1 2018-05-16 Richard Biener <rguenther@suse.de>
3 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4 (dump_stmt_cost): Declare.
5 (add_stmt_cost): Dump cost we add.
6 (add_stmt_costs): New function.
7 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
9 (vect_analyze_stmt): Adjust prototype.
10 (vectorizable_condition): Likewise.
11 (vectorizable_live_operation): Likewise.
12 (vectorizable_reduction): Likewise.
13 (vectorizable_induction): Likewise.
14 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
15 cost vector to pass to vectorizable_ and record afterwards.
16 (vect_model_reduction_cost): Take cost vector argument and adjust.
17 (vect_model_induction_cost): Likewise.
18 (vectorizable_reduction): Likewise.
19 (vectorizable_induction): Likewise.
20 (vectorizable_live_operation): Likewise.
21 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
22 SLP_TREE_NUMBER_OF_VEC_STMTS.
23 (vect_analyze_slp_cost_1): Remove.
24 (vect_analyze_slp_cost): Likewise.
25 (vect_slp_analyze_node_operations): Take visited args and
26 a target cost vector. Avoid processing already visited stmt sets.
27 (vect_slp_analyze_operations): Use a local cost vector to gather
28 costs and register those of non-discarded instances.
29 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
30 (vect_schedule_slp_instance): Remove copying of
31 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
33 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
34 adding cost. Record cost entry location.
35 (vect_prologue_cost_for_slp_op): Function to compute cost of
36 a constant or invariant generated for SLP vect in the prologue,
37 split out from vect_analyze_slp_cost_1.
38 (vect_model_simple_cost): Make static. Adjust for SLP costing.
39 (vect_model_promotion_demotion_cost): Likewise.
40 (vect_model_store_cost): Likewise, make static.
41 (vect_model_load_cost): Likewise.
42 (vectorizable_bswap): Add cost vector arg and adjust.
43 (vectorizable_call): Likewise.
44 (vectorizable_simd_clone_call): Likewise.
45 (vectorizable_conversion): Likewise.
46 (vectorizable_assignment): Likewise.
47 (vectorizable_shift): Likewise.
48 (vectorizable_operation): Likewise.
49 (vectorizable_store): Likewise.
50 (vectorizable_load): Likewise.
51 (vectorizable_condition): Likewise.
52 (vectorizable_comparison): Likewise.
53 (can_vectorize_live_stmts): Likewise.
54 (vect_analyze_stmt): Likewise.
55 (vect_transform_stmt): Adjust calls to vectorizable_*.
56 * tree-vectorizer.c: Include gimple-pretty-print.h.
57 (dump_stmt_cost): New function.
59 2018-05-16 Richard Biener <rguenther@suse.de>
61 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
62 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
63 * tree-ssa-dse.c: Include tree-ssa-loop.h.
64 (check_name): New callback.
65 (dse_classify_store): Track cycles via a visited bitmap of PHI
66 defs and simplify handling of in-loop and across loop dead stores
67 and properly fail for loop-variant refs. Handle byte-tracking with
68 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
71 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
73 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
74 (vect_get_mask_type_for_stmt): Likewise.
75 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
77 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
78 to determine the statement's vector type and the vector type that
79 should be used for calculating nunits. Deal with cases in which
80 the type has to be deferred.
81 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
82 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
83 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
84 (vect_determine_vf_for_stmt): New functions, split out from...
85 (vect_determine_vectorization_factor): ...here.
86 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
87 (vect_get_mask_type_for_stmt): New functions, split out from
88 vect_determine_vectorization_factor.
90 2018-05-16 Richard Biener <rguenther@suse.de>
92 * tree-cfg.c (verify_gimple_assign_ternary): Properly
93 verify the [VEC_]COND_EXPR embedded comparison.
95 2018-05-15 Martin Sebor <msebor@redhat.com>
97 PR tree-optimization/85753
98 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
99 RECORD_TYPE in addition to ARRAY_TYPE.
101 2018-05-15 Martin Sebor <msebor@redhat.com>
104 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
106 2018-05-15 Richard Biener <rguenther@suse.de>
108 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
109 add by_clobber_p one. Change algorithm to collect all defs
110 representing uses we need to walk and try reducing them to
111 a single one before failing.
112 (dse_dom_walker::dse_optimize_stmt): Adjust.
114 2018-05-13 Mark Wielaard <mark@klomp.org>
116 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
117 (size_of_loc_descr): Likewise.
118 (output_loc_operands): Likewise.
119 (output_loc_operands_raw): Likewise.
120 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
121 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
122 (hash_loc_operands): Likewise.
123 (compare_loc_operands): Likewise.
125 2018-05-14 Mark Wielaard <mark@klomp.org>
127 * dwarf2out.c (count_index_addrs): New function.
128 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
130 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
132 PR tree-optimization/83648
133 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
134 return value as malloc candidate.
136 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
139 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
140 as true in call to suggest_attribute.
142 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
144 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
147 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
149 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
151 (aarch64_crypto_aesd_fused): Likewise.
153 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
155 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
156 (movsi_aarch64): Likewise.
157 (load_pairsi): Likewise.
158 (load_pairdi): Likewise.
159 (store_pairsi): Likewise.
160 (store_pairdi): Likewise.
161 (load_pairsf): Likewise.
162 (load_pairdf): Likewise.
163 (store_pairsf): Likewise.
164 (store_pairdf): Likewise.
165 (zero_extend): Likewise.
166 (trunc): Swap alternatives.
167 (fcvt_target): Add '?' to prefer w over r.
169 2018-05-14 Jakub Jelinek <jakub@redhat.com>
172 * config/i386/i386.md: Disallow non-commutative arithmetics in
173 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
174 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
175 in the peephole2 before it.
177 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
179 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
180 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
181 (ix86_handle_option): Handle -mcldemote.
182 * config.gcc: New header.
183 * config/i386/cldemoteintrin.h: New file.
184 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
185 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
187 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
188 OPTION_MASK_ISA_CLDEMOTE.
189 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
190 (ix86_valid_target_attribute_inner_p): Ditto.
191 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
192 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
193 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
194 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
195 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
197 * config/i386/i386.opt: Add -mcldemote.
198 * config/i386/x86intrin.h: New header.
199 * doc/invoke.texi: Add -mcldemote.
201 2018-05-14 Richard Biener <rguenther@suse.de>
203 * doc/match-and-simplify.texi: Adjust :s documentation.
205 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
207 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
208 intended memcpy size.
209 (REORDER_45): Likewise.
211 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
214 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
215 * vec.c (qsort_chk): Use gcc_qsort.
216 * Makefile.in (OBJS-libcommon): Add sort.o.
217 (build/sort.o): New target. Use it...
218 (BUILD_RTL): ... here, and...
219 (build/gencfn-macros): ... here, and...
220 (build/genmatch): ... here.
222 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
223 Chung-Ju Wu <jasonwucj@gmail.com>
225 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
226 * config/nds32/nds32-graywolf.md: New file.
227 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
228 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
230 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
231 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
232 * config/nds32/nds32.md (pipeline_model): Add graywolf.
233 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
234 * config/nds32/pipelines.md: Include n15 settings.
236 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
237 Chung-Ju Wu <jasonwucj@gmail.com>
239 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
240 * config/nds32/nds32-n13.md: New file.
241 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
242 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
244 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
245 * config/nds32/nds32.md (pipeline_model): Add n13.
246 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
247 * config/nds32/pipelines.md: Include n13 settings.
249 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
250 Chung-Ju Wu <jasonwucj@gmail.com>
252 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
253 * config/nds32/nds32-n10.md: New file.
254 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
255 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
257 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
258 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
259 * config/nds32/nds32.md (pipeline_model): Add n10.
260 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
261 * config/nds32/pipelines.md: Include n10 settings.
263 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
264 Kito Cheng <kito.cheng@gmail.com>
265 Chung-Ju Wu <jasonwucj@gmail.com>
267 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
268 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
269 Add enum values for DSP extension instructions.
270 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
272 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
273 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
275 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
276 * config/nds32/nds32-dspext.md: New file for DSP implementation.
277 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
278 * config/nds32/nds32-intrinsic.md: Likewise.
279 * config/nds32/nds32_intrinsic.h: Likewise.
280 * config/nds32/nds32-md-auxiliary.c: Likewise.
281 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
282 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
283 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
284 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
285 * config/nds32/nds32-protos.h: New declarations for DSP extension.
286 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
287 TYPE_DMAC in switch statement.
288 * config/nds32/nds32.c: New checking and implementation for DSP
289 extension instructions.
290 * config/nds32/nds32.h: Likewise.
291 * config/nds32/nds32.md: Likewise.
292 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
293 * config/nds32/predicates.md: Implement new predicates for DSP
296 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
298 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
299 Reformat alternatives and attributes so it is easier to identify
300 which constraints/attributes go with which instruction.
301 (mov<mode>_hardfloat32, FMOVE64): Likewise.
302 (mov<mode>_softfloat32, FMOVE64): Likewise.
303 (mov<mode>_hardfloat64, FMOVE64): Likewise.
304 (mov<mode>_softfloat64, FMOVE64): Likewise.
306 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
308 * doc/extend.texi (PowerPC Built-in Functions): Rename this
310 (Basic PowerPC Built-in Functions): The new name of the
311 subsection previously known as "PowerPC Built-in Functions".
312 (Basic PowerPC Built-in Functions Available on all Configurations):
314 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
315 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
316 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
317 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
319 2018-05-11 Martin Jambor <mjambor@suse.cz>
322 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
325 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
328 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
330 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
332 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
333 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
334 (ix86_handle_option): Handle -mwaitpkg.
335 * config.gcc: New header.
336 * config/i386/cpuid.h (bit_WAITPKG): New bit.
337 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
338 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
340 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
341 OPTION_MASK_ISA_WAITPKG.
342 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
343 (ix86_option_override_internal): Add PTA_WAITPKG.
344 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
345 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
347 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
348 __builtin_ia32_umwait and __builtin_ia32_tpause.
349 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
350 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
351 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
352 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
353 UNSPECV_TPAUSE): New.
354 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
355 * config/i386/i386.opt: Add -mwaitpkg.
356 * config/i386/waitpkgintrin.h: New file.
357 * config/i386/x86intrin.h: New header.
358 * doc/invoke.texi: Add -mwaitpkg.
360 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
363 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
365 (cortex-m0): Use armv6s-m isa.
366 (cortex-m0plus): Likewise.
367 (cortex-m1): Likewise.
368 (cortex-m0.small-multiply): Likewise.
369 (cortex-m0plus.small-multiply): Likewise.
370 (cortex-m1.small-multiply): Likewise.
372 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
373 Jakub Jelinek <jakub@redhat.com>
375 PR tree-optimization/85692
376 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
377 source permute as well.
379 2018-05-11 Martin Liska <mliska@suse.cz>
382 * doc/extend.texi: Document LLVM style format for no_sanitize
385 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
387 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
388 mode_supports_vsx_dform_quad to mode_supports_dq_form.
389 (mode_supports_vsx_dform_quad): Likewise.
390 (mode_supports_vmx_dform): Move these functions to be next to the
391 other mode_supports functions.
392 (mode_supports_dq_form): Likewise.
393 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
394 mode_supports_dq_form.
395 (reg_offset_addressing_ok_p): Likewise.
396 (offsettable_ok_by_alignment): Likewise.
397 (rs6000_legitimate_offset_address_p): Likewise.
398 (legitimate_lo_sum_address_p): Likewise.
399 (rs6000_legitimize_address): Likewise.
400 (rs6000_legitimize_reload_address): Likewise.
401 (rs6000_secondary_reload_inner): Likewise.
402 (rs6000_preferred_reload_class): Likewise.
403 (rs6000_output_move_128bit): Likewise.
405 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
407 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
408 Generate SImode target register for null target.
409 <case IX86_BUILTIN_XGETBV>: Ditto.
410 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
411 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
413 2018-05-10 Carl Love <cel@us.ibm.com>
415 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
416 dcbtt and dcbtstt if operands[2] is 0.
418 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
421 * config/i386/sse.md (usadv64qi): New expander.
423 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
425 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
426 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
427 -maltivec=be support.
428 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
429 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
430 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
431 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
432 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
433 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
434 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
435 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
436 altivec_vsumsws): Adjust.
437 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
438 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
439 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
441 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
442 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
443 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
444 (altivec_lve<VI_char>x): Delete expand.
445 (*altivec_lve<VI_char>x_internal): Rename to...
446 (altivec_lve<VI_char>x): ... this.
447 (altivec_lvxl_<mode>): Delete expand.
448 (*altivec_lvxl_<mode>_internal): Rename to ...
449 (altivec_lvxl_<mode>): ... this.
450 (altivec_stvxl_<mode>): Delete expand.
451 (*altivec_stvxl_<mode>_internal): Rename to ...
452 (altivec_stvxl_<mode>): ... this.
453 (altivec_stve<VI_char>x): Delete expand.
454 (*altivec_stve<VI_char>x_internal): Rename to ...
455 (altivec_stve<VI_char>x): ... this.
456 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
457 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
458 reduc_plus_scal_<mode>): Adjust.
459 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
461 (rs6000_cpu_cpp_builtins): Adjust.
462 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
463 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
464 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
465 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
466 -maltivec=be support.
467 (rs6000_split_vec_extract_var): Adjust.
468 (rs6000_split_v4si_init): Adjust.
469 (swap_selector_for_mode): Delete.
470 (altivec_expand_lvx_be, altivec_expand_stvx_be,
471 altivec_expand_stvex_be): Delete.
472 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
473 -maltivec=be support.
474 (rs6000_gimple_fold_builtin): Ditto.
475 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
477 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
478 (TARGET_DIRECT_MOVE_64BIT): Adjust.
479 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
480 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
481 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
482 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
483 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
484 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
485 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
486 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
487 anonymous split): Adjust.
488 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
489 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
491 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
493 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
494 when --with-gxx-include-dir is also specified.
495 * configure: Regenerate.
497 2018-05-09 Jim Wilson <jimw@sifive.com>
500 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
501 * config/riscv/t-withmultilib: New.
502 * config/riscv/withmultilib.h: New.
503 * doc/install.texi: Document RISC-V --with-multilib-list support.
505 2018-05-09 Richard Biener <rguenther@suse.de>
507 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
509 (vect_bb_vectorization_profitable_p): Adjust. Compute
510 actual scalar cost using the cost vector and the add_stmt_cost
513 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
515 PR rtl-optimization/85645
516 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
517 in the REG_CFA_REGISTER note for LR, don't leave it empty.
519 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
521 PR rtl-optimization/85645
522 * shrink-wrap.c (spread_components): Return a boolean saying if
523 anything was changed.
524 (try_shrink_wrapping_separate): Iterate spread_components until
525 nothing changes anymore.
527 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
529 PR rtl-optimization/85645
530 * regrename.c (build_def_use): Also kill the chains that include the
531 destination of a REG_CFA_REGISTER note.
533 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
535 PR rtl-optimization/85645
536 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
537 insn that has a REG_CFA_REGISTER note.
539 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
541 * cfgexpand.c (expand_clobber): New function.
542 (expand_gimple_stmt_1): Use it.
543 * tree-vect-stmts.c (vect_clobber_variable): New function,
545 (vectorizable_simd_clone_call): ...here.
546 (vectorizable_store): Emit a clobber either side of an
547 IFN_STORE_LANES sequence.
548 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
550 2018-05-09 Tom de Vries <tom@codesourcery.com>
553 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
554 (define_insn "trap_if_false"): Add exit after trap.
556 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
558 PR rtl-optimization/85638
559 * bb-reorder.c: Include common/common-target.h.
560 (create_forwarder_block): New function extracted from...
561 (fix_up_crossing_landing_pad): ...here. Rename into...
562 (dw2_fix_up_crossing_landing_pad): ...this.
563 (sjlj_fix_up_crossing_landing_pad): New function.
564 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
565 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
566 from both partitions and exit the loop after one iteration.
568 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
571 * doc/extend.texi (PowerPC Built-in Functions): Rename this
573 (Basic PowerPC Built-in Functions): The new name of the
574 subsection previously known as "PowerPC Built-in Functions".
575 (Basic PowerPC Built-in Functions Available on all Configurations):
577 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
579 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
581 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
583 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
586 2018-05-08 Jim Wilson <jimw@sifive.com>
588 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
589 (LD_EMUL_SUFFIX): New.
592 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
594 * doc/extend.texi (PowerPC Built-in Functions): Rename this
596 (Basic PowerPC Built-in Functions): The new name of the
597 subsection previously known as "PowerPC Built-in Functions".
598 (Basic PowerPC Built-in Functions Available on all Configurations):
600 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
602 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
604 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
606 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
609 2018-05-08 Jakub Jelinek <jakub@redhat.com>
612 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
613 after cmpelim optimization.
615 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
617 * config.gcc: Support "goldmont".
618 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
619 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
621 * config/i386/i386.c (m_GOLDMONT): Define.
622 (processor_target_table): Add "goldmont".
623 (PTA_GOLDMONT): Define.
624 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
625 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
626 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
627 (fold_builtin_cpu): Add "goldmont".
628 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
629 (ix86_option_override_internal): Add "goldmont".
630 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
631 (processor_type): Add PROCESSOR_GOLDMONT.
632 * config/i386/i386.md: Add CPU "glm".
633 * config/i386/glm.md: New file.
634 * config/i386/x86-tune.def: Add m_GOLDMONT.
635 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
637 2018-05-08 Jakub Jelinek <jakub@redhat.com>
640 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
642 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
643 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
644 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
647 * config/i386/i386.c (ix86_fold_builtin): Handle
648 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
651 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
652 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
654 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
657 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
658 (check_arch): Likewise.
659 (check_fpu): Return the result rather than printing it.
660 (end arch): Fix operator precedence.
662 (END): Print the result from check_fpu.
664 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
665 Alan Hayward <alan.hayward@arm.com>
666 David Sherwood <david.sherwood@arm.com>
668 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
669 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
670 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
671 (*fcmuo<mode>_and): New patterns.
673 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
675 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
676 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
677 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
678 (cmp_op, sve_imm_con): New code attributes.
679 (SVE_COND_INT_CMP, imm_con): Delete.
680 (cmp_op): Remove above unspecs from int attribute.
681 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
683 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
684 comparison-specific unspecs.
685 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
686 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
687 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
688 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
689 (*vec_fcm<cmp_op><mode>): Rename to...
690 (*fcm<cmp_op><mode>): ...this and adjust likewise.
691 (*vec_fcmuo<mode>): Rename to...
692 (*fcmuo<mode>): ...this and adjust likewise.
693 (*pred_fcm<cmp_op><mode>): New pattern.
694 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
695 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
697 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
699 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
700 (aarch64_emit_sve_predicated_cond): New function.
701 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
702 (aarch64_emit_unspec_cond_or): Replace with...
703 (aarch64_emit_sve_or_conds): ...this new function. Use
704 aarch64_emit_sve_ptrue_op for the individual comparisons and
705 aarch64_emit_binop to OR them together.
706 (aarch64_emit_inverted_unspec_cond): Replace with...
707 (aarch64_emit_sve_inverted_cond): ...this new function. Use
708 aarch64_emit_sve_ptrue_op for the comparison and
709 aarch64_emit_unop to invert the result.
710 (aarch64_expand_sve_vec_cmp_float): Update after the above
711 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
713 2018-05-07 Nathan Sidwell <nathan@acm.org>
715 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
716 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
717 (Backwards Compatibility): Likewise.
719 2018-05-07 Luis Machado <luis.machado@linaro.org>
723 2018-05-07 Luis Machado <luis.machado@linaro.org>
725 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
726 <prefetch_dynamic_strides>: New const bool field.
727 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
728 prefetch_dynamic_strides.
729 (exynosm1_prefetch_tune): Likewise.
730 (thunderxt88_prefetch_tune): Likewise.
731 (thunderx_prefetch_tune): Likewise.
732 (thunderx2t99_prefetch_tune): Likewise.
733 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
734 (aarch64_override_options_internal): Update to set
735 PARAM_PREFETCH_DYNAMIC_STRIDES.
736 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
737 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
738 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
739 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
740 prefetch-dynamic-strides setting.
742 2018-05-07 Luis Machado <luis.machado@linaro.org>
744 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
745 <minimum_stride>: New const int field.
746 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
747 minimum_stride field.
748 (exynosm1_prefetch_tune): Likewise.
749 (thunderxt88_prefetch_tune): Likewise.
750 (thunderx_prefetch_tune): Likewise.
751 (thunderx2t99_prefetch_tune): Likewise.
752 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
753 (aarch64_override_options_internal): Update to set
754 PARAM_PREFETCH_MINIMUM_STRIDE.
755 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
756 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
757 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
758 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
759 stride is constant and is below the minimum stride threshold.
761 2018-05-07 Luis Machado <luis.machado@linaro.org>
763 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
766 2018-05-07 Luis Machado <luis.machado@linaro.org>
768 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
769 <prefetch_dynamic_strides>: New const bool field.
770 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
771 prefetch_dynamic_strides.
772 (exynosm1_prefetch_tune): Likewise.
773 (thunderxt88_prefetch_tune): Likewise.
774 (thunderx_prefetch_tune): Likewise.
775 (thunderx2t99_prefetch_tune): Likewise.
776 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
777 (aarch64_override_options_internal): Update to set
778 PARAM_PREFETCH_DYNAMIC_STRIDES.
779 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
780 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
781 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
782 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
783 prefetch-dynamic-strides setting.
785 2018-05-07 Luis Machado <luis.machado@linaro.org>
787 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
788 <minimum_stride>: New const int field.
789 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
790 minimum_stride field.
791 (exynosm1_prefetch_tune): Likewise.
792 (thunderxt88_prefetch_tune): Likewise.
793 (thunderx_prefetch_tune): Likewise.
794 (thunderx2t99_prefetch_tune): Likewise.
795 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
796 (aarch64_override_options_internal): Update to set
797 PARAM_PREFETCH_MINIMUM_STRIDE.
798 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
799 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
800 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
801 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
802 stride is constant and is below the minimum stride threshold.
804 2018-05-06 Jakub Jelinek <jakub@redhat.com>
807 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
808 the type is addressable. Don't force op into register if it has
811 2018-05-05 Roland McGrath <mcgrathr@google.com>
814 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
815 any section for which we don't know a specific type it should have,
816 regardless of name. Previously this was done only for the exact
817 names ".init_array", ".fini_array", and ".preinit_array".
818 (default_elf_asm_named_section): Add comment about
819 relationship with default_section_type_flags and SECTION_NOTYPE.
820 (get_section): Don't consider it a type conflict if one side has
821 SECTION_NOTYPE and the other doesn't, as long as neither has the
822 SECTION_BSS et al used in the default_section_type_flags logic.
824 2018-05-05 Tom de Vries <tom@codesourcery.com>
827 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
828 (workaround_barsyncs): New function.
829 (nvptx_reorg): Use workaround_barsyncs.
830 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
831 (define_expand "nvptx_membar_cta"): New define_expand.
832 (define_insn "*nvptx_membar_cta"): New insn.
834 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
836 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
837 To improve optimization opportunities.
838 * builtin-types.def: The new needed builtin types for the above.
