomic_mutex): Likewise.
8447 (expand_omp_atomic): Likewise.
8449 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8452 * config/i386/i386.c (output_indirect_thunk_function): Use
8453 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8454 for __x86_return_thunk.
8456 2018-01-15 Richard Biener <rguenther@suse.de>
8459 * expmed.c (extract_bit_field_1): Fix typo.
8461 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8464 * config/arm/iterators.md (VF): New mode iterator.
8465 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8466 Remove integer-related logic from pattern.
8467 (neon_vabd<mode>_3): Likewise.
8469 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8472 * common.opt (fstrict-overflow): No longer an alias.
8473 (fwrapv-pointer): New option.
8474 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8475 also for pointer types based on flag_wrapv_pointer.
8476 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8477 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8478 opts->x_flag_wrapv got set.
8479 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8480 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8481 POINTER_TYPE_OVERFLOW_UNDEFINED.
8482 * match.pd: Likewise in address comparison pattern.
8483 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8485 2018-01-15 Richard Biener <rguenther@suse.de>
8488 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8489 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8490 Reset type names to their identifier if their TYPE_DECL doesn't
8491 have linkage (and thus is used for ODR and devirt).
8492 (save_debug_info_for_decl): Remove.
8493 (save_debug_info_for_type): Likewise.
8494 (add_tree_to_fld_list): Adjust.
8495 * tree-pretty-print.c (dump_generic_node): Make dumping of
8496 type names more robust.
8498 2018-01-15 Richard Biener <rguenther@suse.de>
8500 * BASE-VER: Bump to 8.0.1.
8502 2018-01-14 Martin Sebor <msebor@redhat.com>
8505 * builtins.c (check_access): Avoid warning when the no-warning bit
8508 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8510 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8511 * ira-color (allocno_hard_regs_compare): Likewise.
8513 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8516 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8517 Use .pushsection/.popsection.
8519 2018-01-14 Martin Sebor <msebor@redhat.com>
8522 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8524 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8526 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8527 entry from extra_headers.
8528 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8529 extra_headers, make the list bitwise identical to the i?86-*-* one.
8531 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8533 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8534 -mcmodel=large with -mindirect-branch=thunk,
8535 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8536 -mfunction-return=thunk-extern.
8537 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8538 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8539 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8541 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8543 * config/i386/i386.c (print_reg): Print the name of the full
8544 integer register without '%'.
8545 (ix86_print_operand): Handle 'V'.
8546 * doc/extend.texi: Document 'V' modifier.
8548 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8550 * config/i386/constraints.md (Bs): Disallow memory operand for
8551 -mindirect-branch-register.
8553 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8554 (GOT_memory_operand): Likewise.
8555 (call_insn_operand): Likewise.
8556 (sibcall_insn_operand): Likewise.
8557 (GOT32_symbol_operand): Likewise.
8558 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8559 for -mindirect-branch-register.
8560 (tablejump): Likewise.
8561 (*sibcall_memory): Likewise.
8562 (*sibcall_value_memory): Likewise.
8563 Disallow peepholes of indirect call and jump via memory for
8564 -mindirect-branch-register.
8565 (*call_pop): Replace m with Bw.
8566 (*call_value_pop): Likewise.
8567 (*sibcall_pop_memory): Replace m with Bs.
8568 * config/i386/i386.opt (mindirect-branch-register): New option.
8569 * doc/invoke.texi: Document -mindirect-branch-register option.
8571 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8573 * config/i386/i386-protos.h (ix86_output_function_return): New.
8574 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8575 set function_return_type.
8576 (indirect_thunk_name): Add ret_p to indicate thunk for function
8578 (output_indirect_thunk_function): Pass false to
8579 indirect_thunk_name.
8580 (ix86_output_indirect_branch_via_reg): Likewise.
8581 (ix86_output_indirect_branch_via_push): Likewise.
8582 (output_indirect_thunk_function): Create alias for function
8583 return thunk if regno < 0.
8584 (ix86_output_function_return): New function.
8585 (ix86_handle_fndecl_attribute): Handle function_return.
8586 (ix86_attribute_table): Add function_return.
8587 * config/i386/i386.h (machine_function): Add
8588 function_return_type.
8589 * config/i386/i386.md (simple_return_internal): Use
8590 ix86_output_function_return.
8591 (simple_return_internal_long): Likewise.
8592 * config/i386/i386.opt (mfunction-return=): New option.
8593 (indirect_branch): Mention -mfunction-return=.
8594 * doc/extend.texi: Document function_return function attribute.
8595 * doc/invoke.texi: Document -mfunction-return= option.
8597 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8599 * config/i386/i386-opts.h (indirect_branch): New.
8600 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8601 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8602 with local indirect jump when converting indirect call and jump.
8603 (ix86_set_indirect_branch_type): New.
8604 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8605 (indirectlabelno): New.
8606 (indirect_thunk_needed): Likewise.
8607 (indirect_thunk_bnd_needed): Likewise.
8608 (indirect_thunks_used): Likewise.
8609 (indirect_thunks_bnd_used): Likewise.
8610 (INDIRECT_LABEL): Likewise.
8611 (indirect_thunk_name): Likewise.
8612 (output_indirect_thunk): Likewise.
8613 (output_indirect_thunk_function): Likewise.
8614 (ix86_output_indirect_branch_via_reg): Likewise.
8615 (ix86_output_indirect_branch_via_push): Likewise.
8616 (ix86_output_indirect_branch): Likewise.
8617 (ix86_output_indirect_jmp): Likewise.
8618 (ix86_code_end): Call output_indirect_thunk_function if needed.
8619 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8621 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8622 (ix86_attribute_table): Add indirect_branch.
8623 * config/i386/i386.h (machine_function): Add indirect_branch_type
8624 and has_local_indirect_jump.
8625 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8627 (tablejump): Likewise.
8628 (*indirect_jump): Use ix86_output_indirect_jmp.
8629 (*tablejump_1): Likewise.
8630 (simple_return_indirect_internal): Likewise.
8631 * config/i386/i386.opt (mindirect-branch=): New option.
8632 (indirect_branch): New.
8635 (thunk-inline): Likewise.
8636 (thunk-extern): Likewise.
8637 * doc/extend.texi: Document indirect_branch function attribute.
8638 * doc/invoke.texi: Document -mindirect-branch= option.
8640 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8643 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8645 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8647 * ipa-inline.c (want_inline_small_function_p): Return false if
8648 inlining has already failed with CIF_FINAL_ERROR.
8649 (update_caller_keys): Call want_inline_small_function_p before
8651 (update_callee_keys): Likewise.
8653 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8655 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8657 (rs6000_quadword_masked_address_p): Likewise.
8658 (quad_aligned_load_p): Likewise.
8659 (quad_aligned_store_p): Likewise.
8660 (const_load_sequence_p): Add comment to describe the outer-most loop.
8661 (mimic_memory_attributes_and_flags): New function.
8662 (rs6000_gen_stvx): Likewise.
8663 (replace_swapped_aligned_store): Likewise.
8664 (rs6000_gen_lvx): Likewise.
8665 (replace_swapped_aligned_load): Likewise.
8666 (replace_swapped_load_constant): Capitalize argument name in
8667 comment describing this function.
8668 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8669 and stores that access quad-word aligned addresses and replace
8670 with stvx or lvx instructions when appropriate.
8671 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8672 New function prototype.
8673 (rs6000_quadword_masked_address_p): Likewise.
8674 (rs6000_gen_lvx): Likewise.
8675 (rs6000_gen_stvx): Likewise.
8676 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8677 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8678 when memory address is aligned.
8679 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8680 this split to select lvx instruction when memory address is aligned.
8681 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8682 instruction when memory address is aligned.
8683 (*vsx_le_perm_load_v16qi): Likewise.
8684 (four unnamed splitters): Modify to select the stvx instruction
8685 when memory is aligned.
8687 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8689 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8690 which appears in the queue multiple times.
8692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8693 Alan Hayward <alan.hayward@arm.com>
8694 David Sherwood <david.sherwood@arm.com>
8696 * tree-vectorizer.h (vec_lower_bound): New structure.
8697 (_loop_vec_info): Add check_nonzero and lower_bounds.
8698 (LOOP_VINFO_CHECK_NONZERO): New macro.
8699 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8700 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8701 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8702 fields. Make seg_len the distance travelled, not including the
8704 (dr_direction_indicator): Declare.
8705 (dr_zero_step_indicator): Likewise.
8706 (dr_known_forward_stride_p): Likewise.
8707 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8709 (runtime_alias_check_p): Allow runtime alias checks with
8711 (operator ==): Compare access_size and align.
8712 (prune_runtime_alias_test_list): Rework for new distinction between
8713 the access_size and seg_len.
8714 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8716 (get_segment_min_max): New function.
8717 (create_intersect_range_checks): Use it.
8718 (dr_step_indicator): New function.
8719 (dr_direction_indicator): Likewise.
8720 (dr_zero_step_indicator): Likewise.
8721 (dr_known_forward_stride_p): Likewise.
8722 * tree-loop-distribution.c (data_ref_segment_size): Return
8723 DR_STEP * (niters - 1).
8724 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8726 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8727 (vect_preserves_scalar_order_p): New function, split out from...
8728 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8729 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8730 (vect_vfa_access_size): New function.
8731 (vect_vfa_align): Likewise.
8732 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8733 (dump_lower_bound): New function.
8734 (vect_check_lower_bound): Likewise.
8735 (vect_small_gap_p): Likewise.
8736 (vectorizable_with_step_bound_p): Likewise.
8737 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8738 depencies if the vectorization factor is 1. Convert the checks
8739 for nonzero steps into checks on the bounds of DR_STEP. Try using
8740 a bunds check for variable steps if the minimum required step is
8741 relatively small. Update calls to the dr_with_seg_len
8742 constructor and to vect_compile_time_alias.
8743 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8745 (vect_loop_versioning): Call it.
8746 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8748 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8751 Alan Hayward <alan.hayward@arm.com>
8752 David Sherwood <david.sherwood@arm.com>
8754 * doc/sourcebuild.texi (vect_scatter_store): Document.
8755 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8757 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8759 * genopinit.c (main): Add supports_vec_scatter_store and
8760 supports_vec_scatter_store_cached to target_optabs.
8761 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8762 IFN_MASK_SCATTER_STORE.
8763 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8765 * internal-fn.h (internal_store_fn_p): Declare.
8766 (internal_fn_stored_value_index): Likewise.
8767 * internal-fn.c (scatter_store_direct): New macro.
8768 (expand_scatter_store_optab_fn): New function.
8769 (direct_scatter_store_optab_supported_p): New macro.
8770 (internal_store_fn_p): New function.
8771 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8772 IFN_MASK_SCATTER_STORE.
8773 (internal_fn_mask_index): Likewise.
8774 (internal_fn_stored_value_index): New function.
8775 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8777 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8778 * optabs-query.c (supports_vec_scatter_store_p): New function.
8779 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8780 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8781 true for scatter stores.
8782 (vect_gather_scatter_fn_p): Handle scatter stores too.
8783 (vect_check_gather_scatter): Consider using scatter stores if
8784 supports_vec_scatter_store_p.
8785 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8787 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8788 internal_fn_stored_value_index.
8789 (check_load_store_masking): Handle scatter stores too.
8790 (vect_get_store_rhs): Make public.
8791 (vectorizable_call): Use internal_store_fn_p.
8792 (vectorizable_store): Handle scatter store internal functions.
8793 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8794 when deciding whether the end of the group has been reached.
8795 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8796 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8797 (mask_scatter_store<mode>): New insns.
8799 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8800 Alan Hayward <alan.hayward@arm.com>
8801 David Sherwood <david.sherwood@arm.com>
8803 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8804 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8805 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8807 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8808 Use vect_truncate_gather_scatter_offset if we can't treat the
8809 operation as a normal gather load or scatter store.
8810 (get_group_load_store_type): Take the gather_scatter_info
8811 as argument. Try using a gather load or scatter store for
8812 single-element groups.
8813 (get_load_store_type): Update calls to get_group_load_store_type
8814 and vect_use_strided_gather_scatters_p.
8816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8817 Alan Hayward <alan.hayward@arm.com>
8818 David Sherwood <david.sherwood@arm.com>
8820 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8821 optional tree argument.
8822 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8824 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8825 but continue to use the current value as a fallback.
8826 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8827 to compare the updates.
8828 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8829 (get_load_store_type): Use it when handling a strided access.
8830 (vect_get_strided_load_store_ops): New function.
8831 (vect_get_data_ptr_increment): Likewise.
8832 (vectorizable_load): Handle strided gather loads. Always pass
8833 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8836 Alan Hayward <alan.hayward@arm.com>
8837 David Sherwood <david.sherwood@arm.com>
8839 * doc/md.texi (gather_load@var{m}): Document.
8840 (mask_gather_load@var{m}): Likewise.
8841 * genopinit.c (main): Add supports_vec_gather_load and
8842 supports_vec_gather_load_cached to target_optabs.
8843 * optabs-tree.c (init_tree_optimization_optabs): Use
8844 ggc_cleared_alloc to allocate target_optabs.
8845 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8846 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8848 * internal-fn.h (internal_load_fn_p): Declare.
8849 (internal_gather_scatter_fn_p): Likewise.
8850 (internal_fn_mask_index): Likewise.
8851 (internal_gather_scatter_fn_supported_p): Likewise.
8852 * internal-fn.c (gather_load_direct): New macro.
8853 (expand_gather_load_optab_fn): New function.
8854 (direct_gather_load_optab_supported_p): New macro.
8855 (direct_internal_fn_optab): New function.
8856 (internal_load_fn_p): Likewise.
8857 (internal_gather_scatter_fn_p): Likewise.
8858 (internal_fn_mask_index): Likewise.
8859 (internal_gather_scatter_fn_supported_p): Likewise.
8860 * optabs-query.c (supports_at_least_one_mode_p): New function.
8861 (supports_vec_gather_load_p): Likewise.
8862 * optabs-query.h (supports_vec_gather_load_p): Declare.
8863 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8864 and memory_type field.
8865 (NUM_PATTERNS): Bump to 15.
8866 * tree-vect-data-refs.c: Include internal-fn.h.
8867 (vect_gather_scatter_fn_p): New function.
8868 (vect_describe_gather_scatter_call): Likewise.
8869 (vect_check_gather_scatter): Try using internal functions for
8870 gather loads. Recognize existing calls to a gather load function.
8871 (vect_analyze_data_refs): Consider using gather loads if
8872 supports_vec_gather_load_p.
8873 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8874 (vect_get_gather_scatter_offset_type): Likewise.
8875 (vect_convert_mask_for_vectype): Likewise.
8876 (vect_add_conversion_to_patterm): Likewise.
8877 (vect_try_gather_scatter_pattern): Likewise.
8878 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8879 (vect_vect_recog_func_ptrs): Add it.
8880 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8881 internal_fn_mask_index and internal_gather_scatter_fn_p.
8882 (check_load_store_masking): Take the gather_scatter_info as an
8883 argument and handle gather loads.
8884 (vect_get_gather_scatter_ops): New function.
8885 (vectorizable_call): Check internal_load_fn_p.
8886 (vectorizable_load): Likewise. Handle gather load internal
8888 (vectorizable_store): Update call to check_load_store_masking.
8889 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8890 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8891 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8892 (aarch64_gather_scale_operand_d): New predicates.
8893 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8894 (mask_gather_load<mode>): New insns.
8896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8897 Alan Hayward <alan.hayward@arm.com>
8898 David Sherwood <david.sherwood@arm.com>
8900 * optabs.def (fold_left_plus_optab): New optab.
8901 * doc/md.texi (fold_left_plus_@var{m}): Document.
8902 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8903 * internal-fn.c (fold_left_direct): Define.
8904 (expand_fold_left_optab_fn): Likewise.
8905 (direct_fold_left_optab_supported_p): Likewise.
8906 * fold-const-call.c (fold_const_fold_left): New function.
8907 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8908 * tree-parloops.c (valid_reduction_p): New function.
8909 (gather_scalar_reductions): Use it.
8910 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8911 (vect_finish_replace_stmt): Declare.
8912 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8913 (needs_fold_left_reduction_p): New function, split out from...
8914 (vect_is_simple_reduction): ...here. Accept reductions that
8915 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8916 (vect_force_simple_reduction): Also store the reduction type in
8917 the assignment's STMT_VINFO_REDUC_TYPE.
8918 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8919 (merge_with_identity): New function.
8920 (vect_expand_fold_left): Likewise.
8921 (vectorize_fold_left_reduction): Likewise.
8922 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8923 scalar phi in place for it. Check for target support and reject
8924 cases that would reassociate the operation. Defer the transform
8925 phase to vectorize_fold_left_reduction.
8926 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8927 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8928 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8932 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8933 call to ifc_temp_var.
8935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8936 Alan Hayward <alan.hayward@arm.com>
8937 David Sherwood <david.sherwood@arm.com>
8939 * target.def (legitimize_address_displacement): Take the original
8940 offset as a poly_int.
8941 * targhooks.h (default_legitimize_address_displacement): Update
8943 * targhooks.c (default_legitimize_address_displacement): Likewise.
8944 * doc/tm.texi: Regenerate.
8945 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8946 as an argument, moving assert of ad->disp == ad->disp_term to...
8947 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8948 Try calling targetm.legitimize_address_displacement before expanding
8949 the address rather than afterwards, and adjust for the new interface.
8950 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8951 Match the new hook interface. Handle SVE addresses.
8952 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8955 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8957 * Makefile.in (OBJS): Add early-remat.o.
8958 * target.def (select_early_remat_modes): New hook.
8959 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8960 * doc/tm.texi: Regenerate.
8961 * targhooks.h (default_select_early_remat_modes): Declare.
8962 * targhooks.c (default_select_early_remat_modes): New function.
8963 * timevar.def (TV_EARLY_REMAT): New timevar.
8964 * passes.def (pass_early_remat): New pass.
8965 * tree-pass.h (make_pass_early_remat): Declare.
8966 * early-remat.c: New file.
8967 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8969 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8972 Alan Hayward <alan.hayward@arm.com>
8973 David Sherwood <david.sherwood@arm.com>
8975 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8976 vfm1 with a bound_epilog parameter.
8977 (vect_do_peeling): Update calls accordingly, and move the prologue
8978 call earlier in the function. Treat the base bound_epilog as 0 for
8979 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8980 this base when peeling for gaps.
8981 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8982 with fully-masked loops.
8983 (vect_estimate_min_profitable_iters): Handle the single peeled
8984 iteration in that case.
8986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8987 Alan Hayward <alan.hayward@arm.com>
8988 David Sherwood <david.sherwood@arm.com>
8990 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8991 single-element interleaving even if the size is not a power of 2.
8992 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8993 accesses for single-element interleaving if the group size is
8996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8997 Alan Hayward <alan.hayward@arm.com>
8998 David Sherwood <david.sherwood@arm.com>
9000 * doc/md.texi (fold_extract_last_@var{m}): Document.
9001 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9002 * optabs.def (fold_extract_last_optab): New optab.
9003 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9004 * internal-fn.c (fold_extract_direct): New macro.
9005 (expand_fold_extract_optab_fn): Likewise.
9006 (direct_fold_extract_optab_supported_p): Likewise.
9007 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9008 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9009 EXTRACT_LAST_REDUCTION.
9010 (get_initial_def_for_reduction): Do not create an initial vector
9011 for EXTRACT_LAST_REDUCTION reductions.
9012 (vectorizable_reduction): Leave the scalar phi in place for
9013 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9014 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9015 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9016 transform phase to vectorizable_condition.
9017 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9019 (vect_finish_stmt_generation): ...here.
9020 (vect_finish_replace_stmt): New function.
9021 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9022 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9024 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9027 Alan Hayward <alan.hayward@arm.com>
9028 David Sherwood <david.sherwood@arm.com>
9030 * doc/md.texi (extract_last_@var{m}): Document.
9031 * optabs.def (extract_last_optab): New optab.
9032 * internal-fn.def (EXTRACT_LAST): New internal function.
9033 * internal-fn.c (cond_unary_direct): New macro.
9034 (expand_cond_unary_optab_fn): Likewise.
9035 (direct_cond_unary_optab_supported_p): Likewise.
9036 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9037 loops using EXTRACT_LAST.
9038 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9039 (extract_last_<mode>): ...this optab.
9040 (vec_extract<mode><Vel>): Update accordingly.
9042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9043 Alan Hayward <alan.hayward@arm.com>
9044 David Sherwood <david.sherwood@arm.com>
9046 * target.def (empty_mask_is_expensive): New hook.
9047 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9048 * doc/tm.texi: Regenerate.
9049 * targhooks.h (default_empty_mask_is_expensive): Declare.
9050 * targhooks.c (default_empty_mask_is_expensive): New function.
9051 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9052 if the target says that empty masks are expensive.
9053 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9055 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9058 Alan Hayward <alan.hayward@arm.com>
9059 David Sherwood <david.sherwood@arm.com>
9061 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9062 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9063 (vect_use_loop_mask_for_alignment_p): New function.
9064 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9065 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9066 niters_skip argument. Make sure that the first niters_skip elements
9067 of the first iteration are inactive.
9068 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9069 Update call to vect_set_loop_masks_directly.
9070 (get_misalign_in_elems): New function, split out from...
9071 (vect_gen_prolog_loop_niters): ...here.
9072 (vect_update_init_of_dr): Take a code argument that specifies whether
9073 the adjustment should be added or subtracted.
9074 (vect_update_init_of_drs): Likewise.
9075 (vect_prepare_for_masked_peels): New function.
9076 (vect_do_peeling): Skip prologue peeling if we're using a mask
9077 instead. Update call to vect_update_inits_of_drs.
9078 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9080 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9081 alignment. Do not include the number of peeled iterations in
9082 the minimum threshold in that case.
9083 (vectorizable_induction): Adjust the start value down by
9084 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9085 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9086 Take the number of skipped iterations into account when calculating
9088 * tree-vect-stmts.c (vect_gen_while_not): New function.
9090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9091 Alan Hayward <alan.hayward@arm.com>
9092 David Sherwood <david.sherwood@arm.com>
9094 * doc/sourcebuild.texi (vect_fully_masked): Document.
9095 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9097 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9099 (vect_analyze_loop_2): ...here. Don't check the vectorization
9100 factor against the number of loop iterations if the loop is
9103 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9104 Alan Hayward <alan.hayward@arm.com>
9105 David Sherwood <david.sherwood@arm.com>
9107 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9108 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9109 (dump_groups): Update accordingly.
9110 (iv_use::mem_type): New member variable.
9111 (address_p): New function.
9112 (record_use): Add a mem_type argument and initialize the new
9114 (record_group_use): Add a mem_type argument. Use address_p.
9115 Remove obsolete null checks of base_object. Update call to record_use.
9116 (find_interesting_uses_op): Update call to record_group_use.
9117 (find_interesting_uses_cond): Likewise.
9118 (find_interesting_uses_address): Likewise.
9119 (get_mem_type_for_internal_fn): New function.
9120 (find_address_like_use): Likewise.
9121 (find_interesting_uses_stmt): Try find_address_like_use before
9122 calling find_interesting_uses_op.
9123 (addr_offset_valid_p): Use the iv mem_type field as the type
9124 of the addressed memory.
9125 (add_autoinc_candidates): Likewise.
9126 (get_address_cost): Likewise.
9127 (split_small_address_groups_p): Use address_p.
9128 (split_address_groups): Likewise.
9129 (add_iv_candidate_for_use): Likewise.
9130 (autoinc_possible_for_pair): Likewise.
9131 (rewrite_groups): Likewise.
9132 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9133 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9134 (get_alias_ptr_type_for_ptr_address): New function.
9135 (rewrite_use_address): Rewrite address uses in calls that were
9136 identified by find_address_like_use.
9138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9139 Alan Hayward <alan.hayward@arm.com>
9140 David Sherwood <david.sherwood@arm.com>
9142 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9144 * gimple-expr.h (is_gimple_addressable: Likewise.
9145 * gimple-expr.c (is_gimple_address): Likewise.
9146 * internal-fn.c (expand_call_mem_ref): New function.
9147 (expand_mask_load_optab_fn): Use it.
9148 (expand_mask_store_optab_fn): Likewise.
9150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9151 Alan Hayward <alan.hayward@arm.com>
9152 David Sherwood <david.sherwood@arm.com>
9154 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9155 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9156 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9157 (cond_umax@var{mode}): Document.
9158 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9159 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9160 (cond_umin_optab, cond_umax_optab): New optabs.
9161 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9162 (COND_IOR, COND_XOR): New internal functions.
9163 * internal-fn.h (get_conditional_internal_fn): Declare.
9164 * internal-fn.c (cond_binary_direct): New macro.
9165 (expand_cond_binary_optab_fn): Likewise.
9166 (direct_cond_binary_optab_supported_p): Likewise.
9167 (get_conditional_internal_fn): New function.
9168 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9169 Cope with reduction statements that are vectorized as calls rather
9171 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9172 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9173 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9174 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9175 (UNSPEC_COND_EOR): New unspecs.
9176 (optab): Add mappings for them.
9177 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9178 (sve_int_op, sve_fp_op): New int attributes.
9180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9181 Alan Hayward <alan.hayward@arm.com>
9182 David Sherwood <david.sherwood@arm.com>
9184 * optabs.def (while_ult_optab): New optab.
9185 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9186 * internal-fn.def (WHILE_ULT): New internal function.
9187 * internal-fn.h (direct_internal_fn_supported_p): New override
9188 that takes two types as argument.
9189 * internal-fn.c (while_direct): New macro.
9190 (expand_while_optab_fn): New function.
9191 (convert_optab_supported_p): Likewise.
9192 (direct_while_optab_supported_p): New macro.
9193 * wide-int.h (wi::udiv_ceil): New function.
9194 * tree-vectorizer.h (rgroup_masks): New structure.
9195 (vec_loop_masks): New typedef.
9196 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9198 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9199 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9200 (vect_max_vf): New function.
9201 (slpeel_make_loop_iterate_ntimes): Delete.
9202 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9203 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9204 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9205 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9206 internal-fn.h, stor-layout.h and optabs-query.h.
9207 (vect_set_loop_mask): New function.
9208 (add_preheader_seq): Likewise.
9209 (add_header_seq): Likewise.
9210 (interleave_supported_p): Likewise.
9211 (vect_maybe_permute_loop_masks): Likewise.
9212 (vect_set_loop_masks_directly): Likewise.
9213 (vect_set_loop_condition_masked): Likewise.
9214 (vect_set_loop_condition_unmasked): New function, split out from
9215 slpeel_make_loop_iterate_ntimes.
9216 (slpeel_make_loop_iterate_ntimes): Rename to..
9217 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9218 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9219 (vect_do_peeling): Update call accordingly.
9220 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9222 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9223 mask_compare_type, can_fully_mask_p and fully_masked_p.
9224 (release_vec_loop_masks): New function.
9225 (_loop_vec_info): Use it to free the loop masks.
9226 (can_produce_all_loop_masks_p): New function.
9227 (vect_get_max_nscalars_per_iter): Likewise.
9228 (vect_verify_full_masking): Likewise.
9229 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9230 retries, and free the mask rgroups before retrying. Check loop-wide
9231 reasons for disallowing fully-masked loops. Make the final decision
9232 about whether use a fully-masked loop or not.
9233 (vect_estimate_min_profitable_iters): Do not assume that peeling
9234 for the number of iterations will be needed for fully-masked loops.
9235 (vectorizable_reduction): Disable fully-masked loops.
9236 (vectorizable_live_operation): Likewise.
9237 (vect_halve_mask_nunits): New function.
9238 (vect_double_mask_nunits): Likewise.
9239 (vect_record_loop_mask): Likewise.
9240 (vect_get_loop_mask): Likewise.
9241 (vect_transform_loop): Handle the case in which the final loop
9242 iteration might handle a partial vector. Call vect_set_loop_condition
9243 instead of slpeel_make_loop_iterate_ntimes.
9244 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9245 (check_load_store_masking): New function.
9246 (prepare_load_store_mask): Likewise.
9247 (vectorizable_store): Handle fully-masked loops.
9248 (vectorizable_load): Likewise.
9249 (supportable_widening_operation): Use vect_halve_mask_nunits for
9251 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9252 (vect_gen_while): New function.
9253 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9254 (aarch64_uqdec<mode>): New insn.
9256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9257 Alan Hayward <alan.hayward@arm.com>
9258 David Sherwood <david.sherwood@arm.com>
9260 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9261 (reduc_xor_scal_optab): New optabs.
9262 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9263 (reduc_xor_scal_@var{m}): Document.
9264 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9265 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9267 * fold-const-call.c (fold_const_call): Handle them.
9268 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9269 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9270 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9271 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9272 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9273 (UNSPEC_XORV): New unspecs.
9274 (optab): Add entries for them.
9275 (BITWISEV): New int iterator.
9276 (bit_reduc_op): New int attributes.
9278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9279 Alan Hayward <alan.hayward@arm.com>
9280 David Sherwood <david.sherwood@arm.com>
9282 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9283 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9284 * optabs.def (vec_shl_insert_optab): New optab.
9285 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9286 (duplicate_and_interleave): Likewise.
9287 * tree-vect-loop.c: Include internal-fn.h.
9288 (neutral_op_for_slp_reduction): New function, split out from
9289 get_initial_defs_for_reduction.
9290 (get_initial_def_for_reduction): Handle option 2 for variable-length
9291 vectors by loading the neutral value into a vector and then shifting
9292 the initial value into element 0.
9293 (get_initial_defs_for_reduction): Replace the code argument with
9294 the neutral value calculated by neutral_op_for_slp_reduction.
9295 Use gimple_build_vector for constant-length vectors.
9296 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9297 but the first group_size elements have a neutral value.
9298 Use duplicate_and_interleave otherwise.
9299 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9300 Update call to get_initial_defs_for_reduction. Handle SLP
9301 reductions for variable-length vectors by creating one vector
9302 result for each scalar result, with the elements associated
9303 with other scalar results stubbed out with the neutral value.
9304 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9305 Require IFN_VEC_SHL_INSERT for double reductions on
9306 variable-length vectors, or SLP reductions that have
9307 a neutral value. Require can_duplicate_and_interleave_p
9308 support for variable-length unchained SLP reductions if there
9309 is no neutral value, such as for MIN/MAX reductions. Also require
9310 the number of vector elements to be a multiple of the number of
9311 SLP statements when doing variable-length unchained SLP reductions.
9312 Update call to vect_create_epilog_for_reduction.
9313 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9314 and remove initial values.
9315 (duplicate_and_interleave): Make public.
9316 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9317 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9320 Alan Hayward <alan.hayward@arm.com>
9321 David Sherwood <david.sherwood@arm.com>
9323 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9324 (can_duplicate_and_interleave_p): New function.
9325 (vect_get_and_check_slp_defs): Take the vector of statements
9326 rather than just the current one. Remove excess parentheses.
9327 Restriction rejectinon of vect_constant_def and vect_external_def
9328 for variable-length vectors to boolean types, or types for which
9329 can_duplicate_and_interleave_p is false.
9330 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9331 (duplicate_and_interleave): New function.
9332 (vect_get_constant_vectors): Use gimple_build_vector for
9333 constant-length vectors and suitable variable-length constant
9334 vectors. Use duplicate_and_interleave for other variable-length
9335 vectors. Don't defer the update when inserting new statements.
9337 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9338 Alan Hayward <alan.hayward@arm.com>
9339 David Sherwood <david.sherwood@arm.com>
9341 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9342 min_profitable_iters doesn't go negative.
9344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9345 Alan Hayward <alan.hayward@arm.com>
9346 David Sherwood <david.sherwood@arm.com>
9348 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9349 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9350 * optabs.def (vec_mask_load_lanes_optab): New optab.
9351 (vec_mask_store_lanes_optab): Likewise.
9352 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9353 (MASK_STORE_LANES): Likewise.
9354 * internal-fn.c (mask_load_lanes_direct): New macro.
9355 (mask_store_lanes_direct): Likewise.
9356 (expand_mask_load_optab_fn): Handle masked operations.
9357 (expand_mask_load_lanes_optab_fn): New macro.
9358 (expand_mask_store_optab_fn): Handle masked operations.
9359 (expand_mask_store_lanes_optab_fn): New macro.
9360 (direct_mask_load_lanes_optab_supported_p): Likewise.
9361 (direct_mask_store_lanes_optab_supported_p): Likewise.
9362 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9364 (vect_load_lanes_supported): Likewise.
9365 * tree-vect-data-refs.c (strip_conversion): New function.
9366 (can_group_stmts_p): Likewise.
9367 (vect_analyze_data_ref_accesses): Use it instead of checking
9368 for a pair of assignments.
9369 (vect_store_lanes_supported): Take a masked_p parameter.
9370 (vect_load_lanes_supported): Likewise.
9371 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9372 vect_store_lanes_supported and vect_load_lanes_supported.
9373 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9374 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9375 parameter. Don't allow gaps for masked accesses.
9376 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9377 and vect_load_lanes_supported.
9378 (get_load_store_type): Take a masked_p parameter and update
9379 call to get_group_load_store_type.
9380 (vectorizable_store): Update call to get_load_store_type.
9381 Handle IFN_MASK_STORE_LANES.
9382 (vectorizable_load): Update call to get_load_store_type.