840 2018-05-04 Richard Biener <rguenther@suse.de>
842 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
843 * gimple-ssa-store-merging.c
844 (imm_store_chain_info::output_merged_store): Remove redundant create,
845 release split_store vector contents on failure.
846 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
847 scalar stmt vector on cache hit.
849 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
851 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
853 * config.gcc (powerpc-xilinx-eabi*): Remove.
854 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
856 (fusion_addis_mem_combo_load): Ditto.
857 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
859 (rs6000_cpu_cpp_builtins): Ditto.
860 * config/rs6000/rs6000-linux.c
861 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
862 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
863 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
865 (rs6000_setup_reg_addr_masks): Ditto.
866 (rs6000_init_hard_regno_mode_ok): Ditto.
867 (rs6000_option_override_internal): Ditto.
868 (legitimate_lo_sum_address_p): Ditto.
869 (rs6000_legitimize_address): Ditto.
870 (rs6000_legitimize_reload_address): Ditto.
871 (rs6000_legitimate_address_p): Ditto.
872 (abi_v4_pass_in_fpr): Ditto.
873 (setup_incoming_varargs): Ditto.
874 (rs6000_gimplify_va_arg): Ditto.
875 (rs6000_split_multireg_move): Ditto.
876 (rs6000_savres_strategy): Ditto.
877 (rs6000_emit_prologue_components): Ditto.
878 (rs6000_emit_epilogue_components): Ditto.
879 (rs6000_emit_prologue): Ditto.
880 (rs6000_emit_epilogue): Ditto.
881 (rs6000_elf_file_end): Ditto.
882 (rs6000_function_value): Ditto.
883 (rs6000_libcall_value): Ditto.
884 * config/rs6000/rs6000.h: Ditto.
885 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
886 (TARGET_MINMAX): ... this. New.
887 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
888 * config/rs6000/rs6000.md: Remove Xilinx FP support.
889 (*movsi_internal1_single): Delete.
890 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
891 mfpu=, mxilinx-fpu): Delete.
892 * config/rs6000/singlefp.h: Delete.
893 * config/rs6000/sysv4.h: Remove Xilinx FP support.
894 * config/rs6000/t-rs6000: Ditto.
895 * config/rs6000/t-xilinx: Delete.
896 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
897 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
898 (VStype_simple): Delete.
899 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
900 * config/rs6000/xfpu.h: Delete.
901 * config/rs6000/xfpu.md: Delete.
902 * config/rs6000/xilinx.h: Delete.
903 * config/rs6000/xilinx.opt: Delete.
904 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
905 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
907 2018-05-04 Tom de Vries <tom@codesourcery.com>
910 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
913 2018-05-04 Richard Biener <rguenther@suse.de>
916 * tree-complex.c (update_complex_assignment): We are always in SSA form.
917 (expand_complex_div_wide): Likewise.
918 (expand_complex_operations_1): Likewise.
919 (expand_complex_libcall): Preserve EH info of the original stmt.
920 (tree_lower_complex): Handle removed blocks.
921 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
922 on complex multiplication and division libcall builtins.
924 2018-05-04 Richard Biener <rguenther@suse.de>
927 * fold-const.c (negate_expr_p): Restrict negation of operand
928 zero of a division to when we know that can happen without
930 (fold_negate_expr_1): Likewise.
932 2018-05-04 Jakub Jelinek <jakub@redhat.com>
935 * real.h (real_nextafter): Declare.
936 * real.c (real_nextafter): New function.
937 * fold-const-call.c (fold_const_nextafter): New function.
938 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
940 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
941 even when arg1_mode is different from arg0_mode.
943 2018-05-03 Nathan Sidwell <nathan@acm.org>
945 * doc/extend.texi (Deprecated Features): Remove
947 (Backwards Compatibility): Likewise.
948 * doc/invoke.texi (C++ Language Options): Likewise.
949 (C++ Dialect Options): Likewise.
951 2018-05-03 Jakub Jelinek <jakub@redhat.com>
954 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
955 _mm512_mask_mullox_epi64): New intrinsics.
957 2018-05-03 Tom de Vries <tom@codesourcery.com>
960 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
961 dump files): Add offload-tree.
963 2018-05-03 Richard Biener <rguenther@suse.de>
965 PR tree-optimization/85615
966 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
967 to loops not nested in BBs loop father to avoid creating multi-entry
970 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
972 PR tree-optimization/70291
973 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
974 arguments. Change return type to tree. Emit libcall as a new
975 statement rather than replacing existing one when inplace_p is true.
976 (expand_complex_multiplication_components): New function.
977 (expand_complex_multiplication): Expand floating-point complex
978 multiplication using the above.
979 (expand_complex_division): Rename inner_type parameter to type.
980 Update expand_complex_libcall call-site.
981 (expand_complex_operations_1): Update expand_complex_multiplication
982 and expand_complex_division call-sites.
984 2018-05-02 Jakub Jelinek <jakub@redhat.com>
987 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
988 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
989 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
990 the highest significant bit of the shift count mask is clear. In
991 check whether and[sq]i3 is needed verify that all significant bits
992 of the shift count other than the highest are set.
994 2018-05-02 Tom de Vries <tom@codesourcery.com>
997 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
998 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
999 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1000 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1001 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1002 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1003 __builtin_goacc_parlevel_size.
1005 2018-05-02 Richard Biener <rguenther@suse.de>
1007 PR tree-optimization/85597
1008 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1009 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1012 2018-05-02 Tom de Vries <tom@codesourcery.com>
1015 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1016 dump files): Add ltrans-tree.
1018 2018-05-02 Tom de Vries <tom@codesourcery.com>
1021 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1022 dump files): Add wpa-ipa.
1024 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1026 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1027 powerpc*-*-linux*paired* target.
1028 * config/rs6000/750cl.h: Delete.
1029 * config/rs6000/paired.h: Delete.
1030 * config/rs6000/paired.md: Delete.
1031 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1033 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1034 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1035 comment. Remove paired float support.
1036 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1037 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1039 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1040 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1042 * config/rs6000/rs6000.c: Remove paired float support.
1043 (paired_expand_vector_init, paired_expand_vector_move,
1044 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1045 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1046 paired_expand_builtin, paired_expand_predicate_builtin,
1047 paired_init_builtins): Delete.
1048 * config/rs6000/rs6000.h: Remove paired float support.
1049 * config/rs6000/rs6000.md: Remove paired float support.
1050 (move_from_CR_ov_bit): Delete.
1051 * config/rs6000/rs6000.opt (mpaired): Delete.
1052 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1053 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1055 2018-05-02 Richard Biener <rguenther@suse.de>
1058 * gimplify.c (gimplify_save_expr): When in SSA form allow
1059 SAVE_EXPRs to compute to SSA vars.
1061 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1064 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1065 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1066 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1067 clobber operands[2], instead use a new pseudo. Formatting fixes.
1069 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1071 PR tree-optimization/85586
1072 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1073 exit early for statements in the same group if the accesses are
1076 2018-05-02 Tom de Vries <tom@codesourcery.com>
1079 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1082 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1084 PR tree-optimization/85143
1085 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1087 2018-05-01 Tom de Vries <tom@codesourcery.com>
1090 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1091 not found" error message.
1093 2018-05-01 Tom de Vries <tom@codesourcery.com>
1096 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1097 * vec.c (test_ordered_remove_if): New function.
1098 (vec_c_tests): Call test_ordered_remove_if.
1099 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1100 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1101 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1102 VEC_ORDERED_REMOVE_IF.
1104 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1106 PR tree-optimization/82665
1107 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1108 pointer subtraction where arguments come from a memchr call.
1110 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1112 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1113 --push-state --as-needed and --pop-state instead of --as-needed and
1114 --no-as-needed if ld supports it.
1115 * configure: Regenerated.
1118 * doc/install.texi2html: Replace _002d with - and _002a with * in
1119 generated html files using sed.
1121 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1124 * gcc-rich-location.c (blank_line_before_p): New function.
1125 (use_new_line): New function.
1126 (gcc_rich_location::add_fixit_insert_formatted): New function.
1127 * gcc-rich-location.h
1128 (gcc_rich_location::add_fixit_insert_formatted): New function.
1130 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1132 * selftest.c (assert_streq): Rename "expected" and "actual" to
1133 "val1" and "val2". Extend NULL-handling to cover both inputs
1134 symmetrically, while still requiring both to be non-NULL for a pass.
1135 * selftest.h (assert_streq): Rename "expected" and "actual" to
1137 (ASSERT_EQ): Likewise.
1138 (ASSERT_EQ_AT): Likewise.
1139 (ASSERT_KNOWN_EQ): Likewise.
1140 (ASSERT_KNOWN_EQ_AT): Likewise.
1141 (ASSERT_NE): Likewise.
1142 (ASSERT_MAYBE_NE): Likewise.
1143 (ASSERT_MAYBE_NE_AT): Likewise.
1144 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1145 the assertion to pass.
1146 (ASSERT_STREQ_AT): Likewise.
1148 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1150 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1151 interaction with -pie.
1153 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1155 * selftest.h: Fix alphabetization of per-source-file selftest
1158 2018-04-30 Jason Merrill <jason@redhat.com>
1160 PR c++/61982 - dead stores to destroyed objects.
1161 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1164 2018-04-30 Jason Merrill <jason@redhat.com>
1166 * tree.c (build_clobber): New.
1167 * tree.h: Declare it.
1168 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1170 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1172 * diagnostic-show-locus.c (layout::layout): Update for
1173 location_get_source_line returning a char_span.
1174 (struct char_span): Move to input.h.
1175 (struct correction): Update for fields in char_span becoming
1177 (struct source_line): Update for location_get_source_line
1178 returning a char_span.
1179 (layout::print_line): Likewise.
1180 * edit-context.c (edited_file::print_content): Likewise.
1181 (edited_file::print_diff_hunk): Likewise.
1182 (edited_file::print_run_of_changed_lines): Likewise.
1183 (edited_file::get_num_lines): Likewise.
1184 (edited_line::edited_line): Likewise.
1185 * final.c (asm_show_source): Likewise.
1186 * input.c (location_get_source_line): Convert return type
1187 from const char * to char_span, losing the final "line_len"
1189 (dump_location_info): Update for the above.
1190 (get_substring_ranges_for_loc): Likewise. Use a char_span
1191 when handling the literal within the line.
1192 (test_reading_source_line): Update for location_get_source_line
1193 returning a char_span.
1194 * input.h (class char_span): Move here from
1195 diagnostic-show-locus.c, converting from a struct to a class.
1196 Make data members private.
1197 (char_span::operator bool): New.
1198 (char_span::length): New.
1199 (char_span::get_buffer): New.
1200 (char_span::operator[]): New.
1201 (char_span::subspan): Make const.
1202 (char_span::xstrdup): New.
1203 (location_get_source_line): Convert return type from const char *
1204 to char_span, losing the final "line_size" param.
1206 2018-04-30 Jan Hubicka <jh@suse.cz>
1208 * lto-wrapper.c (ltrans_priorities): New static var.
1209 (cmp_priority): New.
1210 (run_gcc): Read priorities and if doing parallel build order
1211 the Makefile by them.
1213 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1215 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1217 2018-04-30 Richard Biener <rguenther@suse.de>
1219 * tree-cfg.c (verify_address): Remove base argument, add
1220 flag whether to check TREE_ADDRESSABLE and do that.
1221 (verify_expr): Remove.
1222 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1223 (verify_gimple_assign_single): Likewise.
1224 (verify_gimple_switch): Likewise.
1225 (verify_expr_location_1): Dereference tp once. Add (disabled)
1226 piece from verify_expr.
1227 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1229 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1231 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1233 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1235 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1236 (small_data_pattern): Likewise.
1237 (arc_rewrite_small_data): Likewise.
1238 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1239 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1240 (get_symbol_alignment): New function.
1241 (legitimate_small_data_address_p): Likewise.
1242 (legitimate_scaled_address): Update, call
1243 legitimate_small_data_address_p.
1244 (output_sdata): New static variable.
1245 (arc_print_operand): Update how we handle small data operands.
1246 (arc_print_operand_address): Likewise.
1247 (arc_legitimate_address_p): Update, use
1248 legitimate_small_data_address_p.
1249 (arc_rewrite_small_data_p): Remove.
1250 (arc_rewrite_small_data_1): Likewise.
1251 (arc_rewrite_small_data): Likewise.
1252 (small_data_pattern): Likewise.
1253 (compact_sda_memory_operand): Update to use
1254 legitimate_small_data_address_p and get_symbol_alignment.
1255 (prepare_move_operands): Don't rewite sdata pattern.
1256 (prepare_extend_operands): Remove.
1257 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1259 (zero_extendqisi2): Likewise.
1260 (zero_extendhisi2): Likewise.
1261 (extendqihi2): Likewise.
1262 (extendqisi2): Likewise.
1263 (extendhisi2): Likewise.
1267 * config/arc/constraints.md (Usd): Change it to memory constraint.
1269 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1271 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1272 as source of std instructions.
1273 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1274 allow 6-bit constants as source for store instructions.
1275 (movdi_insn): Update instruction pattern to allow 6-bit constants
1276 as source for store instructions.
1278 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1280 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1282 2018-04-30 Nathan Sidwell <nathan@acm.org>
1283 Sandra Loosemore <sandra@codesourcery.com>
1285 * dumpfile.c (dump_open): Allow '-' for stdout.
1286 * doc/invoke.texi (Developer Options): Document dump filename
1287 determination early. Document stdin/stdout selection.
1289 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1291 Microblaze Target: PIC data text relative
1293 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1294 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1296 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1297 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1298 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1299 New addressing mode for data-text relative position indepenedent code.
1300 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1301 'ADDRESS_SYMBOLIC_TXT_REL'.
1302 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1303 (microblaze_legitimate_pic_operand): Exclude function calls from
1304 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1305 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1307 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1308 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1309 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1310 for 'address + offset'.
1311 (microblaze_expand_prologue): Add new function prologue call for
1313 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1314 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1315 table in case of TARGET_PIC_DATA_TEXT_REL.
1316 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1317 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1318 Add new macros 'UNSPEC_TEXT',
1319 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1320 + exclude function calls from 'UNSPEC_PLT' in case of data text
1322 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1323 new target hook for generating address diff vector tables in case of
1325 * doc/tm.texi : Regenerate.
1326 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1327 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1328 of addr diff vector generation.
1329 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1330 target hook definition.
1331 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1332 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1333 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1334 option for data text relative.
1336 2018-04-30 Richard Biener <rguenther@suse.de>
1338 * tree-chrec.h (evolution_function_is_constant_p): Remove
1340 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1342 2018-04-30 Richard Biener <rguenther@suse.de>
1345 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1347 2018-04-30 Richard Biener <rguenther@suse.de>
1349 PR tree-optimization/28364
1350 PR tree-optimization/85275
1351 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1352 copying first exit test.
1354 2018-04-28 Mark Wielaard <mark@klomp.org>
1356 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1358 (dwarf_AT): Handle DW_AT_addr_base.
1359 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1361 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1364 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1365 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1366 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1367 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1369 2018-04-28 Richard Biener <rguenther@suse.de>
1371 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1372 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1373 to reflect use. Only add interesting stmts.
1375 2018-04-27 Martin Jambor <mjambor@suse.cz>
1378 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1379 the jump function allows for passing through aggregate values.
1381 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1383 * input.h (in_system_header_at): Convert from macro to inline
1385 (from_macro_expansion_at): Likewise.
1386 (from_macro_definition_at): Likewise.
1388 2018-04-27 Jeff Law <law@redhat.com>
1390 * config.gcc: Mark tile* targets as deprecated/obsolete.
1392 2018-04-27 Richard Biener <rguenther@suse.de>
1394 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1397 2018-04-27 Richard Biener <rguenther@suse.de>
1399 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1401 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1403 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1404 with Yd constraint. Set "preferred_for_speed" attribute from
1405 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1407 (*movdi_internal): Ditto.
1408 (movti_interunit splitters): Remove
1409 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1410 (movdi_interunit splitters): Ditto.
1411 * config/i386/constraints.md (Ye): Remove.
1412 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1414 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1417 * config/aarch64/constraints.md (Usg): Limit to 31.
1420 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1422 PR tree-optimization/85529
1423 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1424 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1425 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1426 zero extension or masking of the MSB bit.
1427 (optimize_range_tests): Add FIRST_BB argument, pass it through
1428 to optimize_range_tests_var_bound.
1429 (maybe_optimize_range_tests, reassociate_bb): Adjust
1430 optimize_range_tests callers.
1432 2018-04-26 Richard Biener <rguenther@suse.de>
1433 Jakub Jelinek <jakub@redhat.com>
1435 * cgraph.h (symbol_table): Just declare debug method here.
1436 * symtab.c (symbol_table::debug): Define.
1438 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1440 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1442 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1444 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1445 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1446 (*movdi_internal): Substitute Yi and Yj constraint with x
1447 and Ym and Yn constraint with y constraint. Update "isa"
1448 attribute and set "preferred_for_speed" attribute from
1449 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1450 (*movsi_internal): Ditto.
1451 (*movdf_internal): Ditto.
1452 (*movsf_internal): Ditto.
1453 (*zero_extendsidi2): Ditto.
1454 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1455 (sse2_loadld): Ditto.
1456 (*vec_extract<ssevecmodelower>_0): Ditto.
1457 (*vec_extractv4si_0_zext_sse4): Ditto.
1458 (vec_concatv2di): Ditto.
1459 (*vec_dup<mode>): Ditto.
1460 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1461 * config/i386/constraints.md (Yi): Remove.
1466 2018-04-26 Nathan Sidwell <nathan@acm.org>
1468 * dumpfile.c (dump_open): New.
1469 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1470 (dump_finish): Detect stdio/stderr by value not name.
1472 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1474 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1476 2018-04-26 Tom de Vries <tom@codesourcery.com>
1479 * config/nvptx/nvptx.c (verify_neutering_jumps)
1480 (verify_neutering_labels): New function
1481 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1483 2018-04-26 Tom de Vries <tom@codesourcery.com>
1486 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1487 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1490 2018-04-26 Richard Biener <rguenther@suse.de>
1491 Tom de Vries <tom@codesourcery.com>
1494 * lto-streamer-out.c (output_function): Fixup loops if required to match
1495 discovery done in the reader.
1497 2018-04-26 Richard Biener <rguenther@suse.de>
1499 PR tree-optimization/85116
1500 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1501 have a loop exit from the single latch predecessor. Remove
1502 case of header with just condition.
1503 (ch_base::copy_headers): Exclude infinite loops from any
1505 (pass_ch::execute): Record exits.
1507 2018-04-26 Richard Biener <rguenther@suse.de>
1509 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1510 prologue cost vector and pass it to vect_get_load_cost.
1511 (vect_get_peeling_costs_all_drs): Likewise.
1512 (vect_peeling_hash_get_lowest_cost): Likewise.
1513 (vect_enhance_data_refs_alignment): Likewise.
1515 2018-04-26 Richard Biener <rguenther@suse.de>
1518 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1519 checking of integer<->pointer conversions.
1520 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1521 sign-/zero-extending pointer types.
1522 (expand_omp_for_static_chunk): Likewise.
1524 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1525 Jean Lee <xiaoyur347@gmail.com>
1527 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1528 (TARGET_ASAN_SHADOW_OFFSET): Define.
1529 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1530 true for -fsanitize=address.
1532 2018-04-25 Mark Wielaard <mark@klomp.org>
1534 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1537 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1539 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1540 than "alu", remove explicit "memory" and "imm_disp" attributes.
1541 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1544 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1545 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1546 gen_lowpart_no_emit.
1548 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1551 * config/i386/i386.c (ix86_expand_builtin): Change memory
1552 operand to XI, extend p0 to Pmode.
1553 * config/i386/i386.md: Change unspec volatile and operand
1554 1 mode to XI, change operand 0 mode to P.
1556 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1558 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1559 GET_MODE_MASK before any checking.
1560 (nds32_can_use_bset_p): Likewise.
1561 (nds32_can_use_btgl_p): Likewise.
1563 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1565 * config/nds32/nds32-doubleword.md: New define_split pattern for
1566 illegal register number.
1568 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1570 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1572 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1574 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1576 2018-04-25 Richard Biener <rguenther@suse.de>
1578 * lto-streamer.h (LTO_major_version): Bump to 8.
1580 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1582 * BASE-VER: Set to 9.0.0.
1584 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1586 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1587 in __abskf2 and __powikf2.
1589 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1592 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1593 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1594 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1595 Use the above on operand 2. Reindent.
1596 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1598 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1601 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1603 * config/i386/i386.opt (mcet): Removed.
1604 * doc/install.texi: Remove -mcet documentation.
1605 * doc/invoke.texi: Likewise.
1607 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1610 * doc/install.texi: Remove -mcet from bootstrap-cet.
1612 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1615 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1616 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1620 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1621 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1622 containing a CONST_VECTOR.
1624 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1626 * doc/install.texi: Update newlib dependency for nvptx.
1628 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1631 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1632 instead of INTVAL when shifting x left.
1634 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1636 PR tree-optimization/85478
1637 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1638 vect_grouped_store_supported for single element vectors.
1640 2018-04-24 Richard Biener <rguenther@suse.de>
1643 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1644 load cost increase to the case of non-constant step.
1646 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1649 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1650 destination if any_malformed_asm.
1652 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1655 * expr.c (store_field): In the bitfield case, if the value comes from
1656 a function call and is returned in registers by means of a PARALLEL,
1657 do not change the mode of the temporary unless BLKmode and VOIDmode.
1659 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1661 PR rtl-optimization/85423
1662 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1663 dependencies to debug insns when the previous insn is non-debug.
1665 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1667 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1668 enums into a single definition.
1669 (fls): Fix predicates and printing.
1672 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1674 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1675 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1676 and short u6 immediate.
1677 (check_if_valid_sleep_operand): Remove.
1678 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1680 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1682 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1683 flag_always_save_lp condition.
1684 * config/nds32/nds32.opt (malways-save-lp): New option.
1686 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1688 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1689 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1690 * config/nds32/nds32.h
1691 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1692 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1694 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1696 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1697 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1699 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1700 Chung-Ju Wu <jasonwucj@gmail.com>
1702 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1704 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1705 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1707 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1709 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1711 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1713 * config/nds32/nds32-protos.h (nds32_data_alignment,
1714 nds32_local_alignment): Declare.
1715 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1716 nds32_local_alignment): New functions.
1717 (TARGET_CONSTANT_ALIGNMENT): Define.
1718 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1720 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1722 * config/nds32/nds32.c
1723 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1724 (TARGET_MODES_TIEABLE_P): Likewise.
1726 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1728 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1731 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1732 Chung-Ju Wu <jasonwucj@gmail.com>
1734 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1735 for unaligned access.
1736 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1738 * config/nds32/nds32-intrinsic.md: Likewise.
1739 * config/nds32/nds32_intrinsic.h: Likewise.
1740 * config/nds32/nds32.h (nds32_builtins): Likewise.
1741 * config/nds32/nds32.opt (munaligned-access): New option.
1742 * config/nds32/nds32.c (nds32_asm_file_start): Display
1743 flag_unaligned_access status.
1745 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1747 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1748 -mno-relax is present.
1749 * config/riscv/linux.h (LINK_SPEC): Ditto.
1751 2018-04-20 Martin Sebor <msebor@redhat.com>
1754 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1756 (gimple_fold_builtin_stxcpy_chk): Same.