9383 Handle IFN_MASK_LOAD_LANES.
9385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9386 Alan Hayward <alan.hayward@arm.com>
9387 David Sherwood <david.sherwood@arm.com>
9389 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9391 * config/aarch64/aarch64-protos.h
9392 (aarch64_sve_struct_memory_operand_p): Declare.
9393 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9394 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9395 (VPRED, vpred): Handle SVE structure modes.
9396 * config/aarch64/constraints.md (Utx): New constraint.
9397 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9398 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9399 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9400 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9401 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9402 structure modes. Split into pieces after RA.
9403 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9404 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9406 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9407 SVE structure modes.
9408 (aarch64_classify_address): Likewise.
9409 (sizetochar): Move earlier in file.
9410 (aarch64_print_operand): Handle SVE register lists.
9411 (aarch64_array_mode): New function.
9412 (aarch64_sve_struct_memory_operand_p): Likewise.
9413 (TARGET_ARRAY_MODE): Redefine.
9415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9416 Alan Hayward <alan.hayward@arm.com>
9417 David Sherwood <david.sherwood@arm.com>
9419 * target.def (array_mode): New target hook.
9420 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9421 * doc/tm.texi: Regenerate.
9422 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9423 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9424 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9426 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9430 Alan Hayward <alan.hayward@arm.com>
9431 David Sherwood <david.sherwood@arm.com>
9433 * fold-const.c (fold_binary_loc): Check the argument types
9434 rather than the result type when testing for a vector operation.
9436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9438 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9439 * doc/tm.texi: Regenerate.
9441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9442 Alan Hayward <alan.hayward@arm.com>
9443 David Sherwood <david.sherwood@arm.com>
9445 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9446 (sve): Document new AArch64 extension.
9447 * doc/md.texi (w): Extend the description of the AArch64
9448 constraint to include SVE vectors.
9449 (Upl, Upa): Document new AArch64 predicate constraints.
9450 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9452 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9453 (msve-vector-bits=): New option.
9454 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9455 SVE when these are disabled.
9456 (sve): New extension.
9457 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9458 modes. Adjust their number of units based on aarch64_sve_vg.
9459 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9460 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9461 aarch64_addr_query_type.
9462 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9463 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9464 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9465 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9466 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9467 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9468 (aarch64_simd_imm_zero_p): Delete.
9469 (aarch64_check_zero_based_sve_index_immediate): Declare.
9470 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9471 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9472 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9473 (aarch64_sve_float_mul_immediate_p): Likewise.
9474 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9476 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9477 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9478 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9479 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9480 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9481 (aarch64_regmode_natural_size): Likewise.
9482 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9483 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9485 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9486 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9487 for VG and the SVE predicate registers.
9488 (V_ALIASES): Add a "z"-prefixed alias.
9489 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9490 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9491 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9492 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9493 (REG_CLASS_NAMES): Add entries for them.
9494 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9495 and the predicate registers.
9496 (aarch64_sve_vg): Declare.
9497 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9498 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9499 (REGMODE_NATURAL_SIZE): Define.
9500 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9502 * config/aarch64/aarch64.c: Include cfgrtl.h.
9503 (simd_immediate_info): Add a constructor for series vectors,
9504 and an associated step field.
9505 (aarch64_sve_vg): New variable.
9506 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9507 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9508 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9509 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9510 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9511 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9512 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9513 (aarch64_get_mask_mode): New functions.
9514 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9515 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9516 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9517 predicate modes and predicate registers. Explicitly restrict
9518 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9519 to store a vector mode if it is recognized by
9520 aarch64_classify_vector_mode.
9521 (aarch64_regmode_natural_size): New function.
9522 (aarch64_hard_regno_caller_save_mode): Return the original mode
9524 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9525 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9526 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9527 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9529 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9530 does not overlap dest if the function is frame-related. Handle
9532 (aarch64_split_add_offset): New function.
9533 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9534 them aarch64_add_offset.
9535 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9536 and update call to aarch64_sub_sp.
9537 (aarch64_add_cfa_expression): New function.
9538 (aarch64_expand_prologue): Pass extra temporary registers to the
9539 functions above. Handle the case in which we need to emit new
9540 DW_CFA_expressions for registers that were originally saved
9541 relative to the stack pointer, but now have to be expressed
9542 relative to the frame pointer.
9543 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9545 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9546 IP0 and IP1 values for SVE frames.
9547 (aarch64_expand_vec_series): New function.
9548 (aarch64_expand_sve_widened_duplicate): Likewise.
9549 (aarch64_expand_sve_const_vector): Likewise.
9550 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9551 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9552 into the register, rather than emitting a SET directly.
9553 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9554 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9555 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9556 (offset_9bit_signed_scaled_p): New functions.
9557 (aarch64_replicate_bitmask_imm): New function.
9558 (aarch64_bitmask_imm): Use it.
9559 (aarch64_cannot_force_const_mem): Reject expressions involving
9560 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9561 (aarch64_classify_index): Handle SVE indices, by requiring
9562 a plain register index with a scale that matches the element size.
9563 (aarch64_classify_address): Handle SVE addresses. Assert that
9564 the mode of the address is VOIDmode or an integer mode.
9565 Update call to aarch64_classify_symbol.
9566 (aarch64_classify_symbolic_expression): Update call to
9567 aarch64_classify_symbol.
9568 (aarch64_const_vec_all_in_range_p): New function.
9569 (aarch64_print_vector_float_operand): Likewise.
9570 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9571 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9572 and the FP immediates 1.0 and 0.5.
9573 (aarch64_print_address_internal): Handle SVE addresses.
9574 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9575 (aarch64_regno_regclass): Handle predicate registers.
9576 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9578 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9579 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9580 (aarch64_convert_sve_vector_bits): New function.
9581 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9582 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9584 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9585 Handle SVE vector and predicate modes. Accept VL-based constants
9586 that need only one temporary register, and VL offsets that require
9587 no temporary registers.
9588 (aarch64_conditional_register_usage): Mark the predicate registers
9589 as fixed if SVE isn't available.
9590 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9591 Return true for SVE vector and predicate modes.
9592 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9593 rather than an unsigned int. Handle SVE modes.
9594 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9596 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9598 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9599 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9600 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9601 (aarch64_sve_float_mul_immediate_p): New functions.
9602 (aarch64_sve_valid_immediate): New function.
9603 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9604 Explicitly reject structure modes. Check for INDEX constants.
9605 Handle PTRUE and PFALSE constants.
9606 (aarch64_check_zero_based_sve_index_immediate): New function.
9607 (aarch64_simd_imm_zero_p): Delete.
9608 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9609 vector modes. Accept constants in the range of CNT[BHWD].
9610 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9611 ask for an Advanced SIMD mode.
9612 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9613 (aarch64_simd_vector_alignment): Handle SVE predicates.
9614 (aarch64_vectorize_preferred_vector_alignment): New function.
9615 (aarch64_simd_vector_alignment_reachable): Use it instead of
9617 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9618 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9620 (MAX_VECT_LEN): Delete.
9621 (expand_vec_perm_d): Add a vec_flags field.
9622 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9623 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9624 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9626 (aarch64_evpc_rev): Rename to...
9627 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9628 (aarch64_evpc_rev_global): New function.
9629 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9630 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9632 (aarch64_evpc_sve_tbl): New function.
9633 (aarch64_expand_vec_perm_const_1): Update after rename of
9634 aarch64_evpc_rev. Handle SVE permutes too, trying
9635 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9636 than aarch64_evpc_tbl.
9637 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9638 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9639 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9640 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9641 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9642 (aarch64_expand_sve_vcond): New functions.
9643 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9644 of aarch64_vector_mode_p.
9645 (aarch64_dwarf_poly_indeterminate_value): New function.
9646 (aarch64_compute_pressure_classes): Likewise.
9647 (aarch64_can_change_mode_class): Likewise.
9648 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9649 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9650 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9651 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9652 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9653 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9654 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9655 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9657 (Dn, Dl, Dr): Accept const as well as const_vector.
9658 (Dz): Likewise. Compare against CONST0_RTX.
9659 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9660 of "vector" where appropriate.
9661 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9662 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9663 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9664 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9665 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9666 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9667 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9668 (v_int_equiv): Extend to SVE modes.
9669 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9671 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9672 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9673 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9674 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9675 (SVE_COND_FP_CMP): New int iterators.
9676 (perm_hilo): Handle the new unpack unspecs.
9677 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9679 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9680 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9681 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9682 (aarch64_equality_operator, aarch64_constant_vector_operand)
9683 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9684 (aarch64_sve_nonimmediate_operand): Likewise.
9685 (aarch64_sve_general_operand): Likewise.
9686 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9687 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9688 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9689 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9690 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9691 (aarch64_sve_float_arith_immediate): Likewise.
9692 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9693 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9694 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9695 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9696 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9697 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9698 (aarch64_sve_float_arith_operand): Likewise.
9699 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9700 (aarch64_sve_float_mul_operand): Likewise.
9701 (aarch64_sve_vec_perm_operand): Likewise.
9702 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9703 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9704 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9705 as well as const_vector.
9706 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9707 in file. Use CONST0_RTX and CONSTM1_RTX.
9708 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9709 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9710 Use aarch64_simd_imm_zero.
9711 * config/aarch64/aarch64-sve.md: New file.
9712 * config/aarch64/aarch64.md: Include it.
9713 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9714 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9715 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9716 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9717 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9718 (sve): New attribute.
9719 (enabled): Disable instructions with the sve attribute unless
9721 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9722 aarch64_expand_mov_immediate.
9723 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9724 CNT[BHSD] immediates.
9725 (movti): Split CONST_POLY_INT moves into two halves.
9726 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9727 Split additions that need a temporary here if the destination
9728 is the stack pointer.
9729 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9730 (*add<mode>3_poly_1): New instruction.
9731 (set_clobber_cc): New expander.
9733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9735 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9736 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9737 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9738 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9739 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9740 Change innermode from fixed_mode_size to machine_mode.
9741 (simplify_subreg): Update call accordingly. Handle a constant-sized
9742 subreg of a variable-length CONST_VECTOR.
9744 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9745 Alan Hayward <alan.hayward@arm.com>
9746 David Sherwood <david.sherwood@arm.com>
9748 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9749 (add_offset_to_base): New function, split out from...
9750 (create_mem_ref): ...here. When handling a scale other than 1,
9751 check first whether the address is valid without the offset.
9752 Add it into the base if so, leaving the index and scale as-is.
9754 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9757 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9758 fold_for_warn before checking if arg2 is INTEGER_CST.
9760 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9762 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9763 (store_multiple_operation): Delete.
9764 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9765 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9766 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9767 guarded by TARGET_STRING.
9768 (rs6000_output_load_multiple): Delete.
9769 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9770 OPTION_MASK_STRING / TARGET_STRING handling.
9771 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9772 (const rs6000_opt_masks) <"string">: Change mask to 0.
9773 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9774 (MASK_STRING): Delete.
9775 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9777 (load_multiple): Delete.
9784 (store_multiple): Delete.
9791 (movmemsi_8reg): Delete.
9792 (corresponding unnamed define_insn): Delete.
9793 (movmemsi_6reg): Delete.
9794 (corresponding unnamed define_insn): Delete.
9795 (movmemsi_4reg): Delete.
9796 (corresponding unnamed define_insn): Delete.
9797 (movmemsi_2reg): Delete.
9798 (corresponding unnamed define_insn): Delete.
9799 (movmemsi_1reg): Delete.
9800 (corresponding unnamed define_insn): Delete.
9801 * config/rs6000/rs6000.opt (mno-string): New.
9802 (mstring): Replace by deprecation warning stub.
9803 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9805 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9807 * regrename.c (regrename_do_replace): If replacing the same
9808 reg multiple times, try to reuse last created gen_raw_REG.
9811 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9812 main to workaround a bug in GDB.
9814 2018-01-12 Tom de Vries <tom@codesourcery.com>
9817 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9819 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9821 PR rtl-optimization/80481
9822 * ira-color.c (get_cap_member): New function.
9823 (allocnos_conflict_by_live_ranges_p): Use it.
9824 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9825 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9827 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9830 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9831 (*saddl_se_1): Ditto.
9833 (*ssubl_se_1): Ditto.
9835 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9837 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9838 rather than wi::to_widest for DR_INITs.
9839 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9840 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9841 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9843 (vect_analyze_group_access_1): Note that here.
9845 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9847 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9848 polynomial type sizes.
9850 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9852 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9853 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9854 (gimple_add_tmp_var): Likewise.
9856 2018-01-12 Martin Liska <mliska@suse.cz>
9858 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9859 (gimple_alloc_sizes): Likewise.
9860 (dump_gimple_statistics): Use PRIu64 in printf format.
9861 * gimple.h: Change uint64_t to int.
9863 2018-01-12 Martin Liska <mliska@suse.cz>
9865 * tree-core.h: Use uint64_t instead of int.
9866 * tree.c (tree_node_counts): Likewise.
9867 (tree_node_sizes): Likewise.
9868 (dump_tree_statistics): Use PRIu64 in printf format.
9870 2018-01-12 Martin Liska <mliska@suse.cz>
9872 * Makefile.in: As qsort_chk is implemented in vec.c, add
9873 vec.o to linkage of gencfn-macros.
9874 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9875 passing the info to record_node_allocation_statistics.
9876 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9878 * ggc-common.c (struct ggc_usage): Add operator== and use
9879 it in operator< and compare function.
9880 * mem-stats.h (struct mem_usage): Likewise.
9881 * vec.c (struct vec_usage): Remove operator< and compare
9882 function. Can be simply inherited.
9884 2018-01-12 Martin Jambor <mjambor@suse.cz>
9887 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9888 * tree-ssa-math-opts.c: Include domwalk.h.
9889 (convert_mult_to_fma_1): New function.
9890 (fma_transformation_info): New type.
9891 (fma_deferring_state): Likewise.
9892 (cancel_fma_deferring): New function.
9893 (result_of_phi): Likewise.
9894 (last_fma_candidate_feeds_initial_phi): Likewise.
9895 (convert_mult_to_fma): Added deferring logic, split actual
9896 transformation to convert_mult_to_fma_1.
9897 (math_opts_dom_walker): New type.
9898 (math_opts_dom_walker::after_dom_children): New method, body moved
9899 here from pass_optimize_widening_mul::execute, added deferring logic
9901 (pass_optimize_widening_mul::execute): Moved most of code to
9902 math_opts_dom_walker::after_dom_children.
9903 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9904 * config/i386/i386.c (ix86_option_override_internal): Added
9905 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9907 2018-01-12 Richard Biener <rguenther@suse.de>
9910 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9911 inline instance vars.
9913 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9916 * config/rx/rx.c (rx_is_restricted_memory_address):
9919 2018-01-12 Richard Biener <rguenther@suse.de>
9921 PR tree-optimization/80846
9922 * target.def (split_reduction): New target hook.
9923 * targhooks.c (default_split_reduction): New function.
9924 * targhooks.h (default_split_reduction): Declare.
9925 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9926 target requests first reduce vectors by combining low and high
9928 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9929 (get_vectype_for_scalar_type_and_size): Export.
9930 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9931 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9932 * doc/tm.texi: Regenerate.
9933 * config/i386/i386.c (ix86_split_reduction): Implement
9934 TARGET_VECTORIZE_SPLIT_REDUCTION.
9936 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9939 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9940 in PIC mode except for TARGET_VXWORKS_RTP.
9941 * config/sparc/sparc.c: Include cfgrtl.h.
9942 (TARGET_INIT_PIC_REG): Define.
9943 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9944 (sparc_pic_register_p): New predicate.
9945 (sparc_legitimate_address_p): Use it.
9946 (sparc_legitimize_pic_address): Likewise.
9947 (sparc_delegitimize_address): Likewise.
9948 (sparc_mode_dependent_address_p): Likewise.
9949 (gen_load_pcrel_sym): Remove 4th parameter.
9950 (load_got_register): Adjust call to above. Remove obsolete stuff.
9951 (sparc_expand_prologue): Do not call load_got_register here.
9952 (sparc_flat_expand_prologue): Likewise.
9953 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9954 (sparc_use_pseudo_pic_reg): New function.
9955 (sparc_init_pic_reg): Likewise.
9956 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9957 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9959 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9961 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9962 Add item for branch_cost.
9964 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9966 PR rtl-optimization/83565
9967 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9968 not extend the result to a larger mode for rotate operations.
9969 (num_sign_bit_copies1): Likewise.
9971 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9974 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9976 Use values-Xc.o for -pedantic.
9977 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9979 2018-01-12 Martin Liska <mliska@suse.cz>
9982 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9984 (possible_polymorphic_call_targets): Use it.
9985 (ipa_devirt): Likewise.
9987 2018-01-12 Martin Liska <mliska@suse.cz>
9989 * profile-count.h (enum profile_quality): Use 0 as invalid
9990 enum value of profile_quality.
9992 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9994 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9995 -mext-string options.
9997 2018-01-12 Richard Biener <rguenther@suse.de>
9999 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10000 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10001 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10003 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10005 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10007 * configure.ac (--with-long-double-format): Add support for the
10008 configuration option to change the default long double format on
10010 * config.gcc (powerpc*-linux*-*): Likewise.
10011 * configure: Regenerate.
10012 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10013 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10014 used without modification.
10016 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10018 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10019 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10020 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10021 MISC_BUILTIN_SPEC_BARRIER.
10022 (rs6000_init_builtins): Likewise.
10023 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10025 (speculation_barrier): New define_insn.
10026 * doc/extend.texi: Document __builtin_speculation_barrier.
10028 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10031 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10032 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10033 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10035 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10036 integral modes instead of "ss" and "sd".
10037 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10038 vectors with 32-bit and 64-bit elements.
10039 (vecdupssescalarmodesuffix): New mode attribute.
10040 (vec_dup<mode>): Use it.
10042 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10045 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10046 frame if argument is passed on stack.
10048 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10051 * ree.c (combine_reaching_defs): Optimize also
10052 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10053 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10055 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10057 PR middle-end/83189
10058 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10060 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10062 PR middle-end/83718
10063 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10064 after they are computed.
10066 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10068 PR tree-optimization/83695
10069 * gimple-loop-linterchange.cc
10070 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10071 reset cached scev information after interchange.
10072 (pass_linterchange::execute): Remove call to scev_reset_htab.
10074 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10076 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10077 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10078 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10079 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10080 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10081 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10082 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10083 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10084 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10085 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10086 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10087 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10088 (V_lane_reg): Likewise.
10089 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10091 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10092 (vfmal_lane_low<mode>_intrinsic,
10093 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10094 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10095 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10096 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10097 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10098 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10100 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10102 * config/arm/arm-cpus.in (fp16fml): New feature.
10103 (ALL_SIMD): Add fp16fml.
10104 (armv8.2-a): Add fp16fml as an option.
10105 (armv8.3-a): Likewise.
10106 (armv8.4-a): Add fp16fml as part of fp16.
10107 * config/arm/arm.h (TARGET_FP16FML): Define.
10108 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10110 * config/arm/arm-modes.def (V2HF): Define.
10111 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10112 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10113 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10114 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10115 vfmsl_low, vfmsl_high): New set of builtins.
10116 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10117 (vfml_op): New code attribute.
10118 (VFMLHALVES): New int iterator.
10119 (VFML, VFMLSEL): New mode attributes.
10120 (V_reg): Define mapping for V2HF.
10121 (V_hi, V_lo): New mode attributes.
10122 (VF_constraint): Likewise.
10123 (vfml_half, vfml_half_selector): New int attributes.
10124 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10126 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10127 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10129 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10130 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10131 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10132 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10134 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10135 Document new effective target and option set.
10137 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10139 * config/arm/arm-cpus.in (armv8_4): New feature.
10140 (ARMv8_4a): New fgroup.
10141 (armv8.4-a): New arch.
10142 * config/arm/arm-tables.opt: Regenerate.
10143 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10144 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10145 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10146 Add matching rules for -march=armv8.4-a and extensions.
10147 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10149 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10152 * config/rx/rx.md (BW): New mode attribute.
10153 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10155 2018-01-11 Richard Biener <rguenther@suse.de>
10157 PR tree-optimization/83435
10158 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10159 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10160 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10162 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10163 Alan Hayward <alan.hayward@arm.com>
10164 David Sherwood <david.sherwood@arm.com>
10166 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10168 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10169 (aarch64_print_address_internal): Use it to check for a zero offset.
10171 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10172 Alan Hayward <alan.hayward@arm.com>
10173 David Sherwood <david.sherwood@arm.com>
10175 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10176 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10177 Return a poly_int64 rather than a HOST_WIDE_INT.
10178 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10179 rather than a HOST_WIDE_INT.
10180 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10181 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10182 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10183 final_offset from HOST_WIDE_INT to poly_int64.
10184 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10185 to_constant when getting the number of units in an Advanced SIMD
10187 (aarch64_builtin_vectorized_function): Check for a constant number
10189 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10191 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10192 attribute instead of GET_MODE_NUNITS.
10193 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10194 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10195 GET_MODE_SIZE for fixed-size registers.
10196 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10197 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10198 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10199 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10200 (aarch64_print_operand, aarch64_print_address_internal)
10201 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10202 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10203 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10204 Handle polynomial GET_MODE_SIZE.
10205 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10206 wider than SImode without modification.
10207 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10208 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10209 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10210 passing and returning SVE modes.
10211 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10212 rather than GEN_INT.
10213 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10214 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10215 (aarch64_allocate_and_probe_stack_space): Likewise.
10216 (aarch64_layout_frame): Cope with polynomial offsets.
10217 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10218 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10219 polynomial offsets.
10220 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10221 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10222 poly_int64 rather than a HOST_WIDE_INT.
10223 (aarch64_get_separate_components, aarch64_process_components)
10224 (aarch64_expand_prologue, aarch64_expand_epilogue)
10225 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10226 (aarch64_anchor_offset): New function, split out from...
10227 (aarch64_legitimize_address): ...here.
10228 (aarch64_builtin_vectorization_cost): Handle polynomial
10229 TYPE_VECTOR_SUBPARTS.
10230 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10232 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10233 number of elements from the PARALLEL rather than the mode.
10234 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10235 rather than GET_MODE_BITSIZE.
10236 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10237 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10238 (aarch64_expand_vec_perm_const_1): Handle polynomial
10239 d->perm.length () and d->perm elements.
10240 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10241 Apply to_constant to d->perm elements.
10242 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10243 polynomial CONST_VECTOR_NUNITS.
10244 (aarch64_move_pointer): Take amount as a poly_int64 rather
10246 (aarch64_progress_pointer): Avoid temporary variable.
10247 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10248 the mode attribute instead of GET_MODE.
10250 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10251 Alan Hayward <alan.hayward@arm.com>
10252 David Sherwood <david.sherwood@arm.com>
10254 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10255 x exists before using it.
10256 (aarch64_add_constant_internal): Rename to...
10257 (aarch64_add_offset_1): ...this. Replace regnum with separate
10258 src and dest rtxes. Handle the case in which they're different,
10259 including when the offset is zero. Replace scratchreg with an rtx.
10260 Use 2 additions if there is no spare register into which we can
10261 move a 16-bit constant.
10262 (aarch64_add_constant): Delete.
10263 (aarch64_add_offset): Replace reg with separate src and dest
10264 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10265 Use aarch64_add_offset_1.
10266 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10267 an rtx rather than an int. Take the delta as a poly_int64
10268 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10269 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10270 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10271 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10272 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10273 and aarch64_add_sp.
10274 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10275 aarch64_add_constant.
10277 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10279 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10280 Use scalar_float_mode.
10282 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10284 * config/aarch64/aarch64-simd.md
10285 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10286 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10287 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10288 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10289 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10290 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10291 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10292 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10293 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10294 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10296 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10299 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10300 targ_options->x_arm_arch_string is non NULL.
10302 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10304 * config/aarch64/aarch64.h
10305 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10307 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10310 * expmed.c (emit_store_flag_force): Swap if const op0
10311 and change VOIDmode to mode of op0.
10313 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10315 PR rtl-optimization/83761
10316 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10317 than bytes to mode_for_size.
10319 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10321 PR middle-end/83189
10322 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10323 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10326 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10328 PR middle-end/83575
10329 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10330 when in layout mode.
10331 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10332 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10335 2018-01-10 Michael Collison <michael.collison@arm.com>
10337 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10338 * config/aarch64/aarch64-option-extension.def: Add
10339 AARCH64_OPT_EXTENSION of 'fp16fml'.
10340 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10341 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10342 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10343 * config/aarch64/constraints.md (Ui7): New constraint.
10344 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10345 (VFMLA_SEL_W): Ditto.
10348 (VFMLA16_LOW): New int iterator.
10349 (VFMLA16_HIGH): Ditto.
10350 (UNSPEC_FMLAL): New unspec.
10351 (UNSPEC_FMLSL): Ditto.
10352 (UNSPEC_FMLAL2): Ditto.
10353 (UNSPEC_FMLSL2): Ditto.
10354 (f16mac): New code attribute.
10355 * config/aarch64/aarch64-simd-builtins.def
10356 (aarch64_fmlal_lowv2sf): Ditto.
10357 (aarch64_fmlsl_lowv2sf): Ditto.
10358 (aarch64_fmlalq_lowv4sf): Ditto.
10359 (aarch64_fmlslq_lowv4sf): Ditto.
10360 (aarch64_fmlal_highv2sf): Ditto.
10361 (aarch64_fmlsl_highv2sf): Ditto.
10362 (aarch64_fmlalq_highv4sf): Ditto.
10363 (aarch64_fmlslq_highv4sf): Ditto.
10364 (aarch64_fmlal_lane_lowv2sf): Ditto.
10365 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10366 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10367 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10368 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10369 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10370 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10371 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10372 (aarch64_fmlal_lane_highv2sf): Ditto.
10373 (aarch64_fmlsl_lane_highv2sf): Ditto.
10374 (aarch64_fmlal_laneq_highv2sf): Ditto.
10375 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10376 (aarch64_fmlalq_lane_highv4sf): Ditto.
10377 (aarch64_fmlsl_lane_highv4sf): Ditto.
10378 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10379 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10380 * config/aarch64/aarch64-simd.md:
10381 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10382 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10383 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10384 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10385 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10386 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10387 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10388 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10389 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10390 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10391 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10392 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10393 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10394 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10395 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10396 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10397 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10398 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10399 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10400 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10401 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10402 (vfmlsl_low_u32): Ditto.
10403 (vfmlalq_low_u32): Ditto.
10404 (vfmlslq_low_u32): Ditto.
10405 (vfmlal_high_u32): Ditto.
10406 (vfmlsl_high_u32): Ditto.
10407 (vfmlalq_high_u32): Ditto.
10408 (vfmlslq_high_u32): Ditto.
10409 (vfmlal_lane_low_u32): Ditto.
10410 (vfmlsl_lane_low_u32): Ditto.
10411 (vfmlal_laneq_low_u32): Ditto.
10412 (vfmlsl_laneq_low_u32): Ditto.
10413 (vfmlalq_lane_low_u32): Ditto.
10414 (vfmlslq_lane_low_u32): Ditto.
10415 (vfmlalq_laneq_low_u32): Ditto.
10416 (vfmlslq_laneq_low_u32): Ditto.
10417 (vfmlal_lane_high_u32): Ditto.
10418 (vfmlsl_lane_high_u32): Ditto.
10419 (vfmlal_laneq_high_u32): Ditto.
10420 (vfmlsl_laneq_high_u32): Ditto.
10421 (vfmlalq_lane_high_u32): Ditto.
10422 (vfmlslq_lane_high_u32): Ditto.
10423 (vfmlalq_laneq_high_u32): Ditto.
10424 (vfmlslq_laneq_high_u32): Ditto.
10425 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10426 (AARCH64_FL_FOR_ARCH8_4): New.
10427 (AARCH64_ISA_F16FML): New ISA flag.
10428 (TARGET_F16FML): New feature flag for fp16fml.
10429 (doc/invoke.texi): Document new fp16fml option.
10431 2018-01-10 Michael Collison <michael.collison@arm.com>
10433 * config/aarch64/aarch64-builtins.c:
10434 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10435 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10436 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10437 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10438 (AARCH64_ISA_SHA3): New ISA flag.
10439 (TARGET_SHA3): New feature flag for sha3.
10440 * config/aarch64/iterators.md (sha512_op): New int attribute.
10441 (CRYPTO_SHA512): New int iterator.
10442 (UNSPEC_SHA512H): New unspec.
10443 (UNSPEC_SHA512H2): Ditto.
10444 (UNSPEC_SHA512SU0): Ditto.
10445 (UNSPEC_SHA512SU1): Ditto.
10446 * config/aarch64/aarch64-simd-builtins.def
10447 (aarch64_crypto_sha512hqv2di): New builtin.
10448 (aarch64_crypto_sha512h2qv2di): Ditto.
10449 (aarch64_crypto_sha512su0qv2di): Ditto.
10450 (aarch64_crypto_sha512su1qv2di): Ditto.
10451 (aarch64_eor3qv8hi): Ditto.
10452 (aarch64_rax1qv2di): Ditto.
10453 (aarch64_xarqv2di): Ditto.
10454 (aarch64_bcaxqv8hi): Ditto.
10455 * config/aarch64/aarch64-simd.md:
10456 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10457 (aarch64_crypto_sha512su0qv2di): Ditto.
10458 (aarch64_crypto_sha512su1qv2di): Ditto.
10459 (aarch64_eor3qv8hi): Ditto.
10460 (aarch64_rax1qv2di): Ditto.
10461 (aarch64_xarqv2di): Ditto.
10462 (aarch64_bcaxqv8hi): Ditto.
10463 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10464 (vsha512h2q_u64): Ditto.
10465 (vsha512su0q_u64): Ditto.
10466 (vsha512su1q_u64): Ditto.
10467 (veor3q_u16): Ditto.
10468 (vrax1q_u64): Ditto.
10469 (vxarq_u64): Ditto.
10470 (vbcaxq_u16): Ditto.
10471 * config/arm/types.md (crypto_sha512): New type attribute.
10472 (crypto_sha3): Ditto.
10473 (doc/invoke.texi): Document new sha3 option.
10475 2018-01-10 Michael Collison <michael.collison@arm.com>
10477 * config/aarch64/aarch64-builtins.c:
10478 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10479 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10480 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10481 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10482 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10483 (AARCH64_ISA_SM4): New ISA flag.
10484 (TARGET_SM4): New feature flag for sm4.
10485 * config/aarch64/aarch64-simd-builtins.def
10486 (aarch64_sm3ss1qv4si): Ditto.
10487 (aarch64_sm3tt1aq4si): Ditto.
10488 (aarch64_sm3tt1bq4si): Ditto.
10489 (aarch64_sm3tt2aq4si): Ditto.
10490 (aarch64_sm3tt2bq4si): Ditto.
10491 (aarch64_sm3partw1qv4si): Ditto.
10492 (aarch64_sm3partw2qv4si): Ditto.
10493 (aarch64_sm4eqv4si): Ditto.
10494 (aarch64_sm4ekeyqv4si): Ditto.
10495 * config/aarch64/aarch64-simd.md:
10496 (aarch64_sm3ss1qv4si): Ditto.
10497 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10498 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10499 (aarch64_sm4eqv4si): Ditto.
10500 (aarch64_sm4ekeyqv4si): Ditto.
10501 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10502 (sm3part_op): Ditto.
10503 (CRYPTO_SM3TT): Ditto.
10504 (CRYPTO_SM3PART): Ditto.
10505 (UNSPEC_SM3SS1): New unspec.
10506 (UNSPEC_SM3TT1A): Ditto.
10507 (UNSPEC_SM3TT1B): Ditto.
10508 (UNSPEC_SM3TT2A): Ditto.
10509 (UNSPEC_SM3TT2B): Ditto.
10510 (UNSPEC_SM3PARTW1): Ditto.
10511 (UNSPEC_SM3PARTW2): Ditto.
10512 (UNSPEC_SM4E): Ditto.
10513 (UNSPEC_SM4EKEY): Ditto.
10514 * config/aarch64/constraints.md (Ui2): New constraint.
10515 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10516 * config/arm/types.md (crypto_sm3): New type attribute.
10517 (crypto_sm4): Ditto.
10518 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10519 (vsm3tt1aq_u32): Ditto.
10520 (vsm3tt1bq_u32): Ditto.
10521 (vsm3tt2aq_u32): Ditto.
10522 (vsm3tt2bq_u32): Ditto.
10523 (vsm3partw1q_u32): Ditto.
10524 (vsm3partw2q_u32): Ditto.
10525 (vsm4eq_u32): Ditto.
10526 (vsm4ekeyq_u32): Ditto.
10527 (doc/invoke.texi): Document new sm4 option.
10529 2018-01-10 Michael Collison <michael.collison@arm.com>
10531 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10532 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10533 (AARCH64_FL_FOR_ARCH8_4): New.
10534 (AARCH64_FL_V8_4): New flag.
10535 (doc/invoke.texi): Document new armv8.4-a option.
10537 2018-01-10 Michael Collison <michael.collison@arm.com>
10539 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10540 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10541 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10542 * config/aarch64/aarch64-option-extension.def: Add
10543 AARCH64_OPT_EXTENSION of 'sha2'.
10544 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10545 (crypto): Disable sha2 and aes if crypto disabled.
10546 (crypto): Enable aes and sha2 if enabled.