1757 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1759 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1762 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1763 __powikf2 when long double is IEEE 128-bit.
1765 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1767 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1768 step to make sure stack always aligned.
1770 2018-04-20 Carl Love <cel@us.ibm.com>
1773 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1774 size check for arg0.
1776 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1777 Tom de Vries <tom@codesourcery.com>
1780 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1781 Emit insns for calls too.
1782 (nvptx_find_par): Always look for worker-level predecessor insn.
1783 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1785 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1786 (nvptx_process_pars): Propagate frames for calls.
1788 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1791 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1793 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1794 (ix86_handle_option): Don't handle OPT_mibt.
1795 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1797 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1799 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1800 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1801 (ix86_target_macros): Define __CET__ with flag_cf_protection
1802 for -fcf-protection.
1803 * config/i386/i386.c (isa2_opts): Remove -mibt.
1804 * config/i386/i386.h (TARGET_IBT): Removed.
1805 (TARGET_IBT_P): Likewise.
1806 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1807 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1808 * config/i386/i386.opt (mcet): Update help message.
1811 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1812 -mcet as an alias for -mshstk.
1814 2018-04-20 Richard Biener <rguenther@suse.de>
1817 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1818 complexity by forcing a single use of the multiply operand.
1820 2018-04-20 Martin Jambor <mjambor@suse.cz>
1823 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1824 recursion dependency to only apply to non-clones.
1826 2018-04-20 Martin Jambor <mjambor@suse.cz>
1829 * ipa-cp.c (create_specialized_node): Check that clones of
1830 self-recursive edges exist during IPA-CP.
1832 2018-04-19 Toon Moene <toon@moene.org>
1834 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1837 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1839 PR tree-optimization/85467
1840 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1841 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1842 VECTOR_CST element to type.
1844 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1847 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1848 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1849 (builtin_longjmp): Likewise.
1850 (save_stack_nonlocal): New pattern.
1851 (restore_stack_nonlocal): Likewise.
1853 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1856 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1857 Replace ASM_OUTPUT_LABEL with fprintf.
1859 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1862 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1863 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1864 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1865 define __IBT__ and __SHSTK__ for -fcf-protection.
1866 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1868 (ix86_trampoline_init): Likewise.
1869 (x86_output_mi_thunk): Likewise.
1870 (ix86_notrack_prefixed_insn_p): Likewise.
1871 (ix86_option_override_internal): Don't disallow -fcf-protection.
1872 * config/i386/i386.md (rdssp<mode>): Also enable for
1874 (incssp<mode>): Likewise.
1875 (nop_endbr): Likewise.
1876 * config/i386/i386.opt (mcet): Change help message to built-in
1880 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1881 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1882 enable CET built-in functions.
1884 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1886 * common/config/i386/i386-common.c
1887 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1888 OPTION_MASK_ISA_MOVDIRI_UNSET,
1889 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1890 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1891 * config.gcc (movdirintrin.h): New header.
1892 * config/i386/cpuid.h (bit_MOVDIRI,
1893 bit_MOVDIR64B): New bits.
1894 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1896 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1897 (VOID, PVOID, PCVOID)): New function types.
1898 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1899 __builtin_ia32_directstoreu_u64,
1900 __builtin_ia32_movdir64b): New builtins.
1901 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1902 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1904 (ix86_valid_target_attribute_inner_p): Ditto.
1905 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1906 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1907 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1908 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1909 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1910 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1911 (movdiri<mode>, movdir64b_<mode>): New.
1912 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1913 * config/i386/immintrin.h: Include movdirintrin.h.
1914 * config/i386/movdirintrin.h: New file.
1915 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1917 2018-04-19 Richard Biener <rguenther@suse.de>
1920 * cfg.c (clear_bb_flags): When loop state says we have
1921 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1923 2018-04-19 Richard Biener <rguenther@suse.de>
1925 PR tree-optimization/84737
1926 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1927 copying restrict info.
1928 (vect_setup_realignment): Use it.
1929 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1930 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1931 the first DR to all generated stores.
1932 (vectorizable_load): Likewise for loads.
1934 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1936 PR tree-optimization/85446
1937 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1938 the integral and pointer types to have the same precision.
1940 * doc/install.texi: Document --disable-cet being the default and
1943 2018-04-18 Martin Liska <mliska@suse.cz>
1945 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1948 2018-04-18 Martin Liska <mliska@suse.cz>
1951 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1954 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1955 arguments if they are comparable.
1957 2018-04-18 Martin Liska <mliska@suse.cz>
1960 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1963 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1966 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1969 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1970 ENDBR after calling __morestack.
1972 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1975 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1976 by using gcc_base_ver to generate a gcc_driver_version, and use
1977 it when generating GCC_DRIVER_NAME.
1978 * configure: Regenerate.
1980 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1983 * config.gcc: Obsolete powerpc*-*-*spe*.
1985 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1988 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1989 (stabstr_D): Change type of unum from unsigned int to
1990 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1993 2018-04-17 Jim Wilson <jimw@sifive.com>
1996 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1997 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1998 Set arg_pointer_offset after using pretend_args_size.
2000 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2002 PR rtl-optimization/85431
2003 * dse.c (record_store): Ignore zero width stores.
2006 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2007 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2008 __builtin_stack_restore rather than after it.
2009 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2010 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2011 argument instead of virtual_dynamic_stack_rtx.
2013 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2015 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2017 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2018 Add note to error message to explain internal mapping of overloaded
2019 built-in function name to non-overloaded built-in function name.
2020 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2023 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2026 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2027 where the inputs overlap with the output.
2029 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2031 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2032 (=v, v) alternative and explicit "memory" attribute.
2033 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2034 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2036 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2037 "sselog1" type instead of "sselog".
2038 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2039 "sselog". Remove explicit "memory" attribute.
2040 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2041 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2043 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2044 "sselog1" type instead of "sselog". Remove explicit "memory"
2046 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2047 use "sselog1" type instead of "sselog". Remove explicit "memory"
2049 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2050 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2052 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2053 "sselog1" type instead of "sselog". Remove explicit "memory"
2055 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2056 use "sselog1" type instead of "sselog". Remove explicit "memory"
2060 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2063 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2066 2018-04-17 Martin Jambor <mjambor@suse.cz>
2069 * ipa-cp.c (create_specialized_node): Call
2070 expand_all_artificial_thunks if necessary.
2072 2018-04-17 Martin Liska <mliska@suse.cz>
2075 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2076 in message, remote space in between '_G' and '('.
2078 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2081 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2082 avx512f_vmcmp<mode>3<round_saeonly_name>,
2083 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2084 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2085 avx512f_rndscale<mode><round_saeonly_name>,
2086 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2087 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2088 Use %<iptr>2 instead of %2 for -masm=intel.
2089 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2090 avx512f_vcvttss2usi<round_saeonly_name>,
2091 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2093 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2094 avx512f_vcvttsd2usi<round_saeonly_name>,
2095 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2096 Use %q1 instead of %1 for -masm=intel.
2097 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2098 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2099 of %3 for -masm=intel.
2100 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2102 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2104 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2106 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2107 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2109 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2111 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2112 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2113 %g1 and one with %0 and %1.
2114 (avx512er_vmrcp28<mode><round_saeonly_name>,
2115 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2117 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2118 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2119 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2120 of %0 and %{%4%} for -masm=intel.
2121 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2122 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2123 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2124 order of %0 and %{%5%}%{z%} for -masm=intel.
2126 2018-04-17 Jan Hubicka <jh@suse.cz>
2129 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2131 2018-04-17 Martin Liska <mliska@suse.cz>
2134 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2135 for target_clone error message. Make default implementation
2136 clone to be a local declaration.
2137 (separate_attrs): Add new argument and check for an empty
2139 (expand_target_clones): Handle it.
2140 (ipa_target_clone): Make redirection just for target_clones
2143 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2144 Tom de Vries <tom@codesourcery.com>
2147 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2148 tiled basic blocks without omp continue statements.
2150 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2153 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2154 vec_extract expression as having side effects to make sure it gets
2157 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2160 * config/i386/i386.c (get_builtin_code_for_version): Check
2163 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2166 * gcc/config.gcc: Support "skylake".
2167 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2169 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2170 (processor_target_table): Add "skylake".
2171 (ix86_option_override_internal): Add "skylake".
2172 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2173 PROCESSOR_CANNONLAKE.
2174 (get_builtin_code_for_version): Fix priority for
2175 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2176 PROCESSOR_SKYLAKE-AVX512.
2177 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2178 (processor_type): Add PROCESSOR_SKYLAKE.
2180 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2181 Jason Merrill <jason@redhat.com>
2184 * convert.c (convert_to_integer_1): Use direct recursion for
2185 enumeral types and types with a precision less than the number
2186 of bits in their mode.
2188 2018-04-16 Julia Koval <julia.koval@intel.com>
2191 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2192 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2194 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2197 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2198 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2199 and -mno-direct-move.
2201 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2204 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2205 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2206 Ensure negative shifts result in {0}.
2208 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2210 PR rtl-optimization/79916
2211 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2212 regs (if any) to define how to gnerate SD moves when LRA is in
2215 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2217 PR rtl-optimization/85393
2218 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2219 * except.c (expand_dw2_landing_pad_for_region): Make static.
2220 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2221 a label and unconditional jump to old_bb, rather than
2222 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2225 PR rtl-optimization/85376
2226 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2227 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2228 instead of a specific value.
2230 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2231 Bin Cheng <bin.cheng@arm.com>
2233 PR tree-optimization/82965
2234 PR tree-optimization/83991
2235 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2236 by_profile_only parameter.
2237 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2238 information if the loop was predicted to iterate too many times.
2239 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2241 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2244 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2247 2018-04-13 Martin Liska <mliska@suse.cz>
2248 Jakub Jelinek <jakub@redhat.com>
2251 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2252 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2253 * builtins.c (expand_builtin_memory_copy_args): Use
2254 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2255 handle dest_addr == pc_rtx.
2257 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2260 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2261 asked to not generate direct moves.
2262 (fix_trunc<mode>si2_stfiwx): Similar.
2263 (fix_trunc<mode>si2_internal): Similar.
2265 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2268 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2269 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2270 lookup if dest in some wider mode is known to be const0_rtx and
2271 if so, record permanent equivalence for it to be ZERO_EXTEND of
2272 the narrower mode destination.
2274 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2276 * lto-streamer-out.c (output_function): Revert 259346.
2277 * omp-expand.c (expand_oacc_for): Likewise.
2279 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2281 PR rtl-optimization/85354
2282 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2283 * sel-sched.c (sel_global_init): ... here.
2285 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2288 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2289 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2290 mode for PE-COFF targets.
2291 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2292 (i386_pe_asm_lto_end): Likewise.
2293 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2294 (TARGET_ASM_LTO_END): Likewise.
2295 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2296 (i386_pe_asm_lto_start): New function.
2297 (i386_pe_asm_lto_end): Likewise.
2299 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2300 Richard Biener <rguenther@suse.de>
2303 * lto-streamer-out.c (output_function): Fix CFG loop state before
2305 * omp-expand.c (expand_oacc_for): Handle calls to internal
2306 functions like regular functions.
2308 2018-04-12 Richard Biener <rguenther@suse.de>
2311 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2312 for the early LTO debug to properly generate references to it
2313 during DIE emission. Do not re-use that for the skeleton for
2315 (dwarf2out_early_finish): Likewise.
2317 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2320 * config/i386/sse.md
2321 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2322 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2323 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2324 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2325 and output is a reg, avoid creating invalid lowpart subreg, but
2326 instead split into a 512-bit move. Don't split if not AVX512VL,
2327 input is xmm16+ reg and output is a mem.
2328 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2329 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2330 xmm16+ reg and output is a mem.
2332 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2334 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2335 also for flag_dwarf2_cfi_asm.
2337 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2339 PR rtl-optimization/85342
2340 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2341 a bool scalar var inside of the loop instead. Don't try to update
2342 recog_data.operand after failed apply_change_group.
2344 2018-04-12 Tom de Vries <tom@codesourcery.com>
2347 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2348 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2349 array with flexible array member as array without given dimension.
2350 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2351 argument for undefined param to true.
2353 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2356 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2357 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2358 from PowerPC section.
2359 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2360 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2361 help text that is too long.
2362 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2363 help text that is too long.
2364 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2365 help text that is too long.
2367 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2369 * config/alpha/alpha.md (stack_probe_internal): Rename
2370 from "probe_stack". Update all callers.
2372 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2374 PR rtl-optimization/84566
2375 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2376 sched_macro_fuse_insns.
2378 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2381 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2382 (compute_block_dependences): ... from here.
2384 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2386 PR tree-optimization/85331
2387 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2388 from int to HOST_WIDE_INT.
2390 2018-04-11 Martin Jambor <mjambor@suse.cz>
2393 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2394 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2395 not the same as the source val.
2396 (cgraph_edge_brings_value_p): New parameter.
2397 (gather_edges_for_value): Pass destination value to
2398 cgraph_edge_brings_value_p.
2399 (perhaps_add_new_callers): Likewise.
2400 (get_info_about_necessary_edges): Likewise and exclude values brought
2401 only by self-recursive edges.
2402 (create_specialized_node): Redirect only clones of self-calling edges.
2403 (+self_recursive_pass_through_p): New function.
2404 (find_more_scalar_values_for_callers_subset): Use it.
2405 (find_aggregate_values_for_callers_subset): Likewise.
2406 (known_aggs_to_agg_replacement_list): Removed.
2407 (decide_whether_version_node): Re-calculate known constants for all
2408 remaining context clones.
2410 2018-04-11 Richard Biener <rguenther@suse.de>
2413 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2414 from early DWARF output.
2415 (dwarf2out_early_finish): Output line info unconditionally into
2416 early DWARF and add reference to it.
2418 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2421 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2422 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2423 other than V2DFmode using iptr mode attribute.
2424 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2426 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2428 PR rtl-optimization/84659
2429 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2431 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2434 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2436 (output_loc_list): Pass address of a dummy size variable even in the
2437 locview handling loop.
2438 (index_location_lists): Add comment on why skip_loc_list_entry can't
2441 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2444 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2447 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2450 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2451 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2452 and -mstring-compare-inline-limit.
2454 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2457 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2458 for stack clash protection in a register whenever we need it to be in
2461 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2463 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2464 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2466 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2469 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2471 (mlong-double-): Ditto.
2472 * config/rs6000/sysv4.opt (msdata=): Ditto.
2473 (mtls-size=): Ditto.
2475 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2477 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2478 erroneous entries for
2479 "vector int vec_ldl (int, long int *)", and
2480 "vector unsigned int vec_ldl (int, unsigned long int *)".
2481 Add comments and entries for
2482 "vector bool char vec_ldl (int, bool char *)",
2483 "vector bool short vec_ldl (int, bool short *)",
2484 "vector bool int vec_ldl (int, bool int *)",
2485 "vector bool long long vec_ldl (int, bool long long *)",
2486 "vector pixel vec_ldl (int, pixel *)",
2487 "vector long long vec_ldl (int, long long *)",
2488 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2489 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2490 type tree bool_long_long_type_node and correct definition of
2491 bool_V2DI_type_node to make reference to this new type tree.
2492 (rs6000_mangle_type): Replace erroneous reference to
2493 bool_long_type_node with bool_long_long_type_node.
2494 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2495 comments to emphasize sign distinctions for char and int types and
2496 replace RS6000_BTI_bool_long constant with
2497 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2498 use of RS6000_BTI_pixel.
2499 (bool_long_type_node): Remove this macro definition.
2500 (bool_long_long_type_node): New macro definition
2502 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2504 PR rtl-optimization/85300
2505 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2506 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2507 simplify_unary_operation fails.
2509 2018-04-10 Martin Liska <mliska@suse.cz>
2511 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2512 cgraph_edge and ipa_ref.
2514 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2518 * config/i386/sse.md
2519 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2520 computation of the VEC_MERGE selector from mask.
2521 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2522 Fix decoding of the VEC_MERGE selector into mask.
2524 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2526 PR tree-optimization/85286
2527 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2529 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2531 * final.c (final_1): Set insn_last_address as well as
2532 insn_current_address.
2534 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2537 * explow.c (emit_stack_probe): Call validize_mem on memory location
2538 before passing it to gen_probe_stack. Create address operand and
2539 legitimize it for the probe_stack_address case.
2541 2018-04-09 Jan Hubicka <jh@suse.cz>
2544 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2545 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2546 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2548 (free_lang_data): Rebuild type inheritance graph.
2550 2018-04-09 Martin Sebor <msebor@redhat.com>
2552 * invoke.texi (-finline-small-functions): Mention other optimization
2554 (-findirect-inlining, -fpartial-inlining): Same.
2555 (-finline-functions-called-once): Same.
2556 (-freorder-blocks-and-partition): Same.
2558 2018-04-09 Jan Hubicka <jh@suse.cz>
2561 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2562 jumps; choose last target that matches the criteria (i.e.
2563 no partition changes for non-crossing jumps).
2564 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2565 support for redirecting crossing jumps to non-crossing.
2567 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2569 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2570 also for naked functions.
2572 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2574 * config/arc/arc.md (add_shift): New pattern.
2575 (add_shift2): Likewise.
2576 (sub_shift): Likewise.
2577 (sub_shift_cmp0_noout): Likewise.
2578 (compare_si_ashiftsi): Likewise.
2579 (xbfu_cmp0_noout): New combine pattern.
2580 (xbfu_cmp0"): Likewise.
2581 (movsi_set_cc_insn): Place the predicable variant first.
2582 (commutative_binary_cmp0_noout): Remove clobber.
2583 (commutative_binary_cmp0): New pattern.
2584 (noncommutative_binary_cmp0): Likewise.
2585 (noncommutative_binary_cmp0_noout): Likewise.
2586 (noncommutative_binary_comparison_result_used): Removed.
2587 (rsub_cmp0): New pattern.
2588 (rsub_cmp0_noout): Likewise.
2589 (extzvsi): Changed, keep only meaningful variants.
2590 (SQH, SEZ): New iterators.
2591 (SQH_postfix): New mode attribute.
2592 (SEZ_prefix): New code attribute.
2593 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2594 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2595 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2597 (noncommutative_operator): Check the availability of barrel
2600 2018-04-09 Richard Biener <rguenther@suse.de>
2602 PR tree-optimization/85284
2603 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2604 Only use the niter constraining form of simple_iv when the exit
2607 2018-04-09 Tom de Vries <tom@codesourcery.com>
2610 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2611 (define_expand "*memory_barrier"): New define_expand.
2612 (define_insn "memory_barrier"): New insn.
2614 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2616 PR rtl-optimization/80463
2617 PR rtl-optimization/83972
2618 PR rtl-optimization/83480
2620 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2621 correct producer for the insn.
2622 (tidy_control_flow): Fixup seqnos in case of debug insns.
2624 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2626 PR rtl-optimization/83913
2628 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2629 different sched-times when merging exprs.
2631 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2633 PR rtl-optimization/83962
2635 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2636 tidy_fallthru_edge and tidy_control_flow.
2638 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2640 PR rtl-optimization/83530
2642 * sel-sched.c (force_next_insn): New global variable.
2643 (remove_insn_for_debug): When force_next_insn is true, also leave only
2644 next insn in the ready list.
2645 (sel_sched_region): When the region wasn't scheduled, make another pass
2646 over it with force_next_insn set to 1.
2648 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2650 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2652 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2653 for interrupt control.
2654 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2655 functions for interrupt control.
2656 * config/nds32/nds32-intrinsic.md: Likewise.
2657 * config/nds32/nds32_intrinsic.h: Likewise.
2658 * config/nds32/nds32.h (nds32_builtins): Likewise.
2660 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2662 * config/nds32/nds32.c (nds32_init_machine_status,
2663 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2664 strict_aligned_p field.
2665 (nds32_expand_to_rtl_hook): New function.
2666 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2667 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2669 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2670 Chung-Ju Wu <jasonwucj@gmail.com>
2672 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2673 * config/nds32/nds32-n7.md: New file.
2674 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2675 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2677 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2678 * config/nds32/nds32.md (pipeline_model): Add n7.
2679 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2680 * config/nds32/pipelines.md: Include n7 settings.
2682 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2683 Chung-Ju Wu <jasonwucj@gmail.com>
2685 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2686 * config/nds32/nds32-e8.md: New file.
2687 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2688 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2690 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2691 * config/nds32/nds32.md (pipeline_model): Add e8.
2692 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2693 * config/nds32/pipelines.md: Include e8 settings.
2695 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2696 Chung-Ju Wu <jasonwucj@gmail.com>
2698 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2699 * config/nds32/nds32-n8.md: New file.
2700 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2701 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2703 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2704 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2705 * config/nds32/nds32.md (pipeline_model): Add n8.
2706 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2707 * config/nds32/pipelines.md: Include n8 settings.
2709 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2710 Chung-Ju Wu <jasonwucj@gmail.com>
2712 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2713 * config/nds32/nds32-n9-2r1w.md: New file.
2714 * config/nds32/nds32-n9-3r2w.md: New file.
2715 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2716 nds32_register_ports): New or modify for cpu n9.
2717 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2719 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2720 * config/nds32/nds32-utils.c: New file.
2721 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2722 TARGET_MUL_SLOW): Define.
2723 * config/nds32/nds32.md (pipeline_model): New attribute.
2724 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2725 New options that support cpu n9.
2726 * config/nds32/pipelines.md: Include n9 settings.
2727 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2729 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2731 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2732 information if necessary.
2733 (output_cond_branch_compare_zero): Likewise.
2734 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2735 (nds32_target_alignment): Refine for alignment.
2736 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2737 (FUNCTION_BOUNDARY): Modify.
2738 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2740 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2742 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2744 * config/nds32/constants.md (unspec_volatile_element): Add values for
2745 TLB operation and data prefetch.
2746 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2747 functions for TLB operation and data prefetch.
2748 * config/nds32/nds32-intrinsic.md: Likewise.
2749 * config/nds32/nds32_intrinsic.h: Likewise.
2750 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2751 (nds32_print_operand): Likewise.
2752 * config/nds32/nds32.h (nds32_builtins): Likewise.
2754 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2755 Andrew Pinski <pinsika@gcc.gnu.org>
2758 * match.pd: Use constant_boolean_node of correct type instead of
2759 boolean_true_node or boolean_false_node for simplifying
2760 pointer comparisons to zero.
2762 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2764 PR tree-optimization/80021
2765 * tree.c (verify_type_variant): Make error call in verify_variant_match
2766 translatable and remove final full stop.
2768 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2770 * config/nds32/constants.md (unspec_volatile_element): Add
2771 UNSPEC_VOLATILE_EH_RETURN.
2772 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2773 nds32_output_stack_pop): Support dwarf exception handling process.
2774 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2775 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2776 exception handling process.
2777 (nds32_compute_stack_frame): Likewise.
2778 (nds32_return_addr_rtx): Likewise.
2779 (nds32_initial_elimination_offset): Likewise.
2780 (nds32_expand_prologue): Likewise.
2781 (nds32_expand_epilogue): Likewise.
2782 (nds32_dynamic_chain_address): New function.
2783 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2785 (DYNAMIC_CHAIN_ADDRESS): Define.
2786 (EH_RETURN_DATA_REGNO): Define.
2787 (EH_RETURN_STACKADJ_RTX): Define.
2788 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2789 patterns for dwarf exception handling.
2791 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2793 * config/nds32/nds32.h: Clean up obsolete macros.
2795 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2797 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2798 Add enum values for particular instructions.
2799 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2800 particular intrinsic functions.
2801 * config/nds32/nds32-intrinsic.md: Likewise.
2802 * config/nds32/nds32_intrinsic.h: Likewise.
2803 * config/nds32/nds32.h (nds32_builtins): Likewise.
2804 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2805 (btst, ave): New patterns for particular instructions.
2807 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2809 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2810 Add enum values for atomic load/store and memory sync.
2811 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2813 * config/nds32/nds32-intrinsic.md: Likewise.
2814 * config/nds32/nds32_intrinsic.h: Likewise.
2815 * config/nds32/nds32.h (nds32_builtins): Likewise.
2817 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2819 PR tree-optimization/85257
2820 * fold-const.c (native_encode_vector): If not all elts could fit
2821 and off is -1, return 0 rather than offset.
2822 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2823 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2824 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2825 adjust buffer in native_interpret_expr call.
2827 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2829 * config/nds32/constants.md (unspec_volatile_element): Add cache
2830 control enum values.
2831 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2832 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2833 * config/nds32/nds32.c (nds32_cctl_names): New.
2834 (nds32_print_operand): Handle cache control register names.
2835 * config/nds32/nds32.h (nds32_builtins): New enum values.
2836 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2838 * config/nds32/nds32.md (type): Add mmu.
2839 * config/nds32/pipelines.md (simple_insn): Add mmu.
2841 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2843 * config/nds32/nds32.md (type): Remove call.
2844 * config/nds32/pipelines.md (simple_insn): Likewise.
2846 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2848 * config/nds32/constants.md (unspec_volatile_element): Add
2849 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2850 UNSPEC_VOLATILE_FMFCFG.
2851 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2852 description for fmfcfg and fmfcsr.
2853 (bdesc_1arg): Add fmtcsr.
2854 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2855 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2856 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2857 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2858 unspec_fmfcfg): New patterns.
2859 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2860 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2861 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2862 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2863 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2864 __nds32__fmfcfg): Define.
2866 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2868 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2869 intrinsic register names.
2870 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2871 intrinsic register enum values and macros.
2873 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2875 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2876 for load/store addressing form.
2877 (nds32_print_operand_address): Likewise.
2879 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2882 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2883 based on LABEL_REF. Remove useless assertion.
2884 (pic_address_needs_scratch): Fix formatting.
2885 (sparc_legitimize_pic_address): Minor tweaks.
2886 (sparc_delegitimize_address): Adjust assertion accordingly.
2887 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2888 into symbolic_operand.
2889 (movsi_high_pic_label_ref): Likewise.
2890 (movsi_lo_sum_pic_label_ref): Likewise.
2891 (movdi_pic_label_ref): Likewise.
2892 (movdi_high_pic_label_ref): Likewise.
2893 (movdi_lo_sum_pic_label_ref): Likewise.
2895 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2897 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2898 custom LIB_SPEC setup.
2900 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2901 Kito Cheng <kito.cheng@gmail.com>
2903 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2904 * config/riscv/freebsd.h: New.
2906 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2908 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2909 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2912 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2913 Kito Cheng <kito.cheng@gmail.com>
2915 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2916 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2917 * config/nds32/nds32-protos.h (nds32_output_call,
2918 nds32_output_return): Declare.
2919 * config/nds32/nds32.md: Refine all the call and return patterns.
2921 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2924 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2925 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2927 PR rtl-optimization/84872
2928 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2929 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2932 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2934 * expr.c (copy_blkmode_to_reg): Revert 254862.
2935 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2937 2018-04-06 Richard Biener <rguenther@suse.de>
2940 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2941 after seeing a component reference with an adjacent field. Treat
2942 refs to arrays at struct end of external decls similar to
2943 refs to unconstrained commons.
2945 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2948 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2949 look through SAVE_EXPRs with non-side-effects argument. Adjust
2951 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2954 2018-04-06 Richard Biener <rguenther@suse.de>
2957 * alias.c (find_base_term): New wrapper around find_base_term
2958 unwinding CSELIB_VAL_PTR changes.
2959 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2962 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2964 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2966 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2967 constant definitions.
2968 ("nop"): lr 0,0 -> nopr r0
2969 ("nop_lr0", "nop_lr1"): New insn definitions.
2971 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2973 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2974 NDS32_V3PUSH_AVAILABLE_P macro.
2976 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2977 Chung-Ju Wu <jasonwucj@gmail.com>
2979 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2980 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2981 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2982 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2983 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2984 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2985 * config/nds32/constraints.md: New constraints and checking for hard
2986 float configuration.
2987 * config/nds32/iterators.md: New mode iterator and attribute for hard
2988 float configuration.
2989 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2991 * config/nds32/nds32-fpu.md: New file.
2992 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2993 deal with hard float code generation.
2994 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2996 (abi_type, float_reg_number): New enum type.
2997 * config/nds32/nds32-predicates.c: New predicates for hard float.
2998 * config/nds32/nds32-protos.h: Declare functions for hard float.
2999 * config/nds32/nds32.c: Implementation for hard float configuration.
3000 * config/nds32/nds32.h: Definitions for hard float configuration.
3001 * config/nds32/nds32.md: Include hard float machine description and
3002 modify patterns for hard float configuration.
3003 * config/nds32/nds32.opt: New options for hard float configuration.
3004 * config/nds32/predicates.md: New predicates for hard float
3007 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3009 * common/config/nds32/nds32-common.c
3010 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3012 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3015 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3016 CONSTRUCTOR_ELT (ctor, ...)->value.
3018 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3021 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3023 2018-04-05 Tom de Vries <tom@codesourcery.com>
3026 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3029 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3030 Kito Cheng <kito.cheng@gmail.com>
3032 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3033 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3034 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3035 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3037 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3038 Kito Cheng <kito.cheng@gmail.com>
3040 * config/nds32/constraints.md (Ufe): New memory constraint.
3041 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3042 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3043 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3045 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3046 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3048 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3050 * config/nds32/nds32.md: Use optimize_size in the condition for
3051 alu-shift instructions.
3053 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3055 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3057 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3059 * config/nds32/nds32.md (negsi2): Refine pattern.
3061 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3062 Chung-Ju Wu <jasonwucj@gmail.com>
3064 * config/nds32/iterators.md (shift_rotate): New code iterator.
3065 (shift): New code attribute.
3066 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3067 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3068 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3069 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3070 bit-wise operations.
3071 (andsi3, *andsi3): Ditto.
3072 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3073 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3074 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3075 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3076 nds32_ior_operand, nds32_xor_operand): New predicates.
3078 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3080 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3081 (addsi3, subsi3): ... this.
3083 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3085 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3087 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3089 * config/nds32/nds32.md: Adjust indention.
3091 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3093 * config/nds32/nds32.md (feature): New attribute.
3095 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3097 * config/nds32/nds32.md (subtype): New attribute.
3099 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3102 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3103 expansion to perform a bitwise AND of the argument followed by a
3104 boolean negation of the result.
3106 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3108 PR rtl-optimization/84878
3109 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3110 the basic block. Assert the use reference is not artificial and that
3111 it has an associated insn.
3113 2018-04-04 Michael Matz <matz@suse.de>
3115 * builtins.c (compute_objsize): Pass correct operand
3116 to array_at_struct_end_p.
3118 2018-04-04 Richard Biener <rguenther@suse.de>
3121 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3122 from contexts for DINFO_LEVEL_TERSE and below.
3124 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3126 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3127 resiter_operand condition.
3128 * config/nds32/nds32.md (*move<mode>): Ditto.
3130 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3131 Monk Chiang <sh.chiang04@gmail.com>
3133 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3135 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3137 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3139 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3140 Kito Cheng <kito.cheng@gmail.com>
3142 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3143 nds32_cond_code_str, output_cond_branch,
3144 output_cond_branch_compare_zero, nds32_expand_cbranch,
3145 nds32_expand_cstore, nds32_expand_movcc,
3146 nds32_output_cbranchsi4_equality_zero,
3147 nds32_output_cbranchsi4_equality_reg,
3148 nds32_output_cbranchsi4_equality_reg_or_const_int,
3149 nds32_output_cbranchsi4_greater_less_zero: New functions.
3150 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3151 nds32_expand_cstore, nds32_expand_movcc,
3152 nds32_output_cbranchsi4_equality_zero,
3153 nds32_output_cbranchsi4_equality_reg,
3154 nds32_output_cbranchsi4_equality_reg_or_const_int,
3155 nds32_output_cbranchsi4_greater_less_zero): Declare.
3156 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3157 nds32_rimm11s_operand): New predicates.
3158 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3159 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3162 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3164 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3165 * config/nds32/nds32.md: Ditto.
3166 * config/nds32/pipelines.md: Ditto.
3168 2018-04-04 Richard Biener <rguenther@suse.de>
3170 PR tree-optimization/85168
3171 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3172 propagating abnormals.
3174 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3176 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3178 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3179 Kito Cheng <kito.cheng@gmail.com>
3181 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3182 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3183 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3184 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3185 * config/nds32/nds32.md (sibcall_internal): New.
3186 (sibcall_register): Remove.
3187 (sibcall_immediate): Remove.
3188 (sibcall_value_internal): New.
3189 (sibcall_value_register): Remove.
3190 (sibcall_value_immediate): Remove.
3191 * config/nds32/predicates.md (nds32_general_register_operand): New.
3192 (nds32_call_address_operand): New.
3194 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3196 PR rtl-optimization/85167
3197 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3198 bb_defs if *split_p, instead preinitialize it to NULL.
3200 PR tree-optimization/85156
3201 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3202 evaluating the argument multiple times.
3204 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3206 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3208 (_mm_cvtpd_ps): Likewise.
3209 (_mm_cvttpd_epi32): Likewise.
3210 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3211 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3212 vector, pixel, and bool following altivec.h include.
3214 2018-04-03 Martin Sebor <msebor@redhat.com>
3216 * doc/extend.texi (Common Function Attributes): Clarify.
3217 (const attribute): Likewise.
3218 (pure attribute): Likewise.
3220 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3223 * config/i386/i386.c (ix86_expand_vector_set): Use
3224 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3226 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3228 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3229 instructions when changing rounding bits to preserve precision bits
3230 in the x87 control word.
3232 2018-04-03 Martin Liska <mliska@suse.cz>
3234 PR tree-optimization/82491
3235 * rtl.h (strip_offset_and_add): Replace += suboffset with
3236 poly_uint64 () + suboffset.
3238 2018-03-29 Martin Liska <mliska@suse.cz>
3239 Martin Jambor <mjambor@suse.cz>
3242 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3243 param_type is not an integral or pointer type.
3245 2018-04-03 Richard Biener <rguenther@suse.de>
3247 * sese.h (recompute_all_dominators): Remove.
3249 2018-04-02 Martin Sebor <msebor@redhat.com>
3251 * doc/invoke.texi (-Wrestrict): Fix typos.
3253 2018-04-02 Jim Wilson <jimw@sifive.com>
3255 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3256 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3257 (<optab>di3, <optab>si3_extend): Likewise.
3258 (<optab>si3_mask, <optab>si3_mask_1): New.
3259 (<optab>di3_mask, <optab>di3_mask_1): New.
3260 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3261 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3262 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3264 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3266 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3269 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3271 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3272 (nds32_canonicalize_comparison): New function.
3274 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3275 Kito Cheng <kito.cheng@gmail.com>
3276 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3278 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3279 * config/nds32/constants.md (unspec_volatile_element): Add
3280 UNSPEC_VOLATILE_RELAX_GROUP.
3281 * config/nds32/nds32-relax-opt.c: New file.
3282 * config/nds32/nds32-predicates.c
3283 (nds32_symbol_load_store_p): New function.
3284 * config/nds32/nds32-protos.h
3285 (nds32_symbol_load_store_p): Declare function.
3286 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3287 * config/nds32/nds32.c
3288 (nds32_register_pass): New function to register pass.
3289 (nds32_register_passes): New function to register passes.
3290 * config/nds32/nds32.md (relax_group): New pattern.
3291 * config/nds32/nds32.opt (mrelax-hint): New option.
3292 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3294 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3296 * config/nds32/t-nds32: Modify files dependency.
3298 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3300 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3301 (PROFILE_HOOK): Define its implementation.
3303 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3305 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3306 type and 32-bit size.
3308 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3311 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3312 (V_128_256): New mode iterator.
3313 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3314 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3315 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3317 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3318 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3320 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3323 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3324 NaN inputs correctly.
3326 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3329 * config/rs6000/vsx.md (??r): New mode attribute.
3330 (*vsx_mov<mode>_64bit): Use it.
3331 (*vsx_mov<mode>_32bit): Likewise.
3333 2018-03-30 Martin Sebor <msebor@redhat.com>
3335 PR tree-optimization/84818
3336 * builtins.c (check_access): Use warning_n.
3338 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3341 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3343 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3346 2018-03-30 Julia Koval <julia.koval@intel.com>
3349 * x86-tune.def (movx, partial_reg_dependency): Enable for
3352 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3355 * lra-constraints.c (process_alt_operands): Move setting
3356 this_alternative_matches below.
3358 2018-03-29 Martin Liska <mliska@suse.cz>
3361 * doc/invoke.texi: Document how LTO works with debug info.
3362 Describe auto-load support of binutils. Mention 'x86-64'
3363 as valid option value of -march option.
3365 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3367 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3370 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3371 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3372 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3375 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3378 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3379 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3380 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3381 for RS6000_BTM_POWERPC64.
3382 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3383 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3384 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3389 2018-03-28 Carl Love <cel@us.ibm.com>
3392 2017-09-27 Carl Love <cel@us.ibm.com>
3394 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3395 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3396 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3399 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3401 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3402 instead of __vector bool.
3403 (_mm_max_pu8): Likewise.
3404 (_mm_min_pi16): Likewise.
3406 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3409 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3410 (DIVWEUO): Likewise.
3412 (DIVDEUO): Likewise.
3413 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3414 DIVWEUO and DIVDEUO.
3415 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3416 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3417 (div_extend): Likewise.
3418 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3420 (__builtin_divweuo): Likewise.
3421 (__builtin_divdeo): Likewise.
3422 (__builtin_divdeuo): Likewise.
3424 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3427 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3428 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3430 PR tree-optimization/82004
3431 * gimple-match-head.c (optimize_pow_to_exp): New function.
3432 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3433 Don't fold to exp if optimize_pow_to_exp is false.
3435 2018-03-28 Martin Liska <mliska@suse.cz>
3438 * calls.c (initialize_argument_information): Fix trailing space.
3439 * common.opt: Fix typo and provide better explanation for
3440 -fsanitize-coverage option.
3441 * config/i386/i386.opt: Fix typo.
3443 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3444 Martin Liska <mliska@suse.cz>
3447 * gimplify.c (asan_poison_variable): Don't do the check for
3448 gimplify_omp_ctxp here.
3449 (gimplify_decl_expr): Do it here.
3450 (gimplify_target_expr): Likewise.
3452 2018-03-28 Martin Liska <mliska@suse.cz>
3455 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3456 chkp_type_bounds_count if MPX is not enabled.
3458 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3460 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3462 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3465 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3466 function to create the function decl for complex long double
3467 multiply and divide for -mabi=ieeelongdouble.
3468 (init_float128_ieee): Call it.
3470 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3473 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3474 -fcf-protection=branch -mibt.
3475 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3477 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3480 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3482 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3485 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3486 extern array declarations.
3488 2018-03-27 Richard Biener <rguenther@suse.de>
3491 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3492 explicit single_use checks.
3494 2018-03-27 Richard Biener <rguenther@suse.de>
3496 PR tree-optimization/85082
3497 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3500 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3502 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3503 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3504 Turn on fasynchronous-unwind-tables and funwind-tables.
3506 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3509 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3510 (*bmi_blsr_<mode>_ccz): Ditto.
3512 2018-03-26 Tom de Vries <tom@codesourcery.com>
3514 PR tree-optimization/85063
3515 * omp-general.c (offloading_function_p): New function. Factor out
3517 * omp-offload.c (pass_omp_target_link::gate): ... here.
3518 * omp-general.h (offloading_function_p): Declare.
3519 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3520 with attribute omp declare target for offloading functions.
3522 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3524 PR tree-optimization/84005
3525 * tree-data-ref.h (get_base_for_alignment): Declare.
3526 * tree-data-ref.c (get_base_for_alignment_1): New function.
3527 (get_base_for_alignment): Likewise.
3528 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3529 get_base_for_alignment to find a suitable base object, instead
3530 of always using drb->base_address.
3532 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3535 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3536 known size by default.
3538 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3541 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3542 and non BLKmode operands.
3544 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3547 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3548 Clean up attributes.
3550 2018-03-23 Richard Biener <rguenther@suse.de>
3553 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3554 we are going to emit early debug for LTO.
3556 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3559 * function.c (match_asm_constraints_1): Don't optimize if input
3560 doesn't satisfy general_operand predicate for output's mode.
3563 * alias.c (write_dependence_p): Don't require for x_canonicalized
3564 non-VOIDmode if x has VOIDmode.
3567 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3568 just don't try to optimize it rather than assert it never happens.
3570 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3572 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3573 macro expansions for definition of ST_INTERNAL_<mode> and
3574 LD_INTERNAL_<mode> builtins.
3575 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3577 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3579 (altivec_expand_st_builtin): Likewise.
3580 (altivec_expand_builtin): Remove calls to deleted functions.
3581 (rs6000_address_for_altivec): Delete this function.
3582 * config/rs6000/vector.md: Remove expands for
3583 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3585 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3588 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3589 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3590 re-computing once computed.
3591 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3592 (arm_init_machine_status): Initialize
3593 machine->static_chain_stack_bytes.
3595 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3598 * doc/extend.texi: Add four new prototypes for vec_ld.
3599 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3600 definitions for more logical presentation.
3601 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3602 entries for V1TI variants of __builtin_altivec_ld builtin.
3603 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3604 handling of V1TI variant of LVX icode pattern.
3605 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3606 (rs6000_gimple_fold_builtin): Likewise.
3607 (altivec_init_builtins): Add code to define
3608 __builtin_altivec_lvx_v1ti function.
3610 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3613 * function.c (match_asm_constraints_1): Don't do the optimization
3614 if input isn't a REG, SUBREG, MEM or constant.
3616 2018-03-22 Tom de Vries <tom@codesourcery.com>
3618 PR tree-optimization/84956
3619 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3620 bb_has_abnormal_pred.
3622 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3625 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3626 DECL_INITIAL (decl) to decl at the end.
3627 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3630 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3632 * doc/extend.texi (__builtin_tgmath): Document when complex
3633 integer types are treated as _Complex _Float64.
3635 2018-03-21 Tom de Vries <tom@codesourcery.com>
3637 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3639 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3641 PR tree-optimization/84960
3642 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3643 if it is ENTRY block, move them into single succ of ENTRY in that case.
3645 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3647 PR tree-optimization/84811
3648 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3649 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3650 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3651 (known_subrange_p): Update accordingly. Cast each value involved
3652 in the size comparison, rather than casting the result of the
3655 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3657 PR tree-optimization/84982
3658 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3659 by flipping the least significant bit rather than all bits from
3660 bitpos to bitpos + bitsize - 1.
3662 2018-03-21 Nathan Sidwell <nathan@acm.org>
3664 * doc/extend.texi (Deprecated Features): Remove mention of
3665 long-deleted deprecations.
3667 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3670 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3671 * configure: Regenerate.
3673 2018-03-21 Tom de Vries <tom@codesourcery.com>
3675 PR tree-optimization/83126
3676 * tree-parloops.c (num_phis): New function.
3677 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3679 2018-03-21 Nathan Sidwell <nathan@acm.org>
3681 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3682 mention anon-struct/union members and trailing attributes.
3684 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3686 PR tree-optimization/84969
3687 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3688 builtin memset partitions if they set different rhs values.
3690 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3692 PR rtl-optimization/84989
3693 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3694 VEC_DUPLICATE with scalar result mode.
3696 2018-03-21 Martin Liska <mliska@suse.cz>
3699 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3700 not intended return statement.
3702 2018-03-21 Martin Liska <mliska@suse.cz>
3705 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3706 (chkp_find_bound_slots_1): Limit number of iterations.
3708 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3711 * Minor grammar fixes for x86 options.
3713 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3716 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3717 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3719 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3722 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3723 (altivec_lvx_<mode>_1op): Likewise.
3724 (altivec_stvx_<mode>_2op): Likewise.
3725 (altivec_stvx_<mode>_1op): Likewise.
3726 (altivec_lvx_<VM2:mode>): New define_expand.
3727 (altivec_stvx_<VM2:mode>): Likewise.
3728 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3729 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3730 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3731 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3732 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3733 (rs6000_gen_lvx): Likewise.
3734 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3735 (altivec_expand_stv_builtin): Likewise.
3736 (altivec_expand_builtin): Likewise.
3737 * config/rs6000/vector.md: Likewise.
3739 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3742 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3745 2018-03-20 Richard Biener <rguenther@suse.de>
3748 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3749 sign-conversions as zero, fall back to standard scalar_stmt
3752 2018-03-20 Martin Liska <mliska@suse.cz>
3755 * predict.c (rebuild_frequencies): Handle case when we have
3756 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3758 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3761 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3762 flag_section_anchors.
3763 * varasm.c (use_blocks_for_decl_p): Remove hack for
3764 dw2_force_const_mem.
3767 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3769 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3770 be created, use lowpart_subreg of operands[0] rather than operands[0]
3772 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3773 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3774 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3775 and n constraint instead of aarch64_shift_imm_di and Usd.
3776 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3777 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3779 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3782 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3783 to favor GPR over NEON registers.
3784 (<shift>di3_neon): Likewise.
3786 2018-03-20 Tom de Vries <tom@codesourcery.com>
3789 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3790 (nvptx_process_pars): Emit bar.sync asap and alap.
3792 2018-03-20 Tom de Vries <tom@codesourcery.com>
3795 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3796 seen_label if seen_label is already set.
3798 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3801 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3802 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3803 Use 1U instead of 1. Formatting fixes.
3806 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3807 instead of TREE_TYPE (s1) for the return value.
3809 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3811 PR tree-optimization/84946
3812 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3813 bitsize + bitsize in poly_uint64 rather than poly_int64.
3816 * dwarf2asm.c: Include fold-const.c.
3817 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3818 of decl rather than decl itself.
3820 PR rtl-optimization/84643
3821 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3823 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3826 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3827 calling assemble_variable.
3829 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3832 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3833 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3835 2018-03-19 Jim Wilson <jimw@sifive.com>
3838 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3839 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3840 (riscv_first_stack_step): Likewise.
3841 (riscv_option_override): Use STACK_BOUNDARY instead of
3843 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3845 (BIGGEST_ALIGNMENT): Set to 128.
3846 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3847 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3850 2018-03-19 Richard Biener <rguenther@suse.de>
3852 PR tree-optimization/84933
3853 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3854 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3856 2018-03-19 Richard Biener <rguenther@suse.de>
3858 PR tree-optimization/84859
3859 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3860 (cond_if_else_store_replacement): Perform sinking operation on
3861 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3862 Generalize what a BB with a single eligible store is.