10547 (simd): Disable sha2 and aes if simd disabled.
10548 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10550 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10551 (TARGET_SHA2): New feature flag for sha2.
10552 (TARGET_AES): New feature flag for aes.
10553 * config/aarch64/aarch64-simd.md:
10554 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10555 conditional on TARGET_AES.
10556 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10557 (aarch64_crypto_sha1hsi): Make pattern conditional
10559 (aarch64_crypto_sha1hv4si): Ditto.
10560 (aarch64_be_crypto_sha1hv4si): Ditto.
10561 (aarch64_crypto_sha1su1v4si): Ditto.
10562 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10563 (aarch64_crypto_sha1su0v4si): Ditto.
10564 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10565 (aarch64_crypto_sha256su0v4si): Ditto.
10566 (aarch64_crypto_sha256su1v4si): Ditto.
10567 (doc/invoke.texi): Document new aes and sha2 options.
10569 2018-01-10 Martin Sebor <msebor@redhat.com>
10571 PR tree-optimization/83781
10572 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10575 2018-01-11 Martin Sebor <msebor@gmail.com>
10576 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10578 PR tree-optimization/83501
10579 PR tree-optimization/81703
10581 * tree-ssa-strlen.c (get_string_cst): Rename...
10582 (get_string_len): ...to this. Handle global constants.
10583 (handle_char_store): Adjust.
10585 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10586 Jim Wilson <jimw@sifive.com>
10588 * config/riscv/riscv-protos.h (riscv_output_return): New.
10589 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10590 (riscv_attribute_table, riscv_output_return),
10591 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10592 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10593 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10595 (riscv_expand_prologue): Add early return for naked function.
10596 (riscv_expand_epilogue): Likewise.
10597 (riscv_function_ok_for_sibcall): Return false for naked function.
10598 (riscv_set_current_function): New.
10599 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10600 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10601 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10602 * doc/extend.texi (RISC-V Function Attributes): New.
10604 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10606 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10607 check for 128-bit long double before checking TCmode.
10608 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10609 128-bit long doubles before checking TFmode or TCmode.
10610 (FLOAT128_IBM_P): Likewise.
10612 2018-01-10 Martin Sebor <msebor@redhat.com>
10614 PR tree-optimization/83671
10615 * builtins.c (c_strlen): Unconditionally return zero for the empty
10617 Use -Warray-bounds for warnings.
10618 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10619 for non-constant array indices with COMPONENT_REF, arrays of
10620 arrays, and pointers to arrays.
10621 (gimple_fold_builtin_strlen): Determine and set length range for
10622 non-constant character arrays.
10624 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10626 PR middle-end/81897
10627 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10630 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10632 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10634 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10637 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10638 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10639 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10640 indexed_or_indirect_operand predicate.
10641 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10642 (*vsx_le_perm_load_v8hi): Likewise.
10643 (*vsx_le_perm_load_v16qi): Likewise.
10644 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10645 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10646 (*vsx_le_perm_store_v8hi): Likewise.
10647 (*vsx_le_perm_store_v16qi): Likewise.
10648 (eight unnamed splitters): Likewise.
10650 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10652 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10653 * config/rs6000/emmintrin.h: Likewise.
10654 * config/rs6000/mmintrin.h: Likewise.
10655 * config/rs6000/xmmintrin.h: Likewise.
10657 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10660 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10662 * tree.c (tree_nop_conversion): Return true for location wrapper
10664 (maybe_wrap_with_location): New function.
10665 (selftest::check_strip_nops): New function.
10666 (selftest::test_location_wrappers): New function.
10667 (selftest::tree_c_tests): Call it.
10668 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10669 (maybe_wrap_with_location): New decl.
10670 (EXPR_LOCATION_WRAPPER_P): New macro.
10671 (location_wrapper_p): New inline function.
10672 (tree_strip_any_location_wrapper): New inline function.
10674 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10677 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10678 stack_realign_offset for the largest alignment of stack slot
10680 (ix86_find_max_used_stack_alignment): New function.
10681 (ix86_finalize_stack_frame_flags): Use it. Set
10682 max_used_stack_alignment if we don't realign stack.
10683 * config/i386/i386.h (machine_function): Add
10684 max_used_stack_alignment.
10686 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10688 * config/arm/arm.opt (-mbranch-cost): New option.
10689 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10692 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10695 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10696 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10698 2018-01-10 Richard Biener <rguenther@suse.de>
10701 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10702 early out so it also covers the case where we have a non-NULL
10705 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10707 PR tree-optimization/83753
10708 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10709 for non-strided grouped accesses if the number of elements is 1.
10711 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10714 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10715 * i386.h (TARGET_USE_GATHER): Define.
10716 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10718 2018-01-10 Martin Liska <mliska@suse.cz>
10721 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10722 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10724 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10725 CLEANUP_NO_PARTITIONING is not set.
10727 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10729 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10730 for vectors, as a partial revert of r254296.
10731 * rtl.h (const_vec_p): Delete.
10732 (const_vec_duplicate_p): Don't test for vector CONSTs.
10733 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10734 * expmed.c (make_tree): Likewise.
10737 * common.md (E, F): Use CONSTANT_P instead of checking for
10739 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10740 checking for CONST_VECTOR.
10742 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10744 PR middle-end/83575
10745 * predict.c (force_edge_cold): Handle in more sane way edges
10746 with no prediction.
10748 2018-01-09 Carl Love <cel@us.ibm.com>
10750 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10752 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10753 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10754 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10755 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10756 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10757 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10758 * config/rs6000/rs6000-protos.h: Add extern defition for
10759 rs6000_generate_float2_double_code.
10760 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10762 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10763 (float2_v2df): Add define_expand.
10765 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10768 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10769 op_mode in the force_to_mode call.
10771 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10773 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10774 instead of checking each element individually.
10775 (aarch64_evpc_uzp): Likewise.
10776 (aarch64_evpc_zip): Likewise.
10777 (aarch64_evpc_ext): Likewise.
10778 (aarch64_evpc_rev): Likewise.
10779 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10780 instead of checking each element individually. Return true without
10782 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10783 whether all selected elements come from the same input, instead of
10784 checking each element individually. Remove calls to gen_rtx_REG,
10785 start_sequence and end_sequence and instead assert that no rtl is
10788 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10790 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10791 order of HIGH and CONST checks.
10793 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10795 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10796 if the destination isn't an SSA_NAME.
10798 2018-01-09 Richard Biener <rguenther@suse.de>
10800 PR tree-optimization/83668
10801 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10803 (canonicalize_loop_form): ... here, renamed from ...
10804 (canonicalize_loop_closed_ssa_form): ... this and amended to
10805 swap successor edges for loop exit blocks to make us use
10806 the RPO order we need for initial schedule generation.
10808 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10810 PR tree-optimization/64811
10811 * match.pd: When optimizing comparisons with Inf, avoid
10812 introducing or losing exceptions from comparisons with NaN.
10814 2018-01-09 Martin Liska <mliska@suse.cz>
10817 * asan.c (shadow_mem_size): Add gcc_assert.
10819 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10821 Don't save registers in main().
10824 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10825 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10826 * config/avr/avr.c (avr_set_current_function): Don't error if
10827 naked, OS_task or OS_main are specified at the same time.
10828 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10830 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10832 * common/config/avr/avr-common.c (avr_option_optimization_table):
10833 Switch on -mmain-is-OS_task for optimizing compilations.
10835 2018-01-09 Richard Biener <rguenther@suse.de>
10837 PR tree-optimization/83572
10838 * graphite.c: Include cfganal.h.
10839 (graphite_transform_loops): Connect infinite loops to exit
10840 and remove fake edges at the end.
10842 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10844 * ipa-inline.c (edge_badness): Revert accidental checkin.
10846 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10849 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10850 symbols; not inline clones.
10852 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10855 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10856 hard registers. Formatting fixes.
10858 PR preprocessor/83722
10859 * gcc.c (try_generate_repro): Pass
10860 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10861 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10864 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10865 Kito Cheng <kito.cheng@gmail.com>
10867 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10868 (riscv_leaf_function_p): Delete.
10869 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10871 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10873 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10875 (do_ifelse): New function.
10876 (do_isel): New function.
10877 (do_sub3): New function.
10878 (do_add3): New function.
10879 (do_load_mask_compare): New function.
10880 (do_overlap_load_compare): New function.
10881 (expand_compare_loop): New function.
10882 (expand_block_compare): Call expand_compare_loop() when appropriate.
10883 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10884 option description.
10885 (-mblock-compare-inline-loop-limit): New option.
10887 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10890 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10891 Reverse order of second and third operands in first alternative.
10892 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10893 of first and second elements in UNSPEC_VPERMR vector.
10894 (altivec_expand_vec_perm_le): Likewise.
10896 2018-01-08 Jeff Law <law@redhat.com>
10898 PR rtl-optimizatin/81308
10899 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10900 (process_switch): If group_case_labels makes a change, then set
10902 (pass_convert_switch::execute): If a switch is converted, then
10903 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10905 PR rtl-optimization/81308
10906 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10909 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10911 PR target/83663 - Revert r255946
10912 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10913 generation for cases where splatting a value is not useful.
10914 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10915 across a vec_duplicate and a paradoxical subreg forming a vector
10916 mode to a vec_concat.
10918 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10920 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10921 -march=armv8.3-a variants.
10922 * config/arm/t-multilib: Likewise.
10923 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10925 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10927 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10929 (cceq_ior_compare_complement): Give it a name so I can use it, and
10930 change boolean_or_operator predicate to boolean_operator so it can
10931 be used to generate a crand.
10932 (eqne): New code iterator.
10933 (bd/bd_neg): New code_attrs.
10934 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10935 a single define_insn.
10936 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10937 decrement (bdnzt/bdnzf/bdzt/bdzf).
10938 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10939 with the new names of the branch decrement patterns, and added the
10940 names of the branch decrement conditional patterns.
10942 2018-01-08 Richard Biener <rguenther@suse.de>
10944 PR tree-optimization/83563
10945 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10948 2018-01-08 Richard Biener <rguenther@suse.de>
10950 PR middle-end/83713
10951 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10953 2018-01-08 Richard Biener <rguenther@suse.de>
10955 PR tree-optimization/83685
10956 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10957 references to abnormals.
10959 2018-01-08 Richard Biener <rguenther@suse.de>
10962 * dwarf2out.c (output_indirect_strings): Handle empty
10963 skeleton_debug_str_hash.
10964 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10966 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10968 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10969 (emit_store_direct): Likewise.
10970 (arc_trampoline_adjust_address): Likewise.
10971 (arc_asm_trampoline_template): New function.
10972 (arc_initialize_trampoline): Use asm_trampoline_template.
10973 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10974 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10975 * config/arc/arc.md (flush_icache): Delete pattern.
10977 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10979 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10980 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10983 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10986 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10987 by not USED_FOR_TARGET.
10988 (make_pass_resolve_sw_modes): Likewise.
10990 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10992 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10995 2018-01-08 Richard Biener <rguenther@suse.de>
10997 PR middle-end/83580
10998 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11000 2018-01-08 Richard Biener <rguenther@suse.de>
11002 PR middle-end/83517
11003 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11005 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11007 PR middle-end/81897
11008 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11009 basic blocks with a small number of successors.
11010 (convert_control_dep_chain_into_preds): Improve handling of
11012 (dump_predicates): Split apart into...
11013 (dump_pred_chain): ...here...
11014 (dump_pred_info): ...and here.
11015 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11016 (can_chain_union_be_invalidated_p): Improve check for invalidation
11018 (uninit_uses_cannot_happen): Avoid unnecessary if
11019 convert_control_dep_chain_into_preds yielded nothing.
11021 2018-01-06 Martin Sebor <msebor@redhat.com>
11023 PR tree-optimization/83640
11024 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11025 subtracting negative offset from size.
11026 (builtin_access::overlap): Adjust offset bounds of the access to fall
11027 within the size of the object if possible.
11029 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11031 PR rtl-optimization/83699
11032 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11033 extract_bit_field_as_subreg to cases in which the extracted
11034 value is also a vector.
11036 * lra-constraints.c (process_alt_operands): Test for the equivalence
11037 substitutions when detecting a possible reload cycle.
11039 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11042 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11043 by default if flag_selective_schedling{,2}. Formatting fixes.
11045 PR rtl-optimization/83682
11046 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11047 if it has non-VECTOR_MODE element mode.
11048 (vec_duplicate_p): Likewise.
11050 PR middle-end/83694
11051 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11052 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11054 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11057 * config/i386/i386-builtin.def
11058 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11059 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11060 Require also OPTION_MASK_ISA_AVX512F in addition to
11061 OPTION_MASK_ISA_GFNI.
11062 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11063 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11064 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11065 to OPTION_MASK_ISA_GFNI.
11066 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11067 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11068 OPTION_MASK_ISA_AVX512BW.
11069 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11070 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11071 addition to OPTION_MASK_ISA_GFNI.
11072 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11073 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11074 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11075 to OPTION_MASK_ISA_GFNI.
11076 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11077 a requirement for all ISAs rather than any of them with a few
11079 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11081 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11082 bitmasks to be enabled with 3 exceptions, instead of requiring any
11083 enabled ISA with lots of exceptions.
11084 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11085 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11086 Change avx512bw in isa attribute to avx512f.
11087 * config/i386/sgxintrin.h: Add license boilerplate.
11088 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11089 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11090 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11091 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11093 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11094 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11095 temporarily sse2 rather than sse if not enabled already.
11098 * config/i386/sse.md (VI248_VLBW): Rename to ...
11099 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11100 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11101 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11102 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11103 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11104 mode iterator instead of VI248_VLBW.
11106 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11108 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11109 (record_modified): Skip clobbers; add debug output.
11110 (param_change_prob): Use sreal frequencies.
11112 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11114 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11115 punt for user-aligned variables.
11117 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11119 * tree-chrec.c (chrec_contains_symbols): Return true for
11122 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11125 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11126 of (x|y) == x for BICS pattern.
11128 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11130 PR tree-optimization/83605
11131 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11132 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11135 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11137 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11138 * config/epiphany/rtems.h: New file.
11140 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11141 Uros Bizjak <ubizjak@gmail.com>
11144 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11145 QIreg_operand instead of register_operand predicate.
11146 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11147 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11148 comments instead of -fmitigate[-_]rop.
11150 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11153 * cgraphunit.c (symbol_table::compile): Switch to text_section
11154 before calling assembly_start debug hook.
11155 * run-rtl-passes.c (run_rtl_passes): Likewise.
11158 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11160 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11161 range_int_cst_p rather than !symbolic_range_p before calling
11162 extract_range_from_multiplicative_op_1.
11164 2018-01-04 Jeff Law <law@redhat.com>
11166 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11167 redundant test in assertion.
11169 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11171 * doc/rtl.texi: Document machine_mode wrapper classes.
11173 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11175 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11176 using tree_to_uhwi.
11178 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11180 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11181 the VEC_PERM_EXPR fold to fail.
11183 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11186 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11187 to switched_sections.
11189 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11192 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11193 test for d.testing.
11195 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11198 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11199 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11201 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11204 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11205 is BLKmode and bitpos not zero or mode change is needed.
11207 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11210 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11213 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11216 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11217 instead of MULT rtx. Update all corresponding splitters.
11218 (*saddl_se): Ditto.
11219 (*ssub<modesuffix>): Ditto.
11220 (*ssubl_se): Ditto.
11221 (*cmp_sadd_di): Update split patterns.
11222 (*cmp_sadd_si): Ditto.
11223 (*cmp_sadd_sidi): Ditto.
11224 (*cmp_ssub_di): Ditto.
11225 (*cmp_ssub_si): Ditto.
11226 (*cmp_ssub_sidi): Ditto.
11227 * config/alpha/predicates.md (const23_operand): New predicate.
11228 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11229 Look for ASHIFT, not MULT inner operand.
11230 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11232 2018-01-04 Martin Liska <mliska@suse.cz>
11234 PR gcov-profile/83669
11235 * gcov.c (output_intermediate_file): Add version to intermediate
11237 * doc/gcov.texi: Document new field 'version' in intermediate
11238 file format. Fix location of '-k' option of gcov command.
11240 2018-01-04 Martin Liska <mliska@suse.cz>
11243 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11245 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11247 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11249 2018-01-03 Martin Sebor <msebor@redhat.com>
11251 PR tree-optimization/83655
11252 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11253 checking calls with invalid arguments.
11255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11257 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11258 (vectorizable_mask_load_store): Delete.
11259 (vectorizable_call): Return false for masked loads and stores.
11260 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11261 instead of gimple_assign_rhs1.
11262 (vectorizable_load): Handle IFN_MASK_LOAD.
11263 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11267 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11269 (vectorizable_mask_load_store): ...here.
11270 (vectorizable_load): ...and here.