3864 2018-03-19 Richard Biener <rguenther@suse.de>
3866 PR tree-optimization/84929
3867 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3868 chrec_is_positive against non-chrec arg.
3870 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3873 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3875 2018-03-18 Martin Liska <mliska@suse.cz>
3877 PR rtl-optimization/84635
3878 * regrename.c (build_def_use): Use matches_mode only when
3881 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3883 PR tree-optimization/84913
3884 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3885 vectorize chains of COND_EXPRs.
3887 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3889 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3891 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3893 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3895 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3897 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3899 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3900 Kito Cheng <kito.cheng@gmail.com>
3902 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3903 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3904 (nds32_adjust_reg_alloc_order): New function.
3905 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3907 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3909 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3910 nds32_print_operand, nds32_print_operand_address): Use
3911 HOST_WIDE_INT_PRINT_DEC instead.
3913 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3915 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3917 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3920 * config/i386/i386.c (initial_ix86_tune_features,
3921 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3923 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3924 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3925 rather than 1u << ix86_tune. Formatting fix.
3926 (ix86_option_override_internal): Change ix86_arch_mask from
3927 unsigned int to unsigned HOST_WIDE_INT, initialize to
3928 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3929 (ix86_function_specific_restore): Likewise.
3931 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3934 * postreload.c (reload_combine_recognize_pattern): Perform
3935 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3936 truncate_int_for_mode the result for the destination's mode.
3939 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3940 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3942 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3945 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3948 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3949 regno_allocno_class_array and sorted_pseudos.
3950 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3951 insns where regno is used.
3953 2018-03-16 Martin Liska <mliska@suse.cz>
3956 * multiple_target.c (create_dispatcher_calls): Redirect
3957 reference in the symbol table.
3959 2018-03-16 Martin Liska <mliska@suse.cz>
3962 * multiple_target.c (create_dispatcher_calls): Redirect also
3965 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3969 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3970 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3971 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3973 2018-03-16 Julia Koval <julia.koval@intel.com>
3975 * doc/invoke.texi (Skylake Server): Add CLWB.
3976 Cannonlake): Remove CLWB.
3978 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3980 PR tree-optimization/84841
3981 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3983 (FLOAT_ONE_CONST_TYPE): Define.
3984 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3985 (sort_by_operand_rank): Put entries with higher constant_type last
3986 rather than first to match comments.
3988 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3990 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3993 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3996 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3997 check and use address of target always.
3999 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4002 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4003 (indirect_thunk_bnd_needed): Likewise.
4004 (indirect_thunks_used): Likewise.
4005 (indirect_thunks_bnd_used): Likewise.
4006 (indirect_return_needed): New.
4007 (indirect_return_bnd_needed): Likewise.
4008 (output_indirect_thunk_function): Add a bool argument for
4010 (output_indirect_thunk_function): Don't generate alias for
4011 function return thunk.
4012 (ix86_code_end): Call output_indirect_thunk_function to generate
4013 function return thunks.
4014 (ix86_output_function_return): Set indirect_return_bnd_needed
4015 and indirect_return_needed instead of indirect_thunk_bnd_needed
4016 and indirect_thunk_needed.
4018 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4020 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4021 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4022 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4024 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4025 Paul Hua <paul.hua.gm@gmail.com>
4028 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4030 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4032 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4033 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4036 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4039 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4040 instead of GET_MODE_SIZE when comparing Units.
4042 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4045 * varasm.c (hash_section): Return an unchangeble hash value
4046 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4047 Return !aarch64_can_use_per_function_literal_pools_p ().
4049 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4052 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4053 rather than address of cmode as last argument to prepare_cmp_insn.
4055 2018-03-15 Julia Koval <julia.koval@intel.com>
4057 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4058 F_AVX512VNNI, F_AVX512BITALG): New.
4060 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4063 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4064 insn for floating-point loads and stores.
4066 2018-03-14 Carl Love <cel@us.ibm.com>
4068 * config/rs6000/rs6000-c.c: Add macro definitions for
4069 ALTIVEC_BUILTIN_VEC_PERMXOR.
4070 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4071 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4072 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4073 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4075 * config/doc/extend.texi: Add prototypes for vec_permxor.
4077 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4080 * diagnostic-show-locus.c (class layout_point): Convert m_line
4081 from int to linenum_type.
4082 (line_span::comparator): Use linenum "compare" function when
4083 comparing line numbers.
4084 (test_line_span): New function.
4085 (layout_range::contains_point): Convert param "row" from int to
4087 (layout_range::intersects_line_p): Likewise.
4088 (layout::will_show_line_p): Likewise.
4089 (layout::print_source_line): Likewise.
4090 (layout::should_print_annotation_line_p): Likewise.
4091 (layout::print_annotation_line): Likewise.
4092 (layout::print_leading_fixits): Likewise.
4093 (layout::annotation_line_showed_range_p): Likewise.
4094 (struct line_corrections): Likewise for field m_row.
4095 (line_corrections::line_corrections): Likewise for param "row".
4096 (layout::print_trailing_fixits): Likewise.
4097 (layout::get_state_at_point): Likewise.
4098 (layout::get_x_bound_for_row): Likewise.
4099 (layout::print_line): Likewise.
4100 (diagnostic_show_locus): Likewise for locals "last_line" and
4102 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4103 * input.c (selftest::test_linenum_comparisons): New function.
4104 (selftest::input_c_tests): Call it.
4105 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4106 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4107 * selftest.h (ASSERT_GT): New macro.
4108 (ASSERT_GT_AT): New macro.
4109 (ASSERT_LT): New macro.
4110 (ASSERT_LT_AT): New macro.
4112 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4114 PR rtl-optimization/84780
4115 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4117 2018-03-14 Martin Liska <mliska@suse.cz>
4119 * tree.c (record_node_allocation_statistics): Use
4120 get_stats_node_kind.
4121 (get_stats_node_kind): New function extracted from
4122 record_node_allocation_statistics.
4123 (free_node): Use get_stats_node_kind.
4125 2018-03-14 Richard Biener <rguenther@suse.de>
4127 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4128 that the value-set of ANTIC_IN doesn't grow.
4131 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4133 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4134 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4136 2018-03-14 Julia Koval <julia.koval@intel.com>
4138 * config.gcc (icelake-client, icelake-server): New.
4140 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4141 (initial_ix86_arch_features): Ditto.
4142 (PTA_SKYLAKE): Add SGX.
4143 (PTA_ICELAKE): Remove.
4144 (PTA_ICELAKE_CLIENT): New.
4145 (PTA_ICELAKE_SERVER): New.
4146 (ix86_option_override_internal): Split up icelake on icelake client and
4148 (get_builtin_code_for_version): Ditto.
4149 (fold_builtin_cpu): Ditto.
4150 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4151 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4152 * config/i386/i386.h (processor_type): Ditto.
4153 * doc/invoke.texi: Ditto.
4155 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4158 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4159 INTEGER_CST offset, add it together with bitpos / 8 and
4160 sign extend based on POINTER_SIZE.
4164 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4167 * config/i386/constraints.md (Yc): New register constraint.
4168 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4169 Use Yc constraint for alternative 2 of operand 0. Remove
4170 preferred_for_speed attribute.
4172 2018-03-14 Richard Biener <rguenther@suse.de>
4174 PR tree-optimization/84830
4175 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4176 with the old one to avoid oscillations.
4178 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4181 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4183 (assign_by_spills): Return a flag of reload assignment failure.
4184 Do not process the reload assignment failures. Do not spill other
4185 reload pseudos if they has the same reg class. Update n if
4187 (lra_assign): Add a return arg. Set up from the result of
4188 assign_by_spills call.
4189 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4190 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4191 usage_insns if it is not NULL.
4192 (spill_hard_reg_in_range): New function.
4193 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4194 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4195 function prototypes.
4196 (lra_assign): Change prototype.
4197 * lra.c (lra): Add code to deal with fails by splitting hard reg
4200 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4202 * config/riscv/riscv.opt (mrelax): New option.
4203 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4204 "norelax" when riscv_mrelax is disabled.
4205 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4207 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4210 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4211 reassociation for int modes.
4213 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4215 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4216 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4218 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4219 * config/aarch64/aarch64-sve.md
4220 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4221 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4222 (*extend<mode><Vwide>2): Rename to...
4223 (aarch64_sve_extend<mode><Vwide>2): ...this.
4224 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4225 renaming the old pattern to...
4226 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4228 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4229 define_expand, renaming the old pattern to...
4230 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4231 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4232 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4233 account when deciding which SVE instruction the optab should use.
4234 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4236 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4238 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4239 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4240 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4241 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4242 (tlsdesc_small_<mode>): Turn a define_expand and use
4243 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4244 (tlsdesc_small_advsimd_<mode>): ...this.
4245 (tlsdesc_small_sve_<mode>): New pattern.
4247 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4249 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4250 (UNSPEC_UMUL_HIGHPART): New constants.
4251 (MUL_HIGHPART): New int iteraor.
4252 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4253 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4255 (*<su>mul<mode>3_highpart): New define_insn.
4257 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4260 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4263 2018-03-13 Martin Liska <mliska@suse.cz>
4266 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4268 (sem_item_optimizer::~sem_item_optimizer): Release it.
4269 (sem_item_optimizer::merge_classes): Register variable aliases.
4270 (sem_item_optimizer::fixup_pt_set): New function.
4271 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4272 * ipa-icf.h: Declare new variables and functions.
4274 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4277 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4278 integer_pow2p@2 and test integer_pow2p in condition.
4279 (A < 0 ? C : 0): Similarly for @1.
4282 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4283 characters starting at p contain '\0' character, don't look beyond
4287 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4288 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4291 * reg-stack.c (change_stack): Change update_end var from int to
4292 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4293 also call set_block_for_insn on the newly added insns and rescan.
4296 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4297 on the last operand.
4300 * tree.c (stabilize_reference_1): Return save_expr (e) for
4301 STATEMENT_LIST even if it doesn't have side-effects.
4303 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4305 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4307 2018-03-12 Renlin Li <renlin.li@arm.com>
4309 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4310 aarch64_output_scalar_simd_mov_immediate.
4312 2018-03-12 Martin Sebor <msebor@redhat.com>
4314 PR tree-optimization/83456
4315 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4316 for perfectly overlapping calls to memcpy.
4317 (gimple_fold_builtin_memory_chk): Same.
4318 (gimple_fold_builtin_strcpy): Handle no-warning.
4319 (gimple_fold_builtin_stxcpy_chk): Same.
4320 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4322 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4324 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4325 parameter. Use it for SFmode.
4326 (rs6000_function_arg_advance_1): Adjust.
4327 (rs6000_function_arg): Adjust.
4328 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4330 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4332 PR rtl-optimization/84169
4333 PR rtl-optimization/84780
4334 * combine.c (can_combine_p): Check for a 2-insn combination whether
4335 the destination register is used between the two insns, too.
4337 2018-03-12 Richard Biener <rguenther@suse.de>
4339 PR tree-optimization/84803
4340 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4341 for refs DR analysis didn't process.
4343 2018-03-12 Richard Biener <rguenther@suse.de>
4345 PR tree-optimization/84777
4346 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4347 force-vectorize loops ignore whether we are optimizing for size.
4349 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4351 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4352 (TARGET_MD_ASM_ADJUST): Define.
4354 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4355 Kito Cheng <kito.cheng@gmail.com>
4356 Chung-Ju Wu <jasonwucj@gmail.com>
4358 * config/nds32/nds32.c (nds32_compute_stack_frame,
4359 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4360 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4361 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4362 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4363 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4364 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4365 * config/nds32/nds32.md (prologue, epilogue): Use macro
4366 NDS32_V3PUSH_AVAILABLE_P to do checking.
4368 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4371 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4372 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4373 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4374 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4375 addition of most attributes on !orig_type_die or the attribute not
4376 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4378 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4379 Chung-Ju Wu <jasonwucj@gmail.com>
4381 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4383 * config/nds32/nds32.opt (mvh): New option.
4385 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4386 Chung-Ju Wu <jasonwucj@gmail.com>
4388 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4390 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4391 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4394 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4395 Chung-Ju Wu <jasonwucj@gmail.com>
4397 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4399 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4400 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4402 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4403 Kito Cheng <kito.cheng@gmail.com>
4404 Chung-Ju Wu <jasonwucj@gmail.com>
4406 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4407 UNSPEC_FFMISM and UNSPEC_FLMISM.
4408 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4409 for ffb, ffmism and flmism.
4410 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4411 (unspec_ffmism): Ditto.
4412 (unspec_flmism): Ditto.
4413 (nds32_expand_builtin_impl): Check if string extension is available.
4414 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4415 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4417 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4420 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4423 * lra-assigns.c (assign_by_spills): Return a flag of reload
4424 assignment failure. Do not process the reload assignment
4425 failures. Do not spill other reload pseudos if they has the same
4427 (lra_assign): Add a return arg. Set up from the result of
4428 assign_by_spills call.
4429 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4430 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4431 usage_insns if it is not NULL.
4432 (spill_hard_reg_in_range): New function.
4433 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4434 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4435 function prototypes.
4436 (lra_assign): Change prototype.
4437 * lra.c (lra): Add code to deal with fails by splitting hard reg
4440 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4443 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4445 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4448 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4449 (dw_val_node): Add val_symbolic_view.
4450 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4451 (symview_upper_bound): New.
4452 (new_line_info_table): Initialize symviews_since_reset.
4453 (dwarf2out_source_line): Count symviews_since_reset and set
4454 symview_upper_bound.
4455 (dw_val_equal_p): Handle symview.
4456 (add_AT_symview): New.
4457 (print_dw_val): Handle symview.
4458 (attr_checksum, attr_checksum_ordered): Likewise.
4459 (same_dw_val_p, size_of_die): Likewise.
4460 (value_format, output_die): Likewise.
4461 (add_high_low_attributes): Use add_AT_symview for entry_view.
4462 (dwarf2out_finish): Reset symview_upper_bound, clear
4465 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4468 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4469 Add strict argument and use it.
4470 (rs6000_split_multireg_move): Update for new strict argument.
4471 (mem_operand_gpr): Disallow all non-offsettable addresses.
4472 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4474 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4477 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4478 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4479 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4482 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4483 decl, use remap_type if we want to use the type.
4485 2018-03-09 Martin Sebor <msebor@redhat.com>
4487 PR tree-optimization/84526
4488 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4490 (builtin_access::generic_overlap): Be prepared to handle non-array
4493 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4495 PR rtl-optimization/84682
4496 * lra-constraints.c (process_address_1): Check is_address flag
4497 for address constraints.
4498 (process_alt_operands): Likewise.
4499 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4500 preprocess_constraints.
4501 * recog.h (preprocess_constraints): Add oploc parameter.
4503 * recog.c (preprocess_constraints): Test address_operand for
4504 CT_ADDRESS constraints.
4506 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4509 * lra-assigns.c (assign_by_spills): Return a flag of reload
4510 assignment failure. Do not process the reload assignment
4511 failures. Do not spill other reload pseudos if they has the same
4513 (lra_assign): Add a return arg. Set up from the result of
4514 assign_by_spills call.
4515 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4516 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4517 usage_insns if it is not NULL.
4518 (spill_hard_reg_in_range): New function.
4519 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4520 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4521 function prototypes.
4522 (lra_assign): Change prototype.
4523 * lra.c (lra): Add code to deal with fails by splitting hard reg
4526 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4529 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4530 Accept complain bool parameter. Only emit errors if it is true.
4531 (arm_parse_cpu_option_name): Likewise.
4532 (arm_target_thumb_only): Adjust callers of the above.
4533 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4534 prototype to take a default true bool parameter.
4535 (arm_parse_arch_option_name): Likewise.
4537 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4538 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4542 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4543 * configure: Regenerate.
4544 * configure.ac ("linker --version-script option"): New.
4545 ("linker soname option"): New.
4547 2018-03-09 Richard Biener <rguenther@suse.de>
4549 PR tree-optimization/84775
4550 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4551 immediate uses of predicate stmts and mark them modified.
4554 PR tree-optimization/84178
4555 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4557 (version_loop_for_if_conversion): Delay update_ssa call.
4558 (tree_if_conversion): Delay update_ssa until after predicate
4561 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4564 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4565 when the function accesses prior frames.
4567 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4570 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4571 gen_llsym, otherwise call maybe_gen_llsym.
4574 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4575 has ',' character inside of it.
4577 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4580 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4581 as clobbering CC_REGNUM.
4583 2018-03-08 Richard Biener <rguenther@suse.de>
4586 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4587 (follow_copies_to_constant): Do not follow SSA names registered
4590 2018-03-08 Richard Biener <rguenther@suse.de>
4592 PR tree-optimization/84178
4593 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4595 (version_loop_for_if_conversion): Delay update_ssa call.
4596 (tree_if_conversion): Delay update_ssa until after predicate
4599 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4601 PR tree-optimization/84178
4602 * tree-if-conv.c (release_bb_predicate): Remove the
4603 the assertion that the stmts have NULL use_ops.
4604 Discard the statements, asserting that they haven't
4605 yet been added to a BB.
4607 2018-03-08 Richard Biener <rguenther@suse.de>
4609 PR tree-optimization/84746
4610 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4611 (phi_translate): Pass in destination ANTIC_OUT set.
4612 (phi_translate_1): Likewise. For a simplified result lookup
4613 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4614 (phi_translate_set): Adjust.
4615 (do_pre_regular_insertion): Likewise.
4616 (do_pre_partial_partial_insertion): Likewise.
4618 2018-03-08 Martin Liska <mliska@suse.cz>
4620 PR gcov-profile/84735
4621 * doc/gcov.texi: Document usage of profile files.
4622 * gcov-io.h: Document changes in the format.
4624 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4628 * dwarf2out.c (struct dw_line_info_table): Update comments for
4630 (FORCE_RESET_NEXT_VIEW): New.
4631 (FORCE_RESETTING_VIEW_P): New.
4632 (RESETTING_VIEW_P): Check for -1 too.
4633 (ZERO_VIEW_P): Likewise.
4634 (new_line_info_table): Force-reset next view.
4635 (dwarf2out_begin_function): Likewise.
4636 (dwarf2out_source_line): Simplify zero_view_p initialization.
4637 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4638 view directly. Omit view when omitting .loc at line 0.
4640 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4642 PR tree-optimization/84740
4643 * tree-switch-conversion.c (process_switch): Call build_constructors
4644 only if info.phi_count is non-zero.
4646 PR tree-optimization/84739
4647 * tree-tailcall.c (find_tail_calls): Check call arguments against
4648 DECL_ARGUMENTS (current_function_decl) rather than
4649 DECL_ARGUMENTS (func) when checking for tail recursion.
4651 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4653 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4654 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4655 Volker Reichelt's entry and add entries for people that perform
4656 GCC fuzzy testing and report numerous bugs.
4658 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4661 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4662 readonly data in sdata, if that is disabled.
4663 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4664 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4665 -mreadonly-in-sdata option.
4667 2018-03-07 Martin Sebor <msebor@redhat.com>
4669 PR tree-optimization/84468
4670 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4671 basic block when looking for nul assignment.
4673 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4676 * except.h (output_function_exception_table): Adjust prototype.
4677 * except.c (output_function_exception_table): Remove FNNAME parameter
4678 and add SECTION parameter. Ouput one part of the table at a time.
4679 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4680 the first part of the exception table and emit unwind directives.
4681 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4682 (i386_pe_seh_cold_init): Likewise.
4683 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4684 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4685 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4686 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4687 * config/i386/winnt.c: Include except.h.
4688 (struct seh_frame_state): Add reg_offset, after_prologue and
4689 in_cold_section fields.
4690 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4691 (i386_pe_seh_cold_init): New function.
4692 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4693 to seh->in_cold_section.
4694 (seh_emit_push): Record the offset of the push.
4695 (seh_emit_save): Record the offet of the save.
4696 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4697 Test seh->after_prologue to disregard the epilogue.
4698 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4699 (i386_pe_end_cold_function): New function.
4701 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4704 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4705 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4708 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4710 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4711 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4714 * multiple_target.c: Include tree-inline.h and intl.h.
4715 (expand_target_clones): Diagnose and fail if node->definition and
4716 !tree_versionable_function_p (node->decl).
4718 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4720 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4722 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4723 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4724 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4726 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4729 * combine.c (try_combine): Use reg_or_subregno instead of handling
4730 just paradoxical SUBREGs and REGs.
4732 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4734 * config/arc/arc.c (arc_finalize_pic): Remove function.
4735 (arc_must_save_register): We use single base PIC register, remove
4736 checks to save/restore the PIC register.
4737 (arc_expand_prologue): Likewise.
4738 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4740 (arc_verify_short): Likewise.
4741 (arc_attr_type): Likewise.
4742 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4743 (walk_stores): Likewise.
4744 (arc_address_cost): Make it static.
4745 (arc_verify_short): Likewise.
4746 (branch_dest): Likewise.
4747 (arc_attr_type): Likewise.
4748 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4749 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4750 (arc_final_prescan_insn): Remove inserting the nops due to
4751 hardware hazards. It is done in reorg step.
4752 (insn_length_variant_t): Remove.
4753 (insn_length_parameters_t): Likewise.
4754 (arc_insn_length_parameters): Likewise.
4755 (arc_get_insn_variants): Likewise.
4756 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4758 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4761 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4764 PR tree-optimization/84687
4765 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4767 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4769 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4771 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4772 Rename to ppc_speculation_barrier.
4773 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4774 __builtin_ppc_speculation_barrier.
4776 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4779 * combine.c (combine_simplify_rtx): Don't try to simplify if
4780 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4783 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4785 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4786 to 32 bytes when compiling for POWER9.
4788 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4791 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4792 regparm >= 3 with no arg reg available also for calls with
4793 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4796 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4798 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4800 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4803 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4805 2018-03-05 Richard Biener <rguenther@suse.de>
4807 PR tree-optimization/84486
4808 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4809 When inserting a __builtin_assume_aligned call set the LHS
4810 SSA name alignment info accordingly.
4812 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4814 PR tree-optimization/84114
4815 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4816 Avoid reassociation of FLOAT_MODE addition.
4818 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4820 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4821 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4822 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4823 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4824 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4825 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4826 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4828 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4829 __builtin_ia32_wbinvd): New builtins.
4830 (SPECIAL_ARGS2): New.
4831 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4832 (SPECIAL_ARGS2): New.
4833 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4834 (ix86_valid_target_attribute_inner_p): Ditto.
4835 (ix86_init_mmx_sse_builtins): Add special_args2.
4836 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4837 TARGET_WBNOINVD_P): New.
4838 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4839 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4840 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4841 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4842 * config/i386/pconfigintrin.h: New file.
4843 * config/i386/wbnoinvdintrin.h: Ditto.
4844 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4845 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4847 2018-03-05 Richard Biener <rguenther@suse.de>
4849 PR tree-optimization/84670
4850 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4852 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4853 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4854 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4855 doesn't grow if all successors (recursively) were visited at least
4858 2018-03-05 Richard Biener <rguenther@suse.de>
4860 PR tree-optimization/84650
4861 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4862 if executed in the loop pipeline.