11272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11274 * tree-vect-stmts.c (vect_build_all_ones_mask)
11275 (vect_build_zero_merge_argument): New functions, split out from...
11276 (vectorizable_load): ...here.
11278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11280 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11282 (vectorizable_mask_load_store): ...here.
11283 (vectorizable_store): ...and here.
11285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11287 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11289 (vectorizable_mask_load_store): ...here.
11291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11293 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11294 (vect_model_store_cost): Take a vec_load_store_type instead of a
11296 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11297 (vect_model_store_cost): Take a vec_load_store_type instead of a
11299 (vectorizable_mask_load_store): Update accordingly.
11300 (vectorizable_store): Likewise.
11301 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11305 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11306 IFN_MASK_LOAD calls here rather than...
11307 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11310 Alan Hayward <alan.hayward@arm.com>
11311 David Sherwood <david.sherwood@arm.com>
11313 * expmed.c (extract_bit_field_1): For vector extracts,
11314 fall back to extract_bit_field_as_subreg if vec_extract
11317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11318 Alan Hayward <alan.hayward@arm.com>
11319 David Sherwood <david.sherwood@arm.com>
11321 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11322 they are variable or constant sized.
11323 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11324 slots for constant-sized data.
11326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11327 Alan Hayward <alan.hayward@arm.com>
11328 David Sherwood <david.sherwood@arm.com>
11330 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11331 handling COND_EXPRs with boolean comparisons, try to find a better
11332 basis for the mask type than the boolean itself.
11334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11336 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11337 is calculated and how it can be overridden.
11338 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11339 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11341 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11345 Alan Hayward <alan.hayward@arm.com>
11346 David Sherwood <david.sherwood@arm.com>
11348 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11349 Remove the mode argument.
11350 (aarch64_simd_valid_immediate): Remove the mode and inverse
11352 * config/aarch64/iterators.md (bitsize): New iterator.
11353 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11354 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11355 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11356 aarch64_simd_valid_immediate.
11357 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11358 (aarch64_reg_or_bic_imm): Likewise.
11359 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11360 with an insn_type enum and msl with a modifier_type enum.
11361 Replace element_width with a scalar_mode. Change the shift
11362 to unsigned int. Add constructors for scalar_float_mode and
11363 scalar_int_mode elements.
11364 (aarch64_vect_float_const_representable_p): Delete.
11365 (aarch64_can_const_movi_rtx_p)
11366 (aarch64_simd_scalar_immediate_valid_for_move)
11367 (aarch64_simd_make_constant): Update call to
11368 aarch64_simd_valid_immediate.
11369 (aarch64_advsimd_valid_immediate_hs): New function.
11370 (aarch64_advsimd_valid_immediate): Likewise.
11371 (aarch64_simd_valid_immediate): Remove mode and inverse
11372 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11373 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11374 and aarch64_float_const_representable_p on the result.
11375 (aarch64_output_simd_mov_immediate): Remove mode argument.
11376 Update call to aarch64_simd_valid_immediate and use of
11377 simd_immediate_info.
11378 (aarch64_output_scalar_simd_mov_immediate): Update call
11381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11382 Alan Hayward <alan.hayward@arm.com>
11383 David Sherwood <david.sherwood@arm.com>
11385 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11386 (mode_nunits): Likewise CONST_MODE_NUNITS.
11387 * machmode.def (ADJUST_NUNITS): Document.
11388 * genmodes.c (mode_data::need_nunits_adj): New field.
11389 (blank_mode): Update accordingly.
11390 (adj_nunits): New variable.
11391 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11393 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11394 listed in adj_nunits.
11395 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11396 listed in adj_nunits. Don't emit case statements for such modes.
11397 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11398 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11399 nothing if adj_nunits is nonnull.
11400 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11401 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11402 (emit_mode_fbit): Update use of print_maybe_const_decl.
11403 (emit_move_size): Likewise. Treat the array as non-const
11405 (emit_mode_adjustments): Handle adj_nunits.
11407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11409 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11410 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11411 (VECTOR_MODES): Use it.
11412 (make_vector_modes): Take the prefix as an argument.
11414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11415 Alan Hayward <alan.hayward@arm.com>
11416 David Sherwood <david.sherwood@arm.com>
11418 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11419 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11420 for MODE_VECTOR_BOOL.
11421 * machmode.def (VECTOR_BOOL_MODE): Document.
11422 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11423 (make_vector_bool_mode): New function.
11424 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11426 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11427 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11429 * stor-layout.c (int_mode_for_mode): Likewise.
11430 * tree.c (build_vector_type_for_mode): Likewise.
11431 * varasm.c (output_constant_pool_2): Likewise.
11432 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11433 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11434 for MODE_VECTOR_BOOL.
11435 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11436 of mode class checks.
11437 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11438 instead of a list of mode class checks.
11439 (expand_vector_scalar_condition): Likewise.
11440 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11443 Alan Hayward <alan.hayward@arm.com>
11444 David Sherwood <david.sherwood@arm.com>
11446 * machmode.h (mode_size): Change from unsigned short to
11448 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11449 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11450 or if measurement_type is not polynomial.
11451 (fixed_size_mode::includes_p): Check for constant-sized modes.
11452 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11453 return a poly_uint16 rather than an unsigned short.
11454 (emit_mode_size): Change the type of mode_size from unsigned short
11455 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11456 (emit_mode_adjustments): Cope with polynomial vector sizes.
11457 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11459 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11461 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11462 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11463 * caller-save.c (setup_save_areas): Likewise.
11464 (replace_reg_with_saved_mem): Likewise.
11465 * calls.c (emit_library_call_value_1): Likewise.
11466 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11467 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11468 (gen_lowpart_for_combine): Likewise.
11469 * convert.c (convert_to_integer_1): Likewise.
11470 * cse.c (equiv_constant, cse_insn): Likewise.
11471 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11472 (cselib_subst_to_values): Likewise.
11473 * dce.c (word_dce_process_block): Likewise.
11474 * df-problems.c (df_word_lr_mark_ref): Likewise.
11475 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11476 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11477 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11478 (rtl_for_decl_location): Likewise.
11479 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11480 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11481 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11482 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11483 (expand_expr_real_1): Likewise.
11484 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11485 (pad_below): Likewise.
11486 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11487 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11488 * ira.c (get_subreg_tracking_sizes): Likewise.
11489 * ira-build.c (ira_create_allocno_objects): Likewise.
11490 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11491 (ira_sort_regnos_for_alter_reg): Likewise.
11492 * ira-costs.c (record_operand_costs): Likewise.
11493 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11494 (resolve_simple_move): Likewise.
11495 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11496 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11497 (lra_constraints): Likewise.
11498 (CONST_POOL_OK_P): Reject variable-sized modes.
11499 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11500 (add_pseudo_to_slot, lra_spill): Likewise.
11501 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11502 * optabs-query.c (get_best_extraction_insn): Likewise.
11503 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11504 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11505 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11506 * recog.c (offsettable_address_addr_space_p): Likewise.
11507 * regcprop.c (maybe_mode_change): Likewise.
11508 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11509 * regrename.c (build_def_use): Likewise.
11510 * regstat.c (dump_reg_info): Likewise.
11511 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11512 (find_reloads, find_reloads_subreg_address): Likewise.
11513 * reload1.c (eliminate_regs_1): Likewise.
11514 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11515 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11516 (simplify_binary_operation_1, simplify_subreg): Likewise.
11517 * targhooks.c (default_function_arg_padding): Likewise.
11518 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11519 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11520 (verify_gimple_assign_ternary): Likewise.
11521 * tree-inline.c (estimate_move_cost): Likewise.
11522 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11523 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11524 (get_address_cost_ainc): Likewise.
11525 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11526 (vect_supportable_dr_alignment): Likewise.
11527 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11528 (vectorizable_reduction): Likewise.
11529 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11530 (vectorizable_operation, vectorizable_load): Likewise.
11531 * tree.c (build_same_sized_truth_vector_type): Likewise.
11532 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11533 * var-tracking.c (emit_note_insn_var_location): Likewise.
11534 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11535 (ADDR_VEC_ALIGN): Likewise.
11537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11538 Alan Hayward <alan.hayward@arm.com>
11539 David Sherwood <david.sherwood@arm.com>
11541 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11543 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11544 or if measurement_type is polynomial.
11545 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11546 * combine.c (make_extraction): Likewise.
11547 * dse.c (find_shift_sequence): Likewise.
11548 * dwarf2out.c (mem_loc_descriptor): Likewise.
11549 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11550 (extract_bit_field, extract_low_bits): Likewise.
11551 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11552 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11553 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11554 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11555 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11556 * reload.c (find_reloads): Likewise.
11557 * reload1.c (alter_reg): Likewise.
11558 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11559 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11560 * tree-if-conv.c (predicate_mem_writes): Likewise.
11561 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11562 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11563 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11564 * valtrack.c (dead_debug_insert_temp): Likewise.
11565 * varasm.c (mergeable_constant_section): Likewise.
11566 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11569 Alan Hayward <alan.hayward@arm.com>
11570 David Sherwood <david.sherwood@arm.com>
11572 * expr.c (expand_assignment): Cope with polynomial mode sizes
11573 when assigning to a CONCAT.
11575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11576 Alan Hayward <alan.hayward@arm.com>
11577 David Sherwood <david.sherwood@arm.com>
11579 * machmode.h (mode_precision): Change from unsigned short to
11581 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11583 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11584 or if measurement_type is not polynomial.
11585 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11586 in which the mode is already known to be a scalar_int_mode.
11587 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11588 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11590 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11591 for GET_MODE_PRECISION.
11592 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11593 for GET_MODE_PRECISION.
11594 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11596 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11597 (expand_field_assignment, make_extraction): Likewise.
11598 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11599 (get_last_value): Likewise.
11600 * convert.c (convert_to_integer_1): Likewise.
11601 * cse.c (cse_insn): Likewise.
11602 * expr.c (expand_expr_real_1): Likewise.
11603 * lra-constraints.c (simplify_operand_subreg): Likewise.
11604 * optabs-query.c (can_atomic_load_p): Likewise.
11605 * optabs.c (expand_atomic_load): Likewise.
11606 (expand_atomic_store): Likewise.
11607 * ree.c (combine_reaching_defs): Likewise.
11608 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11609 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11610 * tree.h (type_has_mode_precision_p): Likewise.
11611 * ubsan.c (instrument_si_overflow): Likewise.
11613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11614 Alan Hayward <alan.hayward@arm.com>
11615 David Sherwood <david.sherwood@arm.com>
11617 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11618 polynomial numbers of units.
11619 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11620 (valid_vector_subparts_p): New function.
11621 (build_vector_type): Remove temporary shim and take the number
11622 of units as a poly_uint64 rather than an int.
11623 (build_opaque_vector_type): Take the number of units as a
11624 poly_uint64 rather than an int.
11625 * tree.c (build_vector_from_ctor): Handle polynomial
11626 TYPE_VECTOR_SUBPARTS.
11627 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11628 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11629 (build_vector_from_val): If the number of units is variable,
11630 use build_vec_duplicate_cst for constant operands and
11631 VEC_DUPLICATE_EXPR otherwise.
11632 (make_vector_type): Remove temporary is_constant ().
11633 (build_vector_type, build_opaque_vector_type): Take the number of
11634 units as a poly_uint64 rather than an int.
11635 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11637 * cfgexpand.c (expand_debug_expr): Likewise.
11638 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11639 (store_constructor, expand_expr_real_1): Likewise.
11640 (const_scalar_mask_from_tree): Likewise.
11641 * fold-const-call.c (fold_const_reduction): Likewise.
11642 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11643 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11644 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11645 (fold_relational_const): Likewise.
11646 (native_interpret_vector): Likewise. Change the size from an
11647 int to an unsigned int.
11648 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11649 TYPE_VECTOR_SUBPARTS.
11650 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11651 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11652 duplicating a non-constant operand into a variable-length vector.
11653 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11654 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11655 * ipa-icf.c (sem_variable::equals): Likewise.
11656 * match.pd: Likewise.
11657 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11658 * print-tree.c (print_node): Likewise.
11659 * stor-layout.c (layout_type): Likewise.
11660 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11661 * tree-cfg.c (verify_gimple_comparison): Likewise.
11662 (verify_gimple_assign_binary): Likewise.
11663 (verify_gimple_assign_ternary): Likewise.
11664 (verify_gimple_assign_single): Likewise.
11665 * tree-pretty-print.c (dump_generic_node): Likewise.
11666 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11667 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11668 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11669 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11670 (vect_shift_permute_load_chain): Likewise.
11671 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11672 (expand_vector_condition, optimize_vector_constructor): Likewise.
11673 (lower_vec_perm, get_compute_type): Likewise.
11674 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11675 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11676 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11677 (vect_recog_mask_conversion_pattern): Likewise.
11678 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11679 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11680 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11681 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11682 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11683 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11684 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11685 (supportable_widening_operation): Likewise.
11686 (supportable_narrowing_operation): Likewise.
11687 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11689 * varasm.c (output_constant): Likewise.
11691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11692 Alan Hayward <alan.hayward@arm.com>
11693 David Sherwood <david.sherwood@arm.com>
11695 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11696 so that both the length == 3 and length != 3 cases set up their
11697 own permute vectors. Add comments explaining why we know the
11698 number of elements is constant.
11699 (vect_permute_load_chain): Likewise.
11701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11702 Alan Hayward <alan.hayward@arm.com>
11703 David Sherwood <david.sherwood@arm.com>
11705 * machmode.h (mode_nunits): Change from unsigned char to
11707 (ONLY_FIXED_SIZE_MODES): New macro.
11708 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11709 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11710 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11712 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11713 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11714 or if measurement_type is not polynomial.
11715 * genmodes.c (ZERO_COEFFS): New macro.
11716 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11718 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11719 Use ZERO_COEFFS when emitting initializers.
11720 * data-streamer.h (bp_pack_poly_value): New function.
11721 (bp_unpack_poly_value): Likewise.
11722 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11723 for GET_MODE_NUNITS.
11724 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11725 for GET_MODE_NUNITS.
11726 * tree.c (make_vector_type): Remove temporary shim and make
11727 the real function take the number of units as a poly_uint64
11728 rather than an int.
11729 (build_vector_type_for_mode): Handle polynomial nunits.
11730 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11731 * emit-rtl.c (const_vec_series_p_1): Likewise.
11732 (gen_rtx_CONST_VECTOR): Likewise.
11733 * fold-const.c (test_vec_duplicate_folding): Likewise.
11734 * genrecog.c (validate_pattern): Likewise.
11735 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11736 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11737 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11738 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11739 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11740 * rtlanal.c (subreg_get_info): Likewise.
11741 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11742 (vect_grouped_load_supported): Likewise.
11743 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11744 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11745 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11746 (simplify_const_unary_operation, simplify_binary_operation_1)
11747 (simplify_const_binary_operation, simplify_ternary_operation)
11748 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11749 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11750 instead of CONST_VECTOR_NUNITS.
11751 * varasm.c (output_constant_pool_2): Likewise.
11752 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11753 explicit-encoded elements in the XVEC for variable-length vectors.
11755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11757 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11760 Alan Hayward <alan.hayward@arm.com>
11761 David Sherwood <david.sherwood@arm.com>
11763 * coretypes.h (fixed_size_mode): Declare.
11764 (fixed_size_mode_pod): New typedef.
11765 * builtins.h (target_builtins::x_apply_args_mode)
11766 (target_builtins::x_apply_result_mode): Change type to
11767 fixed_size_mode_pod.
11768 * builtins.c (apply_args_size, apply_result_size, result_vector)
11769 (expand_builtin_apply_args_1, expand_builtin_apply)
11770 (expand_builtin_return): Update accordingly.
11772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11774 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11775 * cselib.c (cselib_hash_rtx): Likewise.
11776 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11777 CONST_VECTOR encoding.
11779 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11780 Jeff Law <law@redhat.com>
11783 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11784 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11785 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11786 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11789 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11790 explicitly probe *sp in a noreturn function if there were any callee
11791 register saves or frame pointer is needed.
11793 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11796 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11797 BLKmode for ternary, binary or unary expressions.
11800 * var-tracking.c (delete_vta_debug_insn): New inline function.
11801 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11802 insns from get_insns () to NULL instead of each bb separately.
11803 Use delete_vta_debug_insn. No longer static.
11804 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11805 delete_vta_debug_insns callers.
11806 * rtl.h (delete_vta_debug_insns): Declare.
11807 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11808 instead of variable_tracking_main.
11810 2018-01-03 Martin Sebor <msebor@redhat.com>
11812 PR tree-optimization/83603
11813 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11814 arguments past the endof the argument list in functions declared
11815 without a prototype.
11816 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11817 Avoid checking when arguments are null.
11819 2018-01-03 Martin Sebor <msebor@redhat.com>
11822 * doc/extend.texi (attribute const): Fix a typo.
11823 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11824 issuing -Wsuggest-attribute for void functions.
11826 2018-01-03 Martin Sebor <msebor@redhat.com>
11828 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11829 offset_int::from instead of wide_int::to_shwi.
11830 (maybe_diag_overlap): Remove assertion.
11831 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11832 * gimple-ssa-sprintf.c (format_directive): Same.
11833 (parse_directive): Same.
11834 (sprintf_dom_walker::compute_format_length): Same.
11835 (try_substitute_return_value): Same.
11837 2018-01-03 Jeff Law <law@redhat.com>
11839 PR middle-end/83654
11840 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11841 non-constant residual for zero at runtime and avoid probing in
11842 that case. Reorganize code for trailing problem to mirror handling
11845 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11847 PR tree-optimization/83501
11848 * tree-ssa-strlen.c (get_string_cst): New.
11849 (handle_char_store): Call get_string_cst.
11851 2018-01-03 Martin Liska <mliska@suse.cz>
11853 PR tree-optimization/83593
11854 * tree-ssa-strlen.c: Include tree-cfg.h.
11855 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11856 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11857 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11859 (strlen_dom_walker::before_dom_children): Call
11860 gimple_purge_dead_eh_edges. Dump tranformation with details
11862 (strlen_dom_walker::before_dom_children): Update call by adding
11863 new argument cleanup_eh.
11864 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11866 2018-01-03 Martin Liska <mliska@suse.cz>
11869 * cif-code.def (VARIADIC_THUNK): New enum value.
11870 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11873 2018-01-03 Jan Beulich <jbeulich@suse.com>
11875 * sse.md (mov<mode>_internal): Tighten condition for when to use
11876 vmovdqu<ssescalarsize> for TI and OI modes.
11878 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11880 Update copyright years.
11882 2018-01-03 Martin Liska <mliska@suse.cz>
11885 * ipa-visibility.c (function_and_variable_visibility): Skip
11886 functions with noipa attribure.
11888 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11890 * gcc.c (process_command): Update copyright notice dates.
11891 * gcov-dump.c (print_version): Ditto.
11892 * gcov.c (print_version): Ditto.
11893 * gcov-tool.c (print_version): Ditto.
11894 * gengtype.c (create_file): Ditto.
11895 * doc/cpp.texi: Bump @copying's copyright year.
11896 * doc/cppinternals.texi: Ditto.
11897 * doc/gcc.texi: Ditto.
11898 * doc/gccint.texi: Ditto.
11899 * doc/gcov.texi: Ditto.
11900 * doc/install.texi: Ditto.
11901 * doc/invoke.texi: Ditto.
11903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11905 * vector-builder.h (vector_builder::m_full_nelts): Change from
11906 unsigned int to poly_uint64.
11907 (vector_builder::full_nelts): Update prototype accordingly.
11908 (vector_builder::new_vector): Likewise.
11909 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11910 (vector_builder::operator ==): Likewise.
11911 (vector_builder::finalize): Likewise.
11912 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11913 Take the number of elements as a poly_uint64 rather than an
11915 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11916 from unsigned int to poly_uint64.
11917 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11918 (vec_perm_indices::new_vector): Likewise.
11919 (vec_perm_indices::length): Likewise.
11920 (vec_perm_indices::nelts_per_input): Likewise.
11921 (vec_perm_indices::input_nelts): Likewise.
11922 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11923 number of elements per input as a poly_uint64 rather than an
11924 unsigned int. Use the original encoding for variable-length
11925 vectors, rather than clamping each individual element.
11926 For the second and subsequent elements in each pattern,
11927 clamp the step and base before clamping their sum.
11928 (vec_perm_indices::series_p): Handle polynomial element counts.
11929 (vec_perm_indices::all_in_range_p): Likewise.
11930 (vec_perm_indices_to_tree): Likewise.
11931 (vec_perm_indices_to_rtx): Likewise.
11932 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11933 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11934 (tree_vector_builder::new_binary_operation): Handle polynomial
11935 element counts. Return false if we need to know the number
11936 of elements at compile time.
11937 * fold-const.c (fold_vec_perm): Punt if the number of elements
11938 isn't known at compile time.
11940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11942 * vec-perm-indices.h (vec_perm_builder): Change element type
11943 from HOST_WIDE_INT to poly_int64.
11944 (vec_perm_indices::element_type): Update accordingly.
11945 (vec_perm_indices::clamp): Handle polynomial element_types.
11946 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11947 (vec_perm_indices::all_in_range_p): Likewise.
11948 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11950 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11951 polynomial vec_perm_indices element types.
11952 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11953 * fold-const.c (fold_vec_perm): Likewise.
11954 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11955 * tree-vect-generic.c (lower_vec_perm): Likewise.
11956 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11957 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11958 element type to HOST_WIDE_INT.
11960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11961 Alan Hayward <alan.hayward@arm.com>
11962 David Sherwood <david.sherwood@arm.com>
11964 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11965 rather than an int. Use plus_constant.
11966 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11967 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11970 Alan Hayward <alan.hayward@arm.com>
11971 David Sherwood <david.sherwood@arm.com>
11973 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11974 a HOST_WIDE_INT to a poly_int64.
11976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11977 Alan Hayward <alan.hayward@arm.com>
11978 David Sherwood <david.sherwood@arm.com>
11980 * calls.c (load_register_parameters): Cope with polynomial
11981 mode sizes. Require a constant size for BLKmode parameters
11982 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11983 forces a parameter to be padded at the lsb end in order to
11984 fill a complete number of words, require the parameter size
11985 to be ordered wrt UNITS_PER_WORD.
11987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11988 Alan Hayward <alan.hayward@arm.com>
11989 David Sherwood <david.sherwood@arm.com>
11991 * reload1.c (spill_stack_slot_width): Change element type
11992 from unsigned int to poly_uint64_pod.
11993 (alter_reg): Treat mode sizes as polynomial.
11995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11996 Alan Hayward <alan.hayward@arm.com>
11997 David Sherwood <david.sherwood@arm.com>
11999 * reload.c (complex_word_subreg_p): New function.
12000 (reload_inner_reg_of_subreg, push_reload): Use it.
12002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12003 Alan Hayward <alan.hayward@arm.com>
12004 David Sherwood <david.sherwood@arm.com>
12006 * lra-constraints.c (process_alt_operands): Reject matched
12007 operands whose sizes aren't ordered.
12008 (match_reload): Refer to this check here.
12010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12011 Alan Hayward <alan.hayward@arm.com>
12012 David Sherwood <david.sherwood@arm.com>
12014 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12015 that the mode size is in the set {1, 2, 4, 8, 16}.
12017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12018 Alan Hayward <alan.hayward@arm.com>
12019 David Sherwood <david.sherwood@arm.com>
12021 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12022 Use plus_constant instead of gen_rtx_PLUS.
12024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12025 Alan Hayward <alan.hayward@arm.com>
12026 David Sherwood <david.sherwood@arm.com>
12028 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12029 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12030 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12031 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12032 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12033 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12034 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12035 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12036 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12037 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12039 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12040 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12041 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12042 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12043 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12044 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12045 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12046 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12047 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12048 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12050 * expr.c (emit_move_resolve_push): Treat the input and result
12051 of PUSH_ROUNDING as a poly_int64.
12052 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12053 (emit_push_insn): Likewise.
12054 * lra-eliminations.c (mark_not_eliminable): Likewise.
12055 * recog.c (push_operand): Likewise.
12056 * reload1.c (elimination_effects): Likewise.
12057 * rtlanal.c (nonzero_bits1): Likewise.
12058 * calls.c (store_one_arg): Likewise. Require the padding to be
12059 known at compile time.
12061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12062 Alan Hayward <alan.hayward@arm.com>
12063 David Sherwood <david.sherwood@arm.com>
12065 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12066 Use plus_constant instead of gen_rtx_PLUS.
12068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12069 Alan Hayward <alan.hayward@arm.com>
12070 David Sherwood <david.sherwood@arm.com>
12072 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12073 rather than an int.
12075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12076 Alan Hayward <alan.hayward@arm.com>
12077 David Sherwood <david.sherwood@arm.com>
12079 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12080 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12081 via stack temporaries. Treat the mode size as polynomial too.
12083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12084 Alan Hayward <alan.hayward@arm.com>
12085 David Sherwood <david.sherwood@arm.com>
12087 * expr.c (expand_expr_real_2): When handling conversions involving
12088 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12089 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12090 as a poly_uint64 too.
12092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12093 Alan Hayward <alan.hayward@arm.com>
12094 David Sherwood <david.sherwood@arm.com>
12096 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12099 Alan Hayward <alan.hayward@arm.com>
12100 David Sherwood <david.sherwood@arm.com>
12102 * combine.c (can_change_dest_mode): Handle polynomial
12103 REGMODE_NATURAL_SIZE.
12104 * expmed.c (store_bit_field_1): Likewise.
12105 * expr.c (store_constructor): Likewise.
12106 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12107 and polynomial REGMODE_NATURAL_SIZE.
12108 (gen_lowpart_common): Likewise.
12109 * reginfo.c (record_subregs_of_mode): Likewise.
12110 * rtlanal.c (read_modify_subreg_p): Likewise.
12112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12113 Alan Hayward <alan.hayward@arm.com>
12114 David Sherwood <david.sherwood@arm.com>
12116 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12117 numbers of elements.
12119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12120 Alan Hayward <alan.hayward@arm.com>
12121 David Sherwood <david.sherwood@arm.com>
12123 * match.pd: Cope with polynomial numbers of vector elements.
12125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12126 Alan Hayward <alan.hayward@arm.com>
12127 David Sherwood <david.sherwood@arm.com>
12129 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12130 in a POINTER_PLUS_EXPR.
12132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12133 Alan Hayward <alan.hayward@arm.com>
12134 David Sherwood <david.sherwood@arm.com>
12136 * omp-simd-clone.c (simd_clone_subparts): New function.
12137 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12138 (ipa_simd_modify_function_body): Likewise.
12140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12141 Alan Hayward <alan.hayward@arm.com>
12142 David Sherwood <david.sherwood@arm.com>
12144 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12145 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12146 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12147 (expand_vector_condition, vector_element): Likewise.
12148 (subparts_gt): New function.
12149 (get_compute_type): Use subparts_gt.
12150 (count_type_subparts): Delete.
12151 (expand_vector_operations_1): Use subparts_gt instead of
12152 count_type_subparts.
12154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12155 Alan Hayward <alan.hayward@arm.com>
12156 David Sherwood <david.sherwood@arm.com>
12158 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12159 (vect_compile_time_alias): ...this new function. Do the calculation
12160 on poly_ints rather than trees.
12161 (vect_prune_runtime_alias_test_list): Update call accordingly.
12163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12164 Alan Hayward <alan.hayward@arm.com>
12165 David Sherwood <david.sherwood@arm.com>
12167 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12169 (vect_schedule_slp_instance): Likewise.
12171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12172 Alan Hayward <alan.hayward@arm.com>
12173 David Sherwood <david.sherwood@arm.com>
12175 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12176 constant and extern definitions for variable-length vectors.
12177 (vect_get_constant_vectors): Note that the number of units
12178 is known to be constant.
12180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12181 Alan Hayward <alan.hayward@arm.com>
12182 David Sherwood <david.sherwood@arm.com>
12184 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12185 of units as polynomial. Choose between WIDE and NARROW based
12188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12189 Alan Hayward <alan.hayward@arm.com>
12190 David Sherwood <david.sherwood@arm.com>
12192 * tree-vect-stmts.c (simd_clone_subparts): New function.
12193 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12196 Alan Hayward <alan.hayward@arm.com>
12197 David Sherwood <david.sherwood@arm.com>
12199 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12200 vectors as polynomial. Use build_index_vector for
12201 IFN_GOMP_SIMD_LANE.
12203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12204 Alan Hayward <alan.hayward@arm.com>
12205 David Sherwood <david.sherwood@arm.com>
12207 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12208 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12209 for variable-length vectors.
12210 (vectorizable_mask_load_store): Treat the number of units as
12211 polynomial, asserting that it is constant if the condition has
12212 already been enforced.
12213 (vectorizable_store, vectorizable_load): Likewise.
12215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12216 Alan Hayward <alan.hayward@arm.com>
12217 David Sherwood <david.sherwood@arm.com>
12219 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12220 of units as polynomial. Punt if we can't tell at compile time
12221 which vector contains the final result.
12223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12224 Alan Hayward <alan.hayward@arm.com>
12225 David Sherwood <david.sherwood@arm.com>
12227 * tree-vect-loop.c (vectorizable_induction): Treat the number
12228 of units as polynomial. Punt on SLP inductions. Use an integer
12229 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12230 cast of such a series for variable-length floating-point
12233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12234 Alan Hayward <alan.hayward@arm.com>
12235 David Sherwood <david.sherwood@arm.com>
12237 * tree.h (build_index_vector): Declare.
12238 * tree.c (build_index_vector): New function.
12239 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12240 of units as polynomial, forcibly converting it to a constant if
12241 vectorizable_reduction has already enforced the condition.
12242 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12243 to create a {1,2,3,...} vector.
12244 (vectorizable_reduction): Treat the number of units as polynomial.
12245 Choose vectype_in based on the largest scalar element size rather
12246 than the smallest number of units. Enforce the restrictions
12249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12250 Alan Hayward <alan.hayward@arm.com>
12251 David Sherwood <david.sherwood@arm.com>
12253 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12254 number of units as polynomial.
12256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12257 Alan Hayward <alan.hayward@arm.com>
12258 David Sherwood <david.sherwood@arm.com>
12260 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12261 * target.def (autovectorize_vector_sizes): Return the vector sizes
12262 by pointer, using vector_sizes rather than a bitmask.
12263 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12264 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12265 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12267 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12268 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12269 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12270 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12271 * omp-general.c (omp_max_vf): Likewise.
12272 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12273 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12274 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12275 * tree-vect-slp.c (vect_slp_bb): Likewise.
12276 * doc/tm.texi: Regenerate.
12277 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12279 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12280 the vector size as a poly_uint64 rather than an unsigned int.
12281 (current_vector_size): Change from an unsigned int to a poly_uint64.
12282 (get_vectype_for_scalar_type): Update accordingly.
12283 * tree.h (build_truth_vector_type): Take the size and number of
12284 units as a poly_uint64 rather than an unsigned int.
12285 (build_vector_type): Add a temporary overload that takes
12286 the number of units as a poly_uint64 rather than an unsigned int.
12287 * tree.c (make_vector_type): Likewise.
12288 (build_truth_vector_type): Take the number of units as a poly_uint64
12289 rather than an unsigned int.
12291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12292 Alan Hayward <alan.hayward@arm.com>
12293 David Sherwood <david.sherwood@arm.com>
12295 * target.def (get_mask_mode): Take the number of units and length
12296 as poly_uint64s rather than unsigned ints.
12297 * targhooks.h (default_get_mask_mode): Update accordingly.
12298 * targhooks.c (default_get_mask_mode): Likewise.
12299 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12300 * doc/tm.texi: Regenerate.
12302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12303 Alan Hayward <alan.hayward@arm.com>
12304 David Sherwood <david.sherwood@arm.com>
12306 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12307 * omp-general.c (omp_max_vf): Likewise.
12308 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12309 (expand_omp_simd): Handle polynomial safelen.
12310 * omp-low.c (omplow_simd_context): Add a default constructor.
12311 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12312 (lower_rec_simd_input_clauses): Update accordingly.
12313 (lower_rec_input_clauses): Likewise.
12315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12316 Alan Hayward <alan.hayward@arm.com>
12317 David Sherwood <david.sherwood@arm.com>
12319 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12320 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12321 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12322 (vect_analyze_slp_cost): Likewise.
12323 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12324 (vect_model_load_cost): Likewise.
12326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12327 Alan Hayward <alan.hayward@arm.com>
12328 David Sherwood <david.sherwood@arm.com>
12330 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12331 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12332 from an unsigned int * to a poly_uint64_pod *.
12333 (calculate_unrolling_factor): New function.
12334 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12337 Alan Hayward <alan.hayward@arm.com>
12338 David Sherwood <david.sherwood@arm.com>
12340 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12341 from an unsigned int to a poly_uint64.
12342 (_loop_vec_info::slp_unrolling_factor): Likewise.
12343 (_loop_vec_info::vectorization_factor): Change from an int
12345 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12346 (vect_get_num_vectors): New function.
12347 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12348 (vect_get_num_copies): Use vect_get_num_vectors.
12349 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12350 to an unsigned int *.
12351 (vect_analyze_data_refs): Change min_vf from an int * to a
12353 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12354 than an unsigned HOST_WIDE_INT.