4864 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4866 * doc/configfiles.texi (Configuration Files): Move info about
4867 conditionalizing $target-protos.h to...
4868 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4869 differs from $target-protos.h.
4871 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4872 Chung-Ju Wu <jasonwucj@gmail.com>
4874 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4875 * config/nds32/nds32-multiple.md (setmemsi): Define.
4876 * config/nds32/nds32-memory-manipulation.c
4877 (nds32_gen_dup_4_byte_to_word_value): New.
4878 (emit_setmem_word_loop): New.
4879 (emit_setmem_byte_loop): New.
4880 (nds32_expand_setmem_loop): New.
4881 (nds32_expand_setmem_loop_v3m): New.
4882 (nds32_expand_setmem_unroll): New.
4883 (nds32_expand_setmem): New.
4885 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4886 Chung-Ju Wu <jasonwucj@gmail.com>
4888 * config/nds32/nds32-memory-manipulation.c
4889 (nds32_emit_load_store): New.
4890 (nds32_emit_post_inc_load_store): New.
4891 (nds32_emit_mem_move): New.
4892 (nds32_emit_mem_move_block): New.
4893 (nds32_expand_movmemsi_loop_unknown_size): New.
4894 (nds32_expand_movmemsi_loop_known_size): New.
4895 (nds32_expand_movmemsi_loop): New.
4896 (nds32_expand_movmemsi_unroll): New.
4897 (nds32_expand_movmemqi): Rename ...
4898 (nds32_expand_movmemsi): ... to this.
4899 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4900 (movmemsi): ... to this.
4901 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4902 (nds32_expand_movmemsi): ... to this.
4904 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4905 Monk Chiang <sh.chiang04@gmail.com>
4906 Chung-Ju Wu <jasonwucj@gmail.com>
4908 * config/nds32/nds32-protos.h
4909 (nds32_expand_load_multiple): New arguments.
4910 (nds32_expand_store_multiple): Ditto.
4911 (nds32_valid_multiple_load_store): Rename ...
4912 (nds32_valid_multiple_load_store_p): ... to this.
4913 * config/nds32/nds32-memory-manipulation.c
4914 (nds32_expand_load_multiple): Refine implementation.
4915 (nds32_expand_store_multiple): Ditto.
4916 * config/nds32/nds32-multiple.md
4917 (load_multiple): Update nds32_expand_load_multiple interface.
4918 (store_multiple): Update nds32_expand_store_multiple interface.
4919 * config/nds32/nds32-predicates.c
4920 (nds32_valid_multiple_load_store): Rename ...
4921 (nds32_valid_multiple_load_store_p): ... to this and refine
4923 * config/nds32/predicates.md
4924 (nds32_load_multiple_and_update_address_operation): New predicate.
4925 (nds32_store_multiple_and_update_address_operation): New predicate.
4927 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4928 Chung-Ju Wu <jasonwucj@gmail.com>
4930 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4931 (combo): New attribute.
4932 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4934 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4936 * config/nds32/nds32.opt: Change -mcmodel= default value.
4938 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4939 Monk Chiang <sh.chiang04@gmail.com>
4940 Chung-Ju Wu <jasonwucj@gmail.com>
4942 * config/nds32/constants.md (unspec_element): New enum.
4943 * config/nds32/constraints.md (Umw): New constraint.
4944 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4945 * config/nds32/nds32-intrinsic.md: Likewise.
4946 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4947 (nds32_valid_smw_lwm_base_p): New.
4948 (nds32_output_smw_single_word): New.
4949 (nds32_output_lmw_single_word): New.
4950 (nds32_expand_unaligned_load): New.
4951 (nds32_expand_unaligned_store): New.
4952 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4953 (nds32_output_smw_single_word): Declare.
4954 (nds32_output_lmw_single_word): Declare.
4955 (nds32_expand_unaligned_load): Declare.
4956 (nds32_expand_unaligned_store): Declare.
4957 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4958 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4959 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4960 NDS32_BUILTIN_UASTORE_DW.
4961 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4964 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4965 Kito Cheng <kito.cheng@gmail.com>
4966 Chung-Ju Wu <jasonwucj@gmail.com>
4968 * config/nds32/nds32-intrinsic.c
4969 (nds32_expand_builtin_null_ftype_reg): Delete.
4970 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4971 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4972 (nds32_read_argument): New.
4973 (nds32_legitimize_target): Ditto.
4974 (nds32_legitimize_argument): Ditto.
4975 (nds32_check_constant_argument): Ditto.
4976 (nds32_expand_unop_builtin): Ditto.
4977 (nds32_expand_unopimm_builtin): Ditto.
4978 (nds32_expand_binop_builtin): Ditto.
4979 (nds32_builtin_decl_impl): Ditto.
4980 (builtin_description): Ditto.
4981 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4982 (nds32_init_builtins_impl): Ditto.
4983 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4984 (nds32_builtin_decl): New.
4985 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4986 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4988 2018-03-02 Jeff Law <law@redhat.com>
4990 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4991 (redundant_insn, fill_simple_delay_slots): Likewise.
4992 (fill_slots_from_thread): Likewise.
4993 * resource.c (mark_referenced_resources): Likewise.
4994 (mark_set_resources, find_dead_or_set_registers): Likewise.
4996 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4998 * substring-locations.h (format_warning_va): Formatting fix for
5000 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5002 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5003 * substring-locations.c: Include intl.h.
5004 (format_warning_va): Turned into small wrapper around
5005 format_warning_n_va, renamed to ...
5006 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5007 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5009 (format_warning_at_substring_n): New function.
5010 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5011 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5012 format_warning_at_substring with just a shorter name instead of
5013 const function pointer.
5014 (fmtwarn_n): New function.
5015 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5016 appropriate, get rid of all the fmtstr temporaries, move conditionals
5017 with G_() wrapped string literals directly into fmtwarn arguments,
5018 cast dir.len to (int), formatting fixes.
5020 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5022 * doc/invoke.texi: Remove "Cilk Plus" references.
5024 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5025 Richard Biener <rguenther@suse.de>
5028 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5029 for error or warning attributes if CALL_FROM_THUNK_P is set.
5032 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5035 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5036 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5039 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5040 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5042 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5043 instead of -1U in last predictors element's probability member.
5045 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5048 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5049 arguments if they are comparable.
5051 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5053 PR tree-optimization/84634
5054 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5055 masks and masked_loop_p with a single loop_masks, making sure it's
5056 null for bb vectorization.
5058 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5060 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5061 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5062 loop->force_vectorize to check whether there is no alias.
5064 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5067 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5069 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5071 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5072 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5073 instead of a loop around prev_real_insn.
5074 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5078 * config/i386/i386.c (ix86_print_operand): Use conditional
5079 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5082 2018-03-02 Richard Biener <rguenther@suse.de>
5084 PR tree-optimization/84427
5085 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5086 (bitmap_set_subtract_values): Rewrite to handle multiple
5089 (prune_clobbered_mems): Likewise.
5090 (phi_translate): Take edge instead of pred/phiblock.
5091 (phi_translate_1): Likewise.
5092 (phi_translate_set): Likewise. Insert all translated
5093 exprs for a value into the set, keeping possibly multiple
5094 expressions per value.
5095 (compute_antic_aux): Adjust for phi_translate changes.
5096 When intersecting union the expressions and prune those
5097 not in the final value set, keeping possibly multiple
5098 expressions per value. Do not use value-insertion
5099 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5100 all expressions. Add verification that the value-sets
5101 only shrink during iteration.
5102 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5103 (do_pre_regular_insertion): Likewise.
5104 (do_pre_partial_partial_insertion): Likewise.
5106 2018-03-02 Richard Biener <rguenther@suse.de>
5109 * config/darwin.c (saved_debug_info_level): New static global.
5110 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5111 (darwin_asm_lto_end): Restore debug info generation settings.
5113 2018-03-01 Martin Liska <mliska@suse.cz>
5116 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5119 2018-03-01 Richard Biener <rguenther@suse.de>
5122 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5123 type annotation with LTO when debug was disabled at compile-time.
5125 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5127 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5129 (mips_final_postscan_insn): Likewise.
5131 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5133 PR rtl-optimization/84528
5134 * alias.c (init_alias_target): Add commentary.
5135 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5136 a unique base value if the frame pointer is not eliminated
5137 to the stack pointer.
5139 2018-03-01 Tom de Vries <tom@codesourcery.com>
5141 PR rtl-optimization/83327
5142 * lra-int.h (hard_regs_spilled_into): Declare.
5143 * lra.c (hard_regs_spilled_into): Define.
5144 (init_reg_info): Init hard_regs_spilled_into.
5145 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5146 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5147 (process_bb_lives): Handle hard_regs_spilled_into.
5148 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5149 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5151 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5153 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5154 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5155 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5156 * config/rs6000/aix72.h: New file.
5158 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5160 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5161 instead of warning_at with conditional singular and plural messages
5165 * stor-layout.c (update_alignment_for_field): For
5166 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5167 && !DECL_PACKED (field), do the alignment update, just use
5168 only desired_align instead of MAX (type_align, desired_align)
5170 (place_field): Don't do known_align < desired_align handling
5171 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5172 is non-NULL, instead do it after rli->prev_field handling and
5173 only if not within a bitfield word. For DECL_PACKED (field)
5174 use type_align of BITS_PER_UNIT.
5176 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5178 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5179 superfluous parentheses and trailing spaces.
5181 2018-02-28 Richard Biener <rguenther@suse.de>
5183 PR tree-optimization/84584
5184 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5185 SCoPs with fake exit edge.
5187 2018-02-28 Martin Liska <mliska@suse.cz>
5190 * timevar.c (timer::print): Fix format to properly print 100%
5193 2018-02-28 Richard Biener <rguenther@suse.de>
5196 * genmatch.c (capture_info::walk_match): Do not mark
5197 captured expressions without operands as expr_p given
5198 they act more like predicates and should be subject to
5199 "lost tail" side-effect preserving.
5201 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5203 PR rtl-optimization/81611
5204 * auto-inc-dec.c (attempt_change): Move dead note from
5205 mem_insn if it's the next use of regno
5206 (find_address): Take address use of reg holding
5207 non-incremented value. Add parm to limit search to the named
5209 (merge_in_block): Attempt to use a mem insn that is the next
5210 use of the original regno.
5212 2018-02-27 Martin Sebor <msebor@redhat.com>
5215 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5216 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5218 2018-02-27 Martin Sebor <msebor@redhat.com>
5220 PR translation/84207
5221 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5222 n argument to unsigned HOST_WIDE_INT.
5223 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5224 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5225 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5226 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5228 2018-02-27 Richard Biener <rguenther@suse.de>
5230 PR tree-optimization/84512
5231 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5232 Do not use the estimate returned from record_stmt_cost for
5233 the scalar iteration cost but sum properly using add_stmt_cost.
5235 2018-02-27 Richard Biener <rguenther@suse.de>
5237 PR tree-optimization/84466
5238 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5239 Adjust last change to less strictly validate use operands.
5241 2018-02-27 Martin Liska <mliska@suse.cz>
5243 PR gcov-profile/84548
5244 * gcov.c (process_file): Allow partial overlap and consider it
5245 also as group functions.
5246 (output_lines): Properly calculate range of lines for a group.
5248 2018-02-27 Martin Liska <mliska@suse.cz>
5250 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5251 'ggc' suffixes. Change first column width.
5252 (timer::print): Fix formatting of the column.
5254 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5256 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5257 preserve inline entry blocks for the sake of debug inline
5258 entry point markers alone.
5259 (remove_unused_locals): Suggest in comments a better place to
5260 force the preservation of inline entry blocks that are
5261 otherwise unused, but do not preserve them.
5263 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5265 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5267 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5270 * config/i386/constraints.md (Bs): Replace
5271 ix86_indirect_branch_register with
5272 TARGET_INDIRECT_BRANCH_REGISTER.
5274 * config/i386/i386.md (indirect_jump): Likewise.
5275 (tablejump): Likewise.
5276 (*sibcall_memory): Likewise.
5277 (*sibcall_value_memory): Likewise.
5278 Peepholes of indirect call and jump via memory: Likewise.
5279 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5280 (*sibcall_value_GOT_32): Likewise.
5281 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5282 (GOT_memory_operand): Likewise.
5283 (call_insn_operand): Likewise.
5284 (sibcall_insn_operand): Likewise.
5285 (GOT32_symbol_operand): Likewise.
5286 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5288 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5290 PR rtl-optimization/83496
5291 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5292 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5293 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5294 redundant insn, if any.
5295 (relax_delay_slots): Likewise.
5296 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5298 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5300 PR tree-optimization/83965
5301 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5302 that grouped statements are part of a reduction chain. Return
5303 true if the statement is not marked as a reduction itself but
5305 (vect_recog_dot_prod_pattern): Don't check whether the statement
5306 is part of a group here.
5307 (vect_recog_sad_pattern): Likewise.
5308 (vect_recog_widen_sum_pattern): Likewise.
5310 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5313 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5315 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5318 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5320 (ix86_output_indirect_function_return): New prototype.
5321 (ix86_split_simple_return_pop_internal): Likewise.
5322 * config/i386/i386.c (indirect_return_via_cx): New.
5323 (indirect_return_via_cx_bnd): Likewise.
5324 (indirect_thunk_name): Handle return va CX_REG.
5325 (output_indirect_thunk_function): Create alias for
5326 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5327 (ix86_output_indirect_jmp): Remove the bool argument.
5328 (ix86_output_indirect_function_return): New function.
5329 (ix86_split_simple_return_pop_internal): Likewise.
5330 * config/i386/i386.md (*indirect_jump): Don't pass false
5331 to ix86_output_indirect_jmp.
5332 (*tablejump_1): Likewise.
5333 (simple_return_pop_internal): Change it to define_insn_and_split.
5334 Call ix86_split_simple_return_pop_internal to split it for
5336 (simple_return_indirect_internal): Call
5337 ix86_output_indirect_function_return instead of
5338 ix86_output_indirect_jmp.
5340 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5343 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5344 memset and value initialization afterwards.
5346 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5348 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5350 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5353 * common/config/aarch64/aarch64-common.c
5354 (aarch_option_optimization_table[]): Switch
5355 off fomit-frame-pointer
5357 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5358 Chung-Ju Wu <jasonwucj@gmail.com>
5360 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5362 (store_multiple): Ditto.
5364 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5366 * config.gcc: Add --with-cpu support for nds32 target.
5367 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5368 * config/nds32/nds32.opt: Add -mcpu= option.
5370 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5372 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5373 isel=yes): Warn for these deprecated options.
5375 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5377 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5378 ISA_2_5_MASKS_EMBEDDED.
5380 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5382 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5383 p->max as pointers rather than using iterative_hash_expr.
5385 2018-02-23 Carl Love <cel@us.ibm.com>
5387 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5388 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5390 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5391 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5392 P8V_BUILTIN_VEC_VUNSIGNED2.
5394 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5397 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5398 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5400 * lra-constraints.c (curr_insn_transform): Set up
5401 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5403 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5404 * lra-eliminations.c (spill_pseudos): Ditto.
5405 (process_insn_for_elimination): Ditto.
5406 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5407 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5408 LRA_NON_CLOBBERED_ALT.
5410 2018-02-22 Martin Sebor <msebor@redhat.com>
5412 PR tree-optimization/84480
5413 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5414 to maybe_diag_stxncpy_trunc. Call it.
5415 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5416 from gimple_fold_builtin_strcpy. Print inlining stack.
5417 (handle_builtin_stxncpy): Print inlining stack.
5418 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5420 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5423 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5424 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5425 and -fcheck-pointer-bounds are used together.
5426 (indirect_thunk_prefix): New enum.
5427 (indirect_thunk_need_prefix): New function.
5428 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5429 "_nt" instead of "_bnd" for NOTRACK prefix.
5430 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5431 (output_indirect_thunk_function): Likewise.
5433 (ix86_code_end): Update output_indirect_thunk_function calls.
5434 (ix86_output_indirect_branch_via_reg): Replace
5435 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5436 (ix86_output_indirect_branch_via_push): Likewise.
5437 (ix86_output_function_return): Likewise.
5438 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5439 incompatible with -fcf-protection=branch and
5440 -fcheck-pointer-bounds.
5442 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5445 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5446 Change gcc_assert call to output_operand_lossage.
5448 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5450 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5452 2018-02-22 DJ Delorie <dj@redhat.com>
5453 Sebastian Perta <sebastian.perta@renesas.com>
5454 Oleg Endo <olegendo@gcc.gnu.org>
5456 * config/rx/rx.c (rx_rtx_costs): New function.
5457 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5459 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5461 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5463 2018-02-22 Martin Liska <mliska@suse.cz>
5466 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5467 Add "native" as a possible value.
5469 2018-02-22 Martin Liska <mliska@suse.cz>
5472 * config/i386/i386.c (ix86_option_override_internal):
5473 Add "native" as a possible value for -march and -mtune.
5475 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5478 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5479 to all type variants.
5481 PR tree-optimization/84503
5482 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5483 width as info->bitpos + info->bitsize - start.
5484 (merged_store_group::merge_overlapping): Simplify width computation.
5485 (check_no_overlap): New function.
5486 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5487 start + width and last_order of the group, fail if check_no_overlap
5489 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5490 to group if check_no_overlap fails.
5492 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5494 * config/rs6000/altivec.md: Delete contraint arguments to
5495 define_expand, define_split, and define_peephole2, and in
5496 define_insn_and_split if always unused.
5497 * config/rs6000/darwin.md: Ditto.
5498 * config/rs6000/dfp.md: Ditto.
5499 * config/rs6000/rs6000.md: Ditto.
5500 * config/rs6000/sync.md: Ditto.
5501 * config/rs6000/vector.md: Ditto.
5502 * config/rs6000/vsx.md: Ditto.
5504 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5506 * config/rs6000/altivec.md: Write output control strings as braced
5507 blocks instead of double-quoted strings.
5508 * config/rs6000/darwin.md: Ditto.
5509 * config/rs6000/rs6000.md: Ditto.
5510 * config/rs6000/vector.md: Ditto.
5511 * config/rs6000/vsx.md: Ditto.
5513 2018-02-21 Jason Merrill <jason@redhat.com>
5515 PR c++/84314 - ICE with templates and fastcall attribute.
5516 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5518 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5520 * ipa-cp.c (determine_versionability): Fix comment typos.
5522 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5525 * ipa-cp.c (determine_versionability): Do not version functions caling
5528 2018-02-21 Martin Liska <mliska@suse.cz>
5531 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5532 Add "native" as a possible value.
5533 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5534 the macro when native cpu detection is available.
5536 2018-02-21 Martin Liska <mliska@suse.cz>
5539 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5540 Add "native" as a possible value.
5541 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5542 when native cpu detection is available.
5544 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5545 Martin Sebor <msebor@redhat.com>
5547 PR tree-optimization/84478
5548 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5550 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5551 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5552 support which is conservatively correct, for 2 only stay conservative
5553 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5554 argument to the 2 argument get_range_strlen, adjust 6 arg
5555 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5557 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5558 (gimple_fold_builtin_strlen): Pass true as last argument to
5561 2018-02-20 Martin Sebor <msebor@redhat.com>
5564 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5565 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5566 (builtin_memref::builtin_memref): Factor out parts into
5567 set_base_and_offset and call it.
5569 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5572 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5573 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5574 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5575 search at the associated MODE_INT.
5577 2018-02-20 Jeff Law <law@redhat.com>
5580 PR tree-optimization/81592
5582 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5583 for range data rather than using global data.
5584 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5585 range data rather than using global data.
5586 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5587 pass it to children as needed.
5588 (struct directive::fmtresult): Similarly.
5589 (struct directive::set_width): Similarly.
5590 (struct directive::set_precision): Similarly.
5591 (format_integer, format_directive, parse_directive): Similarly.
5592 (format_none): Accept unnamed vr_values parameter.
5593 (format_percent, format_floating, format_character): Similarly.
5594 (format_string, format_plain): Similarly.
5595 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5596 the EVRP range analyzer for range data rather than using global data.
5597 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5598 gimple-ssa-evrp-analyze.h
5599 (class sprintf_dom_walker): Add after_dom_children member function.
5600 Add evrp_range_analyzer member.
5601 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5602 range analyzer as needed.
5603 (sprintf_dom_walker::after_dom_children): New member function.
5604 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5606 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5607 (evrp_range_analyzer::pop_to_marker): Likewise.
5609 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5611 PR tree-optimization/84419
5612 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5613 with the required type if its current type is compatible but
5616 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5619 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5620 after vectorization.
5622 2018-02-20 Martin Liska <mliska@suse.cz>
5625 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5626 possible values if we don't have a hint.
5628 2018-02-20 Martin Liska <mliska@suse.cz>
5632 * final.c (shorten_branches): Build align_tab array with one
5634 * opts.c (finish_options): Add alignment option limit check.
5635 (MAX_CODE_ALIGN): Likewise.
5636 (MAX_CODE_ALIGN_VALUE): Likewise.
5637 * doc/invoke.texi: Document maximum allowed option value for
5638 all -falign-* options.
5640 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5643 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5644 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5645 * var-tracking.c (emit_note_insn_var_location): Remove all references
5646 to NOTE_INSN_CALL_ARG_LOCATION.
5647 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5648 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5649 Use copy_rtx_if_shared.
5650 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5651 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5652 (dwarf2out_var_location): Remove handling of
5653 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5655 * final.c (final_scan_insn): Remove all references to
5656 NOTE_INSN_CALL_ARG_LOCATION.
5657 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5658 before dumping final insns.
5659 * except.c (emit_note_eh_region_end): Remove all references to
5660 NOTE_INSN_CALL_ARG_LOCATION.
5661 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5662 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5663 * config/arc/arc.c (hwloop_optimize): Likewise.
5664 * config/arm/arm.c (create_fix_barrier): Likewise.
5665 * config/s390/s390.c (s390_chunkify_start): Likewise.
5666 * config/sh/sh.c (find_barrier): Likewise.
5667 * config/i386/i386.c (rest_of_insert_endbranch,
5668 ix86_seh_fixup_eh_fallthru): Likewise.
5669 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5670 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5671 * config/frv/frv.c (frv_function_prologue): Likewise.
5672 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5674 (note_outside_basic_block_p): Remove all references to
5675 NOTE_INSN_CALL_ARG_LOCATION.
5676 * gengtype.c (adjust_field_rtx_def): Likewise.
5677 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5679 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5680 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5683 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5686 PR tree-optimization/84452
5687 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5688 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5691 2018-02-19 Martin Liska <mliska@suse.cz>
5694 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5696 2018-02-19 Martin Liska <mliska@suse.cz>
5697 Richard Sandiford <richard.sandiford@linaro.org>
5699 PR tree-optimization/82491
5700 * gimple-fold.c (get_base_constructor): Make earlier bail out
5703 2018-02-19 Carl Love <cel@us.ibm.com>
5705 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5706 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5708 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5709 P8V_BUILTIN_VEC_NEG.
5711 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5713 * config/rl78/rl78.md (movdf): New define expand.
5715 2018-02-19 Martin Liska <mliska@suse.cz>
5718 * doc/invoke.texi: Fix typo.
5719 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5721 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5723 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5724 handle rs6000_single_float and rs6000_double_float specially for
5727 2018-02-16 Jeff Law <law@redhat.com>
5729 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5732 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5735 * ipa-inline.c (inline_small_functions): Fix a typo.
5737 2018-02-16 Nathan Sidwell <nathan@acm.org>
5739 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5741 2018-02-16 Carl Love <cel@us.ibm.com>
5743 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5744 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5745 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5746 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5747 expansion to P8V_BUILTIN_VEC_FLOAT2.
5749 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5751 PR rtl-optimization/70023
5752 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5753 src_regno into account.
5755 2018-02-16 Carl Love <cel@us.ibm.com>
5757 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5758 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5759 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5760 * config/rs6000/rs6000.c: Remove case statements for
5761 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5762 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5763 and P9V_BUILTIN_VEC_VINSERT4B.
5764 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5765 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5766 * config/rs6000/vsx.md:
5767 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5770 2018-02-16 Carl Love <cel@us.ibm.com>
5772 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5774 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5776 * config/rs6000/rs6000-c.c: Add the definitions for
5777 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5778 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5779 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5780 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5781 definition for insert4b and define insn *insert3b_internal.
5782 * doc/extend.texi: Add documentation for vec_extract4b.
5784 2018-02-16 Nathan Sidwell <nathan@acm.org>
5786 * doc/extend.texi (Backwards Compatibility): Mention friend
5787 injection. Note for-scope is deprecated.
5788 * doc/invoke.texi (-ffriend-injection): Deprecate.
5790 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5792 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5793 that moved to I2, also allow destinations that are a paradoxical
5794 subreg (instead of a normal reg).
5796 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5799 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5802 2018-02-16 Richard Biener <rguenther@suse.de>
5804 PR tree-optimization/84037
5805 PR tree-optimization/84016
5807 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5808 Adjust vec_construct for the fact we need additional higher latency
5809 128bit inserts for AVX256 and AVX512 vector builds.
5810 (ix86_add_stmt_cost): Scale vector construction cost for
5813 2018-02-16 Richard Biener <rguenther@suse.de>
5815 PR tree-optimization/84417
5816 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5817 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5818 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5820 2018-02-16 Martin Liska <mliska@suse.cz>
5823 * internal-fn.def (ASAN_CHECK): Set proper flags.
5824 (ASAN_MARK): Likewise.
5826 2018-02-16 Julia Koval <julia.koval@intel.com>
5828 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5829 from PTA_CANNONLAKE.
5831 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5834 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5835 Use ++iter rather than iter++ for std::list iterators.
5836 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5837 defer deleting them until all nodes in the forest are processed. Do
5838 free even leaf nodes. Change to_process into auto_vec.
5841 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5842 * vec.h (vec_default_construct): Use memset instead of placement new
5843 if BROKEN_VALUE_INITIALIZATION is defined.
5844 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5845 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5848 PR rtl-optimization/83723
5849 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5850 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5851 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5853 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5855 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5857 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5859 PR rtl-optimization/81443
5860 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5861 from inner REGs to paradoxical SUBREGs.
5863 2018-02-16 Richard Biener <rguenther@suse.de>
5865 PR tree-optimization/84399
5866 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5867 For operands we can analyze at their definition make sure we can
5868 analyze them at each use as well.
5870 2018-02-16 Richard Biener <rguenther@suse.de>
5872 PR tree-optimization/84190
5873 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5874 volatile accesses if the decl isn't volatile.
5876 2018-02-15 Jason Merrill <jason@redhat.com>
5878 PR c++/84314 - ICE with templates and fastcall attribute.
5879 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5880 TYPE_CANONICAL on an existing type.
5882 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5884 PR tree-optimization/84383
5885 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5886 dstoff nor call operand_equal_p if dstbase is NULL.
5888 PR tree-optimization/84334
5889 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5890 also a CONSTANT_CLASS_P, punt.
5892 2018-02-14 Jim Wilson <jimw@sifive.com>
5894 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5895 first SMALL_OPERAND check. New local min_second_step. Move assert
5896 to where locals are set. Add TARGET_RVC support.
5897 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5899 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5901 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5903 2018-02-14 Martin Sebor <msebor@redhat.com>
5905 PR tree-optimization/83698
5906 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5907 arrays constrain the offset range to their bounds.
5908 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5909 (builtin_access::overlap): Avoid setting the size of overlap if it's
5911 (maybe_diag_overlap): Also consider arrays when deciding what values
5912 of offsets to include in diagnostics.
5914 2018-02-14 Martin Sebor <msebor@redhat.com>
5917 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5918 that correspond to the kind of a declaration.
5920 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5923 * config/pa/pa.md: Load address of PIC label using the linkage table
5924 if the label is nonlocal.
5926 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5928 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5929 warning message if user requests -maltivec=be.
5930 * doc/invoke.texi: Document deprecation of -maltivec=be.
5932 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5935 * config/rs6000/rs6000-c.c: Update definitions for
5936 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5937 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5939 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5942 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5943 add _get_ssp intrinsics. Remove argument from
5944 __builtin_ia32_rdssp[d|q].
5945 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5946 * config/i386/i386-builtin.def: Remove argument from
5947 __builtin_ia32_rdssp[d|q].
5948 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5949 ix86_expand_special_args_builtin for _rdssp[d|q].
5950 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5951 Clear register before usage.
5952 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5953 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5955 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5957 PR tree-optimization/84357
5958 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5959 operand 1 of an ARRAY_REF too.
5961 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5964 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5965 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5967 (set_of_reg): New struct.
5968 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5969 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5970 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5972 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5973 Split into bitclr, bitset, bitinvert patterns if appropriate.
5974 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5975 use rx_fuse_in_memory_bitop.
5976 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5977 to named insn, correct maximum insn length.
5979 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5982 * machmode.def: Define a complex mode for PARTIAL_INT.
5983 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5985 * doc/rtl.texi: Document CSPImode.
5986 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5988 (msp430_hard_regno_nregs_with_padding): Likewise.
5990 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5993 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5995 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5997 PR rtl-optimization/84169
5998 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5999 we generated a parallel as new i3 and we split that to new i2 and i3
6000 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6001 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6002 those to i2, not i1. Partially rewrite this scan code.
6004 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6007 * stor-layout.c (place_field): For variable length fields, adjust
6008 offset_align afterwards not just based on the field's alignment,
6009 but also on the size.
6012 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6013 of exps and logs in the use_exp2 case.
6015 2018-02-13 Jeff Law <law@redhat.com>
6017 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6020 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6023 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6027 * common.opt (gas-loc-support, gas-locview-support): New.
6028 (ginline-points, ginternal-reset-location-views): New.
6029 * doc/invoke.texi: Document them. Use @itemx where intended.
6030 (gvariable-location-views): Adjust.
6031 * target.def (reset_location_view): New.
6032 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6033 (TARGET_RESET_LOCATION_VIEW): New.
6034 * doc/tm.texi: Rebuilt.
6035 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6036 (dwarf2out_default_as_locview_support): New.
6037 (output_asm_line_debug_info): Use option variables.
6038 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6039 (output_loc_list): Likewise.
6040 (add_high_low_attributes): Check option variables.
6041 Don't output entry view attribute in strict mode.
6042 (gen_inlined_subroutine_die): Check option variables.
6043 (dwarf2out_inline_entry): Likewise.
6044 (init_sections_and_labels): Likewise.
6045 (dwarf2out_early_finish): Likewise.
6046 (maybe_reset_location_view): New, from...
6047 (dwarf2out_var_location): ... here. Call it.
6048 * debug.h (dwarf2out_default_as_loc_support): Declare.
6049 (dwarf2out_default_as_locview_support): Declare.
6050 * hooks.c (hook_int_rtx_insn_0): New.
6051 * hooks.h (hook_int_rtx_insn_0): Declare.
6052 * toplev.c (process_options): Take -gas-loc-support and
6053 -gas-locview-support from dwarf2out. Enable
6054 -gvariable-location-views by default only with locview
6055 assembler support. Enable -ginternal-reset-location-views by
6056 default only if the target defines the corresponding hook.
6057 Enable -ginline-points by default if location views are
6058 enabled; force it disabled if statement frontiers are
6060 * tree-inline.c (expand_call_inline): Check option variables.
6061 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6063 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6065 PR tree-optimization/84321
6066 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6067 handling. Also check whether the anti-range contains any values
6068 that satisfy the mask; switch to a VR_RANGE if not.
6070 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6073 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6075 2018-02-13 Martin Jambor <mjambor@suse.cz>
6078 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6079 of call statements, also set location of a load to a temporary.
6081 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6083 * config/rl78/rl78.c (add_vector_labels): New function.
6084 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6085 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6086 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6087 which checks that no arguments are passed.
6088 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6089 * doc/extend.texi: Documentation for the new attribute.
6091 2018-02-13 Andreas Schwab <schwab@suse.de>
6093 * config/riscv/linux.h (CPP_SPEC): Define.
6095 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6098 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6099 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6100 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6101 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6102 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6103 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6104 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6105 temporarily for AES and PCLMUL builtins.
6107 PR tree-optimization/84339
6108 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6109 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6113 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6114 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6115 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6117 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6119 * omp-simd-clone.h: New file.
6120 * omp-simd-clone.c: Include omp-simd-clone.h.
6121 (expand_simd_clones): No longer static.
6122 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6123 cgraph.h and omp-simd-clone.h.
6124 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6125 (vect_recog_widen_shift_pattern): Formatting fix.
6126 (vect_pattern_recog_1): Don't check optab for calls.
6129 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6130 operands[2] into a REG before using gen_lowpart on it.
6132 2018-02-12 Jeff Law <law@redhat.com>
6135 * config/sh/sh.c (find_barrier): Consider a sibling call
6138 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6139 successfully back substituting a reg.
6141 2018-02-12 Richard Biener <rguenther@suse.de>
6143 PR tree-optimization/84037
6144 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6145 parameter, move visited init to caller.
6146 (vect_slp_analyze_operations): Separate cost from validity
6147 check, initialize visited once for all instances.
6148 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6150 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6152 (vect_model_promotion_demotion_cost): Likewise.
6153 (vectorizable_bswap): Guard cost modeling with !slp_node
6154 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6156 (vectorizable_call): Likewise.
6157 (vectorizable_conversion): Likewise.
6158 (vectorizable_assignment): Likewise.
6159 (vectorizable_shift): Likewise.
6160 (vectorizable_operation): Likewise.
6161 (vectorizable_store): Likewise.
6162 (vectorizable_load): Likewise.
6163 (vectorizable_condition): Likewise.
6164 (vectorizable_comparison): Likewise.
6166 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6169 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6170 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6171 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6173 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6176 * params.def (inline-min-speedup): Increase from 8 to 15.
6177 (max-inline-insns-auto): Decrease from 40 to 30.
6178 * ipa-split.c (consider_split): Add some buffer for function to
6179 be considered inlining candidate.
6180 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6183 2018-02-12 Richard Biener <rguenther@suse.de>
6185 PR tree-optimization/84037
6186 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6187 matched stmts if we cannot swap the non-matched ones.
6189 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6191 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6192 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6193 _mm_maskz_scalef_round_ss): New intrinsics.
6194 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6195 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6196 __builtin_ia32_scalefss_round): Remove.
6197 (__builtin_ia32_scalefsd_mask_round,
6198 __builtin_ia32_scalefss_mask_round): New intrinsics.
6199 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6200 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6201 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6202 "<round_constraint>")): Changed to ...
6203 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6204 "<round_scalar_constraint>")): ... this.
6205 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6206 %0, %1, %2<round_op3>}"): Changed to ...
6207 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6208 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6209 %2<round_scalar_mask_op3>}"): ... this.
6210 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6212 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6214 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6215 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6216 (_mm_maskz_sqrt_round_ss): New intrinsics.
6217 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6218 (__builtin_ia32_sqrtsd_mask_round)
6219 (__builtin_ia32_sqrtss_mask_round): New builtins.
6220 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6221 (__builtin_ia32_sqrtss_round): Remove.
6222 (__builtin_ia32_sqrtsd_mask_round)
6223 (__builtin_ia32_sqrtss_mask_round): New builtins.
6224 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6225 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6226 ((match_operand:VF_128 1 "vector_operand"
6227 "xBm,<round_constraint>")): Changed to ...
6228 ((match_operand:VF_128 1 "vector_operand"
6229 "xBm,<round_scalar_constraint>")): ... this.
6230 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6231 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6232 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6233 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6234 %<iptr>1<round_scalar_mask_op3>}): ... this.
6235 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6236 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6238 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6241 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6242 Cast vec_cmpeq result to correct type.
6243 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6244 Cast vec_cmpgt result to correct type.
6246 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6248 * final.c (final_scan_insn_1): Renamed from...
6249 (final_scan_insn): ... this. New wrapper, to recover
6250 seen from the outermost call in recursive ones.
6251 * config/sparc/sparc.c (output_return): Drop seen from call.
6252 (output_sibcall): Likewise.
6253 * config/visium/visium.c (output_branch): Likewise.
6255 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6257 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6260 2018-02-10 Alan Modra <amodra@gmail.com>
6263 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6264 Specify LR as an input.
6266 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6269 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6270 remove_member_access_dummy_vars): New functions.
6271 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6272 lower_omp_1, execute_lower_omp): Use them.
6274 PR rtl-optimization/84308
6275 * shrink-wrap.c (spread_components): Release todo vector.
6277 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6279 PR rtl-optimization/57193
6280 * ira-color.c (struct allocno_color_data): Add member
6281 conflict_allocno_hard_prefs.
6282 (update_conflict_allocno_hard_prefs): New.
6283 (bucket_allocno_compare_func): Add a preference based on
6284 conflict_allocno_hard_prefs.
6285 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6286 (color_allocnos): Remove a dead code. Initiate
6287 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6289 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6292 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6293 constraint from =wa to wa. Avoid a subreg on the output operand,
6294 instead use a pseudo and subreg it in a move.
6295 (p9_xxbrd_<mode>): Changed to ...
6296 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6297 (p9_xxbrd_v2df): New expander.
6298 (p9_xxbrw_<mode>): Changed to ...
6299 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6300 (p9_xxbrw_v4sf): New expander.
6302 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6304 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6306 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6309 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6310 multiply in 32-bit mode.
6311 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6312 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6315 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6317 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6318 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6319 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6320 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6322 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6325 * dwarf2out.c (is_trivial_indirect_ref): New function.
6326 (dwarf2out_late_global_decl): Do not generate a location
6327 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6328 and that are not defined in the current unit.
6330 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6332 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6333 instead of a libcall for UNORDERED.
6335 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6338 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6339 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6341 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6344 * config/s390/s390.c (s390_set_current_function): Invoke
6345 s390_indirect_branch_settings also if fndecl didn't change.
6347 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6349 * config/rs6000/rs6000.md (blockage): Set length to zero.
6351 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6353 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6355 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6358 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6359 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6363 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6364 PARALLEL incoming that failed vt_get_decl_and_offset check.
6367 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6368 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6370 (get_variable_section): For decls in named .bss* sections pass true as
6371 second argument to bss_initializer_p.
6373 2018-02-09 Marek Polacek <polacek@redhat.com>
6374 Jakub Jelinek <jakub@redhat.com>
6377 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6378 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6379 Sync some changes from cxx_fold_indirect_ref.
6381 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6383 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6385 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6386 (BLOCK_INLINE_ENTRY_LABEL): New.
6387 (dwarf2out_var_location): Disregard inline entry markers.
6388 (inline_entry_data): New struct.
6389 (inline_entry_data_hasher): New hashtable type.
6390 (inline_entry_data_hasher::hash): New.
6391 (inline_entry_data_hasher::equal): New.
6392 (inline_entry_data_table): New variable.
6393 (add_high_low_attributes): Add DW_AT_entry_pc and
6394 DW_AT_GNU_entry_view attributes if a pending entry is found
6395 in inline_entry_data_table. Add old entry_pc attribute only
6396 if debug nonbinding markers are disabled.
6397 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6398 markers are enabled.
6399 (block_within_block_p, dwarf2out_inline_entry): New.
6400 (dwarf2out_finish): Check that no entries remained in
6401 inline_entry_data_table.
6402 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6403 (final_scan_insn, notice_source_line): Likewise.
6404 (rest_of_clean_state): Skip inline entry markers.
6405 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6407 * gimple.c (gimple_build_debug_inline_entry): New.
6408 * gimple.h (enum gimple_debug_subcode): Add
6409 GIMPLE_DEBUG_INLINE_ENTRY.
6410 (gimple_build_debug_inline_entry): Declare.
6411 (gimple_debug_inline_entry_p): New.
6412 (gimple_debug_nonbind_marker_p): Adjust.
6413 * insn-notes.def (INLINE_ENTRY): New.
6414 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6415 inline entry marker notes.
6416 (print_insn): Likewise.
6417 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6418 (INSN_DEBUG_MARKER_KIND): Likewise.
6419 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6420 * tree-inline.c (expand_call_inline): Build and insert
6421 debug_inline_entry stmt.
6422 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6423 inline entry blocks early, if nonbind markers are enabled.
6424 (dump_scope_block): Dump fragment info.
6425 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6426 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6427 (gimple_build_debug_inline_entry): New.
6428 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6429 Enable/disable inline entry points too.
6430 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6431 (DEBUG_INSN): Describe inline entry markers.
6433 * common.opt (gvariable-location-views): New.
6434 (gvariable-location-views=incompat5): New.
6435 * config.in: Rebuilt.
6436 * configure: Rebuilt.
6437 * configure.ac: Test assembler for view support.
6438 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6439 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6440 * dwarf2out.c (var_loc_view): New typedef.
6441 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6442 (dwarf2out_locviews_in_attribute): New.
6443 (dwarf2out_locviews_in_loclist): New.
6444 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6445 (enum dw_line_info_opcode): Add LI_adv_address.
6446 (struct dw_line_info_table): Add view.
6447 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6448 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6449 (zero_view_p): New variable.
6450 (ZERO_VIEW_P): New macro.
6451 (output_asm_line_debug_info): New.
6452 (struct var_loc_node): Add view.
6453 (add_AT_view_list, AT_loc_list): New.
6454 (add_var_loc_to_decl): Add view param. Test it against last.
6455 (new_loc_list): Add view params. Record them.
6456 (AT_loc_list_ptr): Handle loc and view lists.
6457 (view_list_to_loc_list_val_node): New.
6458 (print_dw_val): Handle dw_val_class_view_list.
6459 (size_of_die): Likewise.
6460 (value_format): Likewise.
6461 (loc_list_has_views): New.
6462 (gen_llsym): Set vl_symbol too.
6463 (maybe_gen_llsym, skip_loc_list_entry): New.
6464 (dwarf2out_maybe_output_loclist_view_pair): New.
6465 (output_loc_list): Output view list or entries too.
6466 (output_view_list_offset): New.
6467 (output_die): Handle dw_val_class_view_list.
6468 (output_dwarf_version): New.
6469 (output_compilation_unit_header): Use it.
6470 (output_skeleton_debug_sections): Likewise.
6471 (output_rnglists, output_line_info): Likewise.
6472 (output_pubnames, output_aranges): Update version comments.
6473 (output_one_line_info_table): Output view numbers in asm comments.
6474 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6475 Call maybe_gen_llsym.
6476 (loc_list_from_tree_1): Adjust.
6477 (add_AT_location_description): Create view list attribute if
6478 needed, check it's absent otherwise.
6479 (convert_cfa_to_fb_loc_list): Adjust.
6480 (maybe_emit_file): Call output_asm_line_debug_info for test.
6481 (dwarf2out_var_location): Reset views as needed. Precompute
6482 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6483 attribute. Set view.
6484 (new_line_info_table): Reset next view.
6485 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6486 (dwarf2out_source_line): Likewise. Output view resets and labels to
6487 the assembler, or select appropriate line info opcodes.
6488 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6489 (optimize_string_length): Catch it. Adjust.
6490 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6491 dw_val_class_view_list, and remove it if no longer needed.
6492 (hash_loc_list): Hash view numbers.
6493 (loc_list_hasher::equal): Compare them.
6494 (optimize_location_lists): Check whether a view list symbol is
6495 needed, and whether the locview attribute is present, and
6496 whether they match. Remove the locview attribute if no longer
6498 (index_location_lists): Call skip_loc_list_entry for test.
6499 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6500 Use output_dwarf_version.
6501 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6502 (struct dw_val_node): Add val_view_list.
6503 * final.c (SEEN_NEXT_VIEW): New.
6504 (set_next_view_needed): New.
6505 (clear_next_view_needed): New.
6506 (maybe_output_next_view): New.
6507 (final_start_function): Rename to...
6508 (final_start_function_1): ... this. Take pointer to FIRST,
6509 add SEEN parameter. Emit param bindings in the initial view.
6510 (final_start_function): Reintroduce SEEN-less interface.
6511 (final): Rename to...
6512 (final_1): ... this. Take SEEN parameter. Output final pending
6513 next view at the end.
6514 (final): Reintroduce seen-less interface.
6515 (final_scan_insn): Output pending next view before switching
6516 sections or ending a block. Mark the next view as needed when
6517 outputting variable locations. Notify debug backend of section
6518 changes, and of location view changes.
6519 (rest_of_handle_final): Adjust.
6520 * toplev.c (process_options): Autodetect value for debug variable
6521 location views option. Warn on incompat5 without -gdwarf-5.
6522 * doc/invoke.texi (gvariable-location-views): New.
6523 (gvariable-location-views=incompat5): New.
6524 (gno-variable-location-views): New.
6526 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6528 PR tree-optimization/84136
6529 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6530 that the result of find_edge is non-NULL.
6532 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6535 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6536 storing integer register in SImode. Fix cost of 256 and 512
6537 byte aligned SSE register store.
6539 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6541 * config/i386/i386.c (ix86_multiplication_cost): Fix
6542 multiplication cost for TARGET_AVX512DQ.
6544 2018-02-08 Marek Polacek <polacek@redhat.com>
6546 PR tree-optimization/84238
6547 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6550 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6552 PR tree-optimization/84265
6553 * tree-vect-stmts.c (vectorizable_store): Don't treat
6554 VMAT_CONTIGUOUS accesses as grouped.
6555 (vectorizable_load): Likewise.
6557 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6559 PR tree-optimization/81635
6560 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6561 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6562 (test_round_for_mask): New functions.
6563 (wide_int_cc_tests): Call test_round_for_mask.
6564 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6565 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6566 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6567 range returned by get_range_info.
6569 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6572 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6573 * symtab.c: Include builtins.h
6574 (symtab_node::output_to_lto_symbol_table_p): Move here
6575 from lto-streamer-out.c:output_symbol_p.
6576 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6577 (output_symbol_p): Move all logic to symtab.c
6578 (produce_symtab): Update.
6580 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6582 * config/s390/s390-opts.h (enum indirect_branch): Define.
6583 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6584 (s390_indirect_branch_via_thunk)
6585 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6586 (enum s390_indirect_branch_type): Define.
6587 * config/s390/s390.c (struct s390_frame_layout, struct
6588 machine_function): Remove.
6589 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6590 (indirect_branch_table_label_no, indirect_branch_table_name):
6592 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6593 (enum s390_indirect_branch_option): Define.
6594 (s390_return_addr_from_memory): New function.
6595 (s390_handle_string_attribute): New function.
6596 (s390_attribute_table): Add new attribute handler.