12355 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12356 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12357 to an unsigned int *.
12358 (vect_analyze_data_ref_dependences): Likewise.
12359 (vect_compute_data_ref_alignment): Handle polynomial vf.
12360 (vect_enhance_data_refs_alignment): Likewise.
12361 (vect_prune_runtime_alias_test_list): Likewise.
12362 (vect_shift_permute_load_chain): Likewise.
12363 (vect_supportable_dr_alignment): Likewise.
12364 (dependence_distance_ge_vf): Take the vectorization factor as a
12365 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12366 (vect_analyze_data_refs): Change min_vf from an int * to a
12368 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12369 vfm1 as a poly_uint64 rather than an int. Make the same change
12370 for the returned bound_scalar.
12371 (vect_gen_vector_loop_niters): Handle polynomial vf.
12372 (vect_do_peeling): Likewise. Update call to
12373 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12374 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12376 * tree-vect-loop.c (vect_determine_vectorization_factor)
12377 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12378 (vect_get_known_peeling_cost): Likewise.
12379 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12380 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12381 (vect_transform_loop): Likewise. Use the lowest possible VF when
12382 updating the upper bounds of the loop.
12383 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12384 rather than an int.
12385 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12386 polynomial unroll factors.
12387 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12388 (vect_make_slp_decision): Likewise.
12389 (vect_supported_load_permutation_p): Likewise, and polynomial
12391 (vect_analyze_slp_cost): Handle polynomial vf.
12392 (vect_slp_analyze_node_operations): Likewise.
12393 (vect_slp_analyze_bb_1): Likewise.
12394 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12395 than an unsigned HOST_WIDE_INT.
12396 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12397 (vectorizable_load): Handle polynomial vf.
12398 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12400 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12403 Alan Hayward <alan.hayward@arm.com>
12404 David Sherwood <david.sherwood@arm.com>
12406 * match.pd: Handle bit operations involving three constants
12407 and try to fold one pair.
12409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12411 * tree-vect-loop-manip.c: Include gimple-fold.h.
12412 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12413 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12414 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12415 Add a path that uses a step of VF instead of 1, but disable it
12417 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12418 and niters_no_overflow parameters. Update calls to
12419 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12420 Create a new SSA name if the latter choses to use a ste other
12421 than zero, and return it via niters_vector_mult_vf_var.
12422 * tree-vect-loop.c (vect_transform_loop): Update calls to
12423 vect_do_peeling, vect_gen_vector_loop_niters and
12424 slpeel_make_loop_iterate_ntimes.
12425 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12426 (vect_gen_vector_loop_niters): Update declarations after above changes.
12428 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12430 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12431 128-bit round to integer instructions.
12432 (ceil<mode>2): Likewise.
12433 (btrunc<mode>2): Likewise.
12434 (round<mode>2): Likewise.
12436 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12438 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12439 unaligned VSX load/store on P8/P9.
12440 (expand_block_clear): Allow the use of unaligned VSX
12441 load/store on P8/P9.
12443 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12445 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12447 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12448 swap associated with both a load and a store.
12450 2018-01-02 Andrew Waterman <andrew@sifive.com>
12452 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12453 * config/riscv/riscv.md (clear_cache): Use it.
12455 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12457 * web.c: Remove out-of-date comment.
12459 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12461 * expr.c (fixup_args_size_notes): Check that any existing
12462 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12463 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12464 (emit_single_push_insn): ...here.
12466 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12468 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12469 (const_vector_encoded_nelts): New function.
12470 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12471 (const_vector_int_elt, const_vector_elt): Declare.
12472 * emit-rtl.c (const_vector_int_elt_1): New function.
12473 (const_vector_elt): Likewise.
12474 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12475 of CONST_VECTOR_ELT.
12477 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12479 * expr.c: Include rtx-vector-builder.h.
12480 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12481 directly on the tree encoding.
12482 (const_vector_from_tree): Likewise.
12483 * optabs.c: Include rtx-vector-builder.h.
12484 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12485 sequence of "u" values.
12486 * vec-perm-indices.c: Include rtx-vector-builder.h.
12487 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12488 directly on the vec_perm_indices encoding.
12490 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12492 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12493 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12494 * rtx-vector-builder.h: New file.
12495 * rtx-vector-builder.c: Likewise.
12496 * rtl.h (rtx_def::u2): Add a const_vector field.
12497 (CONST_VECTOR_NPATTERNS): New macro.
12498 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12499 (CONST_VECTOR_DUPLICATE_P): Likewise.
12500 (CONST_VECTOR_STEPPED_P): Likewise.
12501 (CONST_VECTOR_ENCODED_ELT): Likewise.
12502 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12503 (unwrap_const_vec_duplicate): Likewise.
12504 (const_vec_series_p): Check for a non-duplicated vector encoding.
12505 Say that the function only returns true for integer vectors.
12506 * emit-rtl.c: Include rtx-vector-builder.h.
12507 (gen_const_vec_duplicate_1): Delete.
12508 (gen_const_vector): Call gen_const_vec_duplicate instead of
12509 gen_const_vec_duplicate_1.
12510 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12511 (gen_const_vec_duplicate): Use rtx_vector_builder.
12512 (gen_const_vec_series): Likewise.
12513 (gen_rtx_CONST_VECTOR): Likewise.
12514 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12515 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12516 Build a new vector rather than modifying a CONST_VECTOR in-place.
12517 (handle_special_swappables): Update call accordingly.
12518 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12519 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12520 Build a new vector rather than modifying a CONST_VECTOR in-place.
12521 (handle_special_swappables): Update call accordingly.
12523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12525 * simplify-rtx.c (simplify_const_binary_operation): Use
12526 CONST_VECTOR_ELT instead of XVECEXP.
12528 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12530 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12531 the selector elements to be different from the data elements
12532 if the selector is a VECTOR_CST.
12533 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12534 ssizetype for the selector.
12536 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12538 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12539 before testing each element individually.
12540 * tree-vect-generic.c (lower_vec_perm): Likewise.
12542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12544 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12545 * selftest-run-tests.c (selftest::run_tests): Call it.
12546 * vector-builder.h (vector_builder::operator ==): New function.
12547 (vector_builder::operator !=): Likewise.
12548 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12549 (vec_perm_indices::all_from_input_p): New function.
12550 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12551 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12552 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12553 instead of reading the VECTOR_CST directly. Detect whether both
12554 vector inputs are the same before constructing the vec_perm_indices,
12555 and update the number of inputs argument accordingly. Use the
12556 utility functions added above. Only construct sel2 if we need to.
12558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12560 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12561 the broadcast of the low byte.
12562 (expand_mult_highpart): Use an explicit encoding for the permutes.
12563 * optabs-query.c (can_mult_highpart_p): Likewise.
12564 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12565 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12566 (vectorizable_bswap): Likewise.
12567 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12568 explicit encoding for the power-of-2 permutes.
12569 (vect_permute_store_chain): Likewise.
12570 (vect_grouped_load_supported): Likewise.
12571 (vect_permute_load_chain): Likewise.
12573 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12575 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12576 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12577 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12578 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12579 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12580 (vect_gen_perm_mask_any): Likewise.
12582 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12584 * int-vector-builder.h: New file.
12585 * vec-perm-indices.h: Include int-vector-builder.h.
12586 (vec_perm_indices): Redefine as an int_vector_builder.
12587 (auto_vec_perm_indices): Delete.
12588 (vec_perm_builder): Redefine as a stand-alone class.
12589 (vec_perm_indices::vec_perm_indices): New function.
12590 (vec_perm_indices::clamp): Likewise.
12591 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12592 (vec_perm_indices::new_vector): New function.
12593 (vec_perm_indices::new_expanded_vector): Update for new
12594 vec_perm_indices class.
12595 (vec_perm_indices::rotate_inputs): New function.
12596 (vec_perm_indices::all_in_range_p): Operate directly on the
12597 encoded form, without computing elided elements.
12598 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12599 encoding. Update for new vec_perm_indices class.
12600 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12601 the given vec_perm_builder.
12602 (expand_vec_perm_var): Update vec_perm_builder constructor.
12603 (expand_mult_highpart): Use vec_perm_builder instead of
12604 auto_vec_perm_indices.
12605 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12606 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12607 or double series encoding as appropriate.
12608 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12609 vec_perm_indices instead of auto_vec_perm_indices.
12610 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12611 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12612 (vect_permute_store_chain): Likewise.
12613 (vect_grouped_load_supported): Likewise.
12614 (vect_permute_load_chain): Likewise.
12615 (vect_shift_permute_load_chain): Likewise.
12616 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12617 (vect_transform_slp_perm_load): Likewise.
12618 (vect_schedule_slp_instance): Likewise.
12619 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12620 (vectorizable_mask_load_store): Likewise.
12621 (vectorizable_bswap): Likewise.
12622 (vectorizable_store): Likewise.
12623 (vectorizable_load): Likewise.
12624 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12625 vec_perm_indices instead of auto_vec_perm_indices. Use
12626 tree_to_vec_perm_builder to read the vector from a tree.
12627 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12628 vec_perm_builder instead of a vec_perm_indices.
12629 (have_whole_vector_shift): Use vec_perm_builder and
12630 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12631 truncation to calc_vec_perm_mask_for_shift.
12632 (vect_create_epilog_for_reduction): Likewise.
12633 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12634 from auto_vec_perm_indices to vec_perm_indices.
12635 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12636 instead of changing individual elements.
12637 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12638 the vector in d.perm.
12639 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12640 from auto_vec_perm_indices to vec_perm_indices.
12641 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12642 instead of changing individual elements.
12643 (arm_vectorize_vec_perm_const): Use new_vector to install
12644 the vector in d.perm.
12645 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12646 Update vec_perm_builder constructor.
12647 (rs6000_expand_interleave): Likewise.
12648 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12649 (rs6000_expand_interleave): Likewise.
12651 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12653 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12654 to qimode could truncate the indices.
12655 * optabs.c (expand_vec_perm_var): Likewise.
12657 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12659 * Makefile.in (OBJS): Add vec-perm-indices.o.
12660 * vec-perm-indices.h: New file.
12661 * vec-perm-indices.c: Likewise.
12662 * target.h (vec_perm_indices): Replace with a forward class
12664 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12665 * optabs.h: Include vec-perm-indices.h.
12666 (expand_vec_perm): Delete.
12667 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12668 (expand_vec_perm_const): Declare.
12669 * target.def (vec_perm_const_ok): Replace with...
12670 (vec_perm_const): ...this new hook.
12671 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12672 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12673 * doc/tm.texi: Regenerate.
12674 * optabs.def (vec_perm_const): Delete.
12675 * doc/md.texi (vec_perm_const): Likewise.
12676 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12677 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12678 expand_vec_perm for constant permutation vectors. Assert that
12679 the mode of variable permutation vectors is the integer equivalent
12680 of the mode that is being permuted.
12681 * optabs-query.h (selector_fits_mode_p): Declare.
12682 * optabs-query.c: Include vec-perm-indices.h.
12683 (selector_fits_mode_p): New function.
12684 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12685 is defined, instead of checking whether the vec_perm_const_optab
12686 exists. Use targetm.vectorize.vec_perm_const instead of
12687 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12688 fit in the vector mode before using a variable permute.
12689 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12690 vec_perm_indices instead of an rtx.
12691 (expand_vec_perm): Replace with...
12692 (expand_vec_perm_const): ...this new function. Take the selector
12693 as a vec_perm_indices rather than an rtx. Also take the mode of
12694 the selector. Update call to shift_amt_for_vec_perm_mask.
12695 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12696 Use vec_perm_indices::new_expanded_vector to expand the original
12697 selector into bytes. Check whether the indices fit in the vector
12698 mode before using a variable permute.
12699 (expand_vec_perm_var): Make global.
12700 (expand_mult_highpart): Use expand_vec_perm_const.
12701 * fold-const.c: Includes vec-perm-indices.h.
12702 * tree-ssa-forwprop.c: Likewise.
12703 * tree-vect-data-refs.c: Likewise.
12704 * tree-vect-generic.c: Likewise.
12705 * tree-vect-loop.c: Likewise.
12706 * tree-vect-slp.c: Likewise.
12707 * tree-vect-stmts.c: Likewise.
12708 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12710 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12711 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12712 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12713 (aarch64_vectorize_vec_perm_const): ...this new function.
12714 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12715 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12716 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12717 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12718 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12719 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12720 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12722 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12723 check for NEON modes.
12724 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12725 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12726 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12727 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12729 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12730 the old VEC_PERM_CONST conditions.
12731 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12732 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12733 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12734 (ia64_vectorize_vec_perm_const_ok): Merge into...
12735 (ia64_vectorize_vec_perm_const): ...this new function.
12736 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12737 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12738 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12739 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12740 * config/mips/mips.c (mips_expand_vec_perm_const)
12741 (mips_vectorize_vec_perm_const_ok): Merge into...
12742 (mips_vectorize_vec_perm_const): ...this new function.
12743 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12744 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12745 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12746 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12747 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12748 (rs6000_expand_vec_perm_const): Delete.
12749 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12751 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12752 (altivec_expand_vec_perm_const_le): Take each operand individually.
12753 Operate on constant selectors rather than rtxes.
12754 (altivec_expand_vec_perm_const): Likewise. Update call to
12755 altivec_expand_vec_perm_const_le.
12756 (rs6000_expand_vec_perm_const): Delete.
12757 (rs6000_vectorize_vec_perm_const_ok): Delete.
12758 (rs6000_vectorize_vec_perm_const): New function.
12759 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12760 an element count and rtx array.
12761 (rs6000_expand_extract_even): Update call accordingly.
12762 (rs6000_expand_interleave): Likewise.
12763 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12764 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12765 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12766 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12767 (rs6000_expand_vec_perm_const): Delete.
12768 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12769 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12770 (altivec_expand_vec_perm_const_le): Take each operand individually.
12771 Operate on constant selectors rather than rtxes.
12772 (altivec_expand_vec_perm_const): Likewise. Update call to
12773 altivec_expand_vec_perm_const_le.
12774 (rs6000_expand_vec_perm_const): Delete.
12775 (rs6000_vectorize_vec_perm_const_ok): Delete.
12776 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12777 reference to the SPE evmerge intructions.
12778 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12779 an element count and rtx array.
12780 (rs6000_expand_extract_even): Update call accordingly.
12781 (rs6000_expand_interleave): Likewise.
12782 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12783 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12785 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12789 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12790 vector mode and that that mode matches the mode of the data
12792 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12793 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12794 directly using expand_vec_perm_1 when forcing selectors into
12796 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12798 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12800 * optabs-query.h (can_vec_perm_p): Delete.
12801 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12802 * optabs-query.c (can_vec_perm_p): Split into...
12803 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12804 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12805 particular selector is valid.
12806 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12807 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12808 (vect_grouped_load_supported): Likewise.
12809 (vect_shift_permute_load_chain): Likewise.
12810 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12811 (vect_transform_slp_perm_load): Likewise.
12812 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12813 (vectorizable_bswap): Likewise.
12814 (vect_gen_perm_mask_checked): Likewise.
12815 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12816 implementations of variable permutation vectors into account
12817 when deciding which selector to use.
12818 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12819 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12820 with a false third argument.
12821 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12822 to test whether the constant selector is valid and can_vec_perm_var_p
12823 to test whether a variable selector is valid.
12825 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12827 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12828 * optabs-query.c (can_vec_perm_p): Likewise.
12829 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12830 instead of vec_perm_indices.
12831 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12832 (vect_gen_perm_mask_checked): Likewise,
12833 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12834 (vect_gen_perm_mask_checked): Likewise,
12836 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12838 * optabs-query.h (qimode_for_vec_perm): Declare.
12839 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12840 (qimode_for_vec_perm): ...this new function.
12841 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12843 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12845 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12846 does not have a conditional at the top.
12848 2018-01-02 Richard Biener <rguenther@suse.de>
12850 * ipa-inline.c (big_speedup_p): Fix expression.
12852 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12855 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12858 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12862 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12863 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12864 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12865 cond_taken_branch_cost 3->4.
12867 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12869 PR tree-optimization/83581
12870 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12871 TODO_cleanup_cfg if any changes have been made.
12873 PR middle-end/83608
12874 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12875 convert_modes if target mode has the right side, but different mode
12878 PR middle-end/83609
12879 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12880 last argument when extracting from CONCAT. If either from_real or
12881 from_imag is NULL, use expansion through memory. If result is not
12882 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12883 the parts directly to inner mode, if even that fails, use expansion
12886 PR middle-end/83623
12887 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12888 check for bswap in mode rather than HImode and use that in expand_unop
12891 Copyright (C) 2018 Free Software Foundation, Inc.
12893 Copying and distribution of this file, with or without modification,
12894 are permitted in any medium without royalty provided the copyright
12895 notice and this notice are preserved.