6597 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6598 (s390_indirect_branch_via_thunk): New function.
6599 (s390_indirect_branch_via_inline_thunk): New function.
6600 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6601 sibling call optimization for non z10 compiles.
6602 (s390_emit_call): Force indirect branch target to be a single
6603 register. Add r1 clobber for non-z10 compiles.
6604 (s390_emit_epilogue): Emit return jump via return_use expander.
6605 (s390_reorg): Handle JUMP_INSNs as execute targets.
6606 (s390_option_override_internal): Perform validity checks for the
6607 new command line options.
6608 (s390_indirect_branch_attrvalue): New function.
6609 (s390_indirect_branch_settings): New function.
6610 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6611 (s390_output_indirect_thunk_function): New function.
6612 (s390_code_end): Implement target hook.
6613 (s390_case_values_threshold): Implement target hook.
6614 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6616 * config/s390/s390.h (struct s390_frame_layout)
6617 (struct machine_function): Move here from s390.c.
6618 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6619 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6620 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6621 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6622 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6623 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6624 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6625 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6626 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6627 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6628 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6629 (mnemonic attribute): Add values which aren't recognized
6631 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6632 pattern for branch conversion. Fix mnemonic attribute.
6633 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6634 indirect branch via thunk if requested.
6635 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6636 ("*indirect_jump"): Disable for branch conversion using out of
6638 ("indirect_jump_via_thunk<mode>_z10")
6639 ("indirect_jump_via_thunk<mode>")
6640 ("indirect_jump_via_inlinethunk<mode>_z10")
6641 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6642 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6643 ("casesi_jump_via_inlinethunk<mode>_z10")
6644 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6645 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6646 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6647 ("*indirect2_jump"): Disable for branch conversion.
6648 ("casesi_jump"): Turn into expander and expand patterns for branch
6650 ("return_use"): New expander.
6651 ("*return"): Emit return via thunk and rename it to ...
6652 ("*return<mode>"): ... this one.
6653 * config/s390/s390.opt: Add new options and and enum for the
6656 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6658 * lra-constraints.c (match_reload): Unconditionally use
6659 gen_lowpart_SUBREG, rather than selecting between that
6660 and equivalent gen_rtx_SUBREG code.
6662 2018-02-08 Richard Biener <rguenther@suse.de>
6664 PR tree-optimization/84233
6665 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6666 changed flag instead of boguously re-using phi_inserted.
6668 2018-02-08 Martin Jambor <mjambor@suse.cz>
6670 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6671 static local variables.
6673 2018-02-08 Richard Biener <rguenther@suse.de>
6675 PR tree-optimization/84278
6676 * tree-vect-stmts.c (vectorizable_store): When looking for
6677 smaller vector types to perform grouped strided loads/stores
6678 make sure the mode is supported by the target.
6679 (vectorizable_load): Likewise.
6681 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6683 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6684 Increase LDP/STP opportunities by adding adjacent callee-saves.
6686 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6688 PR rtl-optimization/84068
6689 PR rtl-optimization/83459
6690 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6692 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6694 PR tree-optimization/84224
6695 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6696 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6699 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6702 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6703 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6704 count, remove one USE.
6706 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6708 * doc/install.texi (Configuration): Document the
6709 --with-long-double-format={ibm,ieee} PowerPC configuration
6713 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6714 Convert from define_expand to be define_insn_and_split. Rework
6715 float/double/_Float128 conversions to QI/HI/SImode to work with
6716 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6717 conversions to QI/HImode types did a store and then a load to
6718 truncate the value. For conversions to VSX registers, don't split
6719 the insn, instead emit the code directly. Use the code iterator
6720 any_fix to combine signed and unsigned conversions.
6721 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6722 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6723 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6724 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6725 (fix_<mode>di2_hw): Likewise.
6726 (fixuns_<mode>di2_hw): Likewise.
6727 (fix_<mode>si2_hw): Likewise.
6728 (fixuns_<mode>si2_hw): Likewise.
6729 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6730 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6731 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6732 fix<uns>_trunc<SFDF:mode>si2_p8.
6733 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6735 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6736 (fix<uns>_<mode>_mem): Likewise.
6737 (fctiw<u>z_<mode>_mem): Likewise.
6738 (fix<uns>_<mode>_mem): Likewise.
6739 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6740 the register allocator from doing a direct move to the GPRs to do
6741 a store, and instead use the ISA 3.0 store byte/half-word from
6742 vector register instruction. For IEEE 128-bit floating point,
6743 also optimize stores of 32-bit ints.
6744 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6746 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6748 * genextract.c (push_pathstr_operand): New function to support
6750 (walk_rtx): Call push_pathstr_operand.
6751 (print_path): Support [a-zA-Z].
6753 2018-02-07 Richard Biener <rguenther@suse.de>
6755 PR tree-optimization/84037
6756 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6757 (cse_and_gimplify_to_preheader): Declare.
6758 (vect_get_place_in_interleaving_chain): Likewise.
6759 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6761 (_loop_vec_info::~_loop_vec_info): Delete it.
6762 (cse_and_gimplify_to_preheader): New function.
6763 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6764 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6765 (vectorizable_load): Likewise. For grouped stores always base
6766 the IV on the first element.
6767 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6768 condition before gimplifying.
6770 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6772 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6773 *DIV_EXPR and *MOD_EXPR.
6775 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6778 * config/i386/i386.c (ix86_option_override_internal): Mask out
6779 the CF_SET bit when checking -fcf-protection.
6781 2018-02-07 Tom de Vries <tom@codesourcery.com>
6784 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6787 2018-02-07 Richard Biener <rguenther@suse.de>
6789 PR tree-optimization/84204
6790 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6793 PR tree-optimization/84205
6794 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6795 special-case isl_ast_op_zdiv_r.
6797 PR tree-optimization/84223
6798 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6799 Only add conditions from within the region.
6800 (gather_bbs::after_dom_children): Adjust.
6802 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6805 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6806 * config/avr/avr.md: Only post-reload split REG-REG moves if
6807 either register is GENERAL_REG_P.
6809 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6811 PR tree-optimization/84235
6812 * tree-ssa-scopedtables.c
6813 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6814 if the subtraction is performed in floating point type where NaNs are
6815 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6816 build 1. Formatting fix.
6818 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6821 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6822 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6823 and skip it regardless of bb boundaries. Use CALL_P macro,
6824 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6825 unnecessarily, formatting fix.
6827 2018-02-06 Michael Collison <michael.collison@arm.com>
6829 * config/arm/thumb2.md:
6830 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6831 (*thumb_mov_notscc): Ditto.
6833 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6836 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6837 unsigned_fix, not "s".
6839 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6841 * configure.ac (gcc_fn_eh_frame_ro): New function.
6842 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6843 correct .eh_frame permissions.
6844 * configure: Regenerate.
6846 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6848 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6851 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6853 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6854 Display warning message for -mno-speculate-indirect-jumps.
6856 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6858 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6860 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6862 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6864 PR tree-optimization/84225
6865 * tree-eh.c (find_trapping_overflow): Only call
6866 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6868 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6871 * config/i386/i386.c: Reimplement the check of possible options
6872 -mibt/-mshstk conbination. Change error messages.
6873 * doc/invoke.texi: Fix a typo: remove extra '='.
6875 2018-02-06 Marek Polacek <polacek@redhat.com>
6877 PR tree-optimization/84228
6878 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6880 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6883 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6884 emitted arch directives.
6885 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6886 __ARM_FEATURE_COPROC before changing architectures.
6888 2018-02-06 Richard Biener <rguenther@suse.de>
6890 * config/i386/i386.c (print_reg): Fix typo.
6891 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6893 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6895 * configure: Regenerate.
6897 2018-02-05 Martin Sebor <msebor@redhat.com>
6899 PR tree-optimization/83369
6900 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6903 2018-02-05 Martin Liska <mliska@suse.cz>
6905 * doc/invoke.texi: Cherry-pick upstream r323995.
6907 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6909 * ira.c (ira_init_register_move_cost): Adjust comment.
6911 2018-02-05 Martin Liska <mliska@suse.cz>
6913 PR gcov-profile/84137
6914 * doc/gcov.texi: Fix typo in documentation.
6916 2018-02-05 Martin Liska <mliska@suse.cz>
6918 PR gcov-profile/83879
6919 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6920 using dlopen functionality.
6922 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6924 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6925 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6926 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6927 _mm_maskz_range_round_ss): New intrinsics.
6928 (__builtin_ia32_rangesd128_round)
6929 (__builtin_ia32_rangess128_round): Remove.
6930 (__builtin_ia32_rangesd128_mask_round,
6931 __builtin_ia32_rangess128_mask_round): New builtins.
6932 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6933 __builtin_ia32_rangess128_round): Remove.
6934 (__builtin_ia32_rangesd128_mask_round,
6935 __builtin_ia32_rangess128_mask_round): New builtins.
6936 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6937 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6938 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6939 "<round_saeonly_constraint>")): Changed to ...
6940 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6941 "<round_saeonly_scalar_constraint>")): ... this.
6942 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6943 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6944 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6945 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6946 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6948 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6950 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6952 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6953 Remove all values except native, 8540 and 8548.
6955 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6957 * config/i386/i386.c (ix86_output_function_return): Pass
6958 INVALID_REGNUM, instead of -1, as invalid register number to
6959 indirect_thunk_name and output_indirect_thunk.
6961 2018-02-02 Julia Koval <julia.koval@intel.com>
6963 * config.gcc: Add -march=icelake.
6964 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6965 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6966 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6967 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6968 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6969 (processor_target_table): Add icelake.
6970 (ix86_option_override_internal): Handle new PTAs.
6971 (get_builtin_code_for_version): Handle icelake.
6972 (M_INTEL_COREI7_ICELAKE): New.
6973 (fold_builtin_cpu): Handle icelake.
6974 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6975 * doc/invoke.texi: Add -march=icelake.
6977 2018-02-02 Julia Koval <julia.koval@intel.com>
6979 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6980 to wide_int_bitmask.
6981 * wide-int-bitmask.h: New.
6983 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6986 * config/i386/i386.md: Replace Pmode with word_mode in
6987 builtin_setjmp_setup and builtin_longjmp to support x32.
6989 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6993 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6995 (rs6000_supported_cpu_names): New static variable.
6996 (linux_cpu_translation_table): Likewise.
6997 (elf_platform) <cpu>: Define new static variable and use it.
6998 Translate kernel AT_PLATFORM name to canonical name if needed.
6999 Error if platform name is unknown.
7001 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7004 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7006 2018-02-01 Jeff Law <law@redhat.com>
7009 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7010 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7011 the scratch if RELEASE_VIA_POP is false.
7012 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7013 If we have to save a temporary register, decrement SIZE appropriately.
7014 Pass new arguments to release_scratch_register_on_entry.
7015 (ix86_adjust_stack_and_probe): Likewise.
7016 (ix86_emit_probe_stack_range): Pass new arguments to
7017 release_scratch_register_on_entry.
7019 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7021 PR rtl-optimization/84157
7022 * combine.c (change_zero_ext): Use REG_P predicate in
7023 front of HARD_REGISTER_P predicate.
7025 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7027 * config/avr/avr.c (avr_option_override): Move disabling of
7028 -fdelete-null-pointer-checks to...
7029 * common/config/avr/avr-common.c (avr_option_optimization_table):
7032 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7034 PR tree-optimization/81635
7035 * tree-data-ref.c (split_constant_offset_1): For types that
7036 wrap on overflow, try to use range info to prove that wrapping
7039 2018-02-01 Renlin Li <renlin.li@arm.com>
7042 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7044 (aarch64_register_move_cost): Likewise.
7045 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7047 (REG_CLASS_NAMES): Likewise.
7048 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7049 TAILCALL_ADDR_REGS. Remove IP registers.
7050 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7052 2018-02-01 Richard Biener <rguenther@suse.de>
7054 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7055 for specifying RPO order and allow NULL for that.
7056 * domwalk.c (dom_walker::dom_walker): Likewise.
7057 (dom_walker::walk): Handle NULL RPO order.
7058 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7060 (rewrite_update_dom_walker): Likewise.
7061 (mark_def_dom_walker): Likewise.
7063 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7065 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7066 (aarch64_maybe_expand_sve_subreg_move): Declare.
7067 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7068 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7070 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7071 that are semantically a reverse operation.
7072 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7073 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7074 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7076 (aarch64_can_change_mode_class): For big-endian, forbid changes
7077 between two SVE modes if they have different element sizes.
7079 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7081 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7082 the TImode handling for big-endian targets.
7084 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7086 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7087 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7089 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7090 Remove BSWAP handing for big-endian targets and use the form of
7091 LD1RQ appropariate for the mode.
7093 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7095 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7096 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7099 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7102 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7103 check for operands that need to go through aarch64_sve_reload_be.
7105 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7107 PR tree-optimization/81661
7108 PR tree-optimization/84117
7109 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7110 * tree-eh.c: Include gimplify.h.
7111 (find_trapping_overflow, replace_trapping_overflow,
7112 rewrite_to_non_trapping_overflow): New functions.
7113 * tree-vect-loop.c: Include tree-eh.h.
7114 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7115 * tree-data-ref.c: Include tree-eh.h.
7116 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7118 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7120 PR rtl-optimization/84123
7121 * combine.c (change_zero_ext): Check if hard register satisfies
7122 can_change_dest_mode before calling gen_lowpart_SUBREG.
7124 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7127 * ira.c (ira_init_register_move_cost): Remove assert.
7129 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7131 PR rtl-optimization/84071
7132 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7133 * doc/tm.texi: Regenerate.
7135 2018-01-31 Richard Biener <rguenther@suse.de>
7137 PR tree-optimization/84132
7138 * tree-data-ref.c (analyze_miv_subscript): Properly
7139 check whether evolution_function_is_affine_multivariate_p
7140 before calling gcd_of_steps_may_divide_p.
7142 2018-01-31 Julia Koval <julia.koval@intel.com>
7145 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7146 * config/i386/i386.md (rdpid_rex64) New.
7147 (rdpid): Make 32bit only.
7149 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7152 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7153 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7155 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7158 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7160 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7162 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7164 PR rtl-optimization/84071
7165 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7166 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7168 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7170 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7171 (arc_attribute_table): Add 'aux' attribute.
7172 (arc_in_small_data_p): Consider aux like variables.
7173 (arc_is_aux_reg_p): New function.
7174 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7175 (arc_get_aux_arg): New function.
7176 (prepare_move_operands): Handle aux-register access.
7177 (arc_handle_aux_attribute): New function.
7178 * doc/extend.texi (ARC Variable attributes): Add subsection.
7180 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7182 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7183 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7184 (arc_attribute_table): Add 'uncached' attribute.
7185 (arc_print_operand): Print '.di' flag for uncached memory
7187 (arc_in_small_data_p): Do not consider for small data the uncached
7189 (arc_is_uncached_mem_p): New function.
7190 * config/arc/predicates.md (compact_store_memory_operand): Check
7191 for uncached memory accesses.
7192 (nonvol_nonimm_operand): Likewise.
7193 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7195 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7198 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7199 falign-loops=): Add Optimization flag.
7201 2018-01-30 Jeff Law <law@redhat.com>
7204 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7205 INT_REGISTERS_SAVED. Check it prior to calling
7206 get_scratch_register_on_entry.
7207 (ix86_adjust_stack_and_probe): Similarly.
7208 (ix86_emit_probe_stack_range): Similarly.
7209 (ix86_expand_prologue): Corresponding changes.
7211 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7214 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7215 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7217 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7220 * lra-constraints.c (curr_insn_transform): Process AND in the
7223 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7225 PR rtl-optimization/83986
7226 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7227 dependence against last_pending_memory_flush in addition to
7230 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7232 PR tree-optimization/81611
7233 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7236 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7239 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7242 2018-01-30 Richard Biener <rguenther@suse.de>
7243 Jakub Jelinek <jakub@redhat.com>
7245 PR tree-optimization/84111
7246 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7247 inner loops added during recursion, as they don't have up-to-date
7250 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7253 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7254 (can_inline_edge_by_limits_p): ... here.
7255 (can_early_inline_edge_p, check_callers,
7256 update_caller_keys, update_callee_keys, recursive_inlining,
7257 add_new_edges_to_heap, speculation_useful_p,
7258 inline_small_functions,
7259 inline_small_functions, flatten_function,
7260 inline_to_all_callers_1): Update.
7262 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7264 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7267 2018-01-30 Richard Biener <rguenther@suse.de>
7269 PR tree-optimization/83008
7270 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7271 invariant and constant vector uses in stmts when they need
7274 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7277 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7278 * configure: Regenerate.
7280 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7282 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7284 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7285 Use gen_rtx_REG rather than gen_lowpart.
7287 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7289 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7290 rather than 0 when creating partial subregs.
7292 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7294 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7297 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7300 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7301 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7302 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7303 flags. This restores the settings used before the 2017-07-24.
7304 Turning off pre increment/decrement/modify allows IVOPTS to
7305 optimize DF/SF loops where the index is an int.
7307 2018-01-29 Richard Biener <rguenther@suse.de>
7308 Kelvin Nilsen <kelvin@gcc.gnu.org>
7311 * tree-vect-stmts.c (vectorizable_call): Don't call
7312 targetm.vectorize_builtin_md_vectorized_function if callee is
7315 2018-01-22 Carl Love <cel@us.ibm.com>
7317 * doc/extend.tex: Fix typo in second arg in
7318 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7320 2018-01-29 Richard Biener <rguenther@suse.de>
7322 PR tree-optimization/84086
7323 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7324 (flush_ssaname_freelist): When SSA names were released reset
7325 the SCEV hash table.
7327 2018-01-29 Richard Biener <rguenther@suse.de>
7329 PR tree-optimization/84057
7330 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7331 removed paths when removing edges.
7333 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7335 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7336 -mfunction-return=@var{choice}.
7338 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7341 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7343 (layout::print_source_line): Likewise.
7344 (test_get_line_width_without_trailing_whitespace): Add test cases.
7346 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7349 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7352 2018-01-26 Jim Wilson <jimw@sifive.com>
7354 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7356 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7359 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7361 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7362 and CMP + SUB-immediate -> SUBS.
7364 2018-01-26 Martin Sebor <msebor@redhat.com>
7366 PR tree-optimization/83896
7367 * tree-ssa-strlen.c (get_string_len): Rename...
7368 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7369 Avoid assuming length is constant.
7370 (handle_char_store): Use HOST_WIDE_INT for string length.
7372 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7375 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7376 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7378 2018-01-26 Richard Biener <rguenther@suse.de>
7380 PR rtl-optimization/84003
7381 * dse.c (record_store): Only record redundant stores when
7382 the earlier store aliases at least all accesses the later one does.
7384 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7386 PR rtl-optimization/83985
7387 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7388 REG_CFA_RESTORE insns.
7389 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7390 REG_CFA_RESTORE insns here.
7393 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7394 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7396 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7398 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7399 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7400 (arc_init): Likewise.
7401 (arc_override_options): Likewise.
7402 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7404 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7406 * config/arc/arc.h (TARGET_DBNZ): Define.
7407 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7408 properly set the tune attribute.
7409 (dbnz): Use TARGET_DBNZ guard.
7410 * config/arc/arc.opt (mtune): Add core3 option.
7412 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7414 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7415 recognize new pic like addresses.
7416 (arc_delegitimize_address): Clean up.
7418 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7420 * config/arc/arc-arches.def: Option mrf16 valid for all
7422 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7423 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7424 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7425 * config/arc/arc-tables.opt: Regenerate.
7426 * config/arc/arc.c (arc_conditional_register_usage): Handle
7427 reduced register file case.
7428 (arc_file_start): Set must have build attributes.
7429 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7431 * config/arc/arc.opt (mrf16): Add new option.
7432 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7433 * config/arc/genmultilib.awk: Handle new mrf16 option.
7434 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7435 * config/arc/t-multilib: Regenerate.
7436 * doc/invoke.texi (ARC Options): Document mrf16 option.
7438 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7440 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7441 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7442 (arc_attribute_table): Add 'secure_call' attribute.
7443 (arc_print_operand): Print secure call operand.
7444 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7446 (arc_is_secure_call_p): New function. * config/arc/arc.md
7447 (call_i): Add support for sjli instruction.
7448 (call_value_i): Likewise.
7449 * config/arc/constraints.md (Csc): New constraint.
7451 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7452 John Eric Martin <John.Martin@emmicro-us.com>
7454 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7455 * config/arc/arc.c (_arc_jli_section): New struct.
7456 (arc_jli_section): New type.
7457 (rc_jli_sections): New static variable.
7458 (arc_handle_jli_attribute): New function.
7459 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7460 (arc_file_end): New function.
7461 (TARGET_ASM_FILE_END): Define.
7462 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7463 (arc_add_jli_section): New function.
7464 (jli_call_scan): Likewise.
7465 (arc_reorg): Call jli_call_scan.
7466 (arc_output_addsi): Remove 'S' from printing asm operand.
7467 (arc_is_jli_call_p): New function.
7468 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7470 (movhi_insn): Likewise.
7471 (movsi_insn): Likewise.
7472 (movsi_set_cc_insn): Likewise.
7473 (loadqi_update): Likewise.
7474 (load_zeroextendqisi_update): Likewise.
7475 (load_signextendqisi_update): Likewise.
7476 (loadhi_update): Likewise.
7477 (load_zeroextendhisi_update): Likewise.
7478 (load_signextendhisi_update): Likewise.
7479 (loadsi_update): Likewise.
7480 (loadsf_update): Likewise.
7481 (movsicc_insn): Likewise.
7482 (bset_insn): Likewise.
7483 (bxor_insn): Likewise.
7484 (bclr_insn): Likewise.
7485 (bmsk_insn): Likewise.
7486 (bicsi3_insn): Likewise.
7487 (cmpsi_cc_c_insn): Likewise.
7488 (movsi_ne): Likewise.
7489 (movsi_cond_exec): Likewise.
7490 (clrsbsi2): Likewise.
7501 (call_i): Remove 'S' asm letter, add jli instruction.
7502 (call_value_i): Likewise.
7503 * config/arc/arc.op (mjli-always): New option.
7504 * config/arc/constraints.md (Cji): New constraint.
7505 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7507 (subsf3_fpx): Likewise.
7508 (mulsf3_fpx): Likewise.
7509 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7511 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7513 * doc/invoke.texi (ARC): Document mjli-always option.
7515 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7517 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7518 avoid addition with 0 and use incw and decw where possible.
7520 2018-01-26 Richard Biener <rguenther@suse.de>
7522 PR tree-optimization/81082
7523 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7524 association if it requires casting to unsigned.
7525 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7526 from fold_plusminus_mult_expr to catch important cases late when
7527 range info is available.
7529 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7531 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7532 * configure.ac (hidden_linkonce): New test.
7533 * configure: Regenerate.
7534 * config.in: Regenerate.
7536 2018-01-26 Julia Koval <julia.koval@intel.com>
7538 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7539 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7540 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7541 _mm_mask_bitshuffle_epi64_mask): Fix type.
7542 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7543 USI_FTYPE_V4DI_V4DI_USI): Remove.
7544 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7545 __builtin_ia32_vpshufbitqmb256_mask,
7546 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7547 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7548 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7550 2018-01-26 Alan Modra <amodra@gmail.com>
7553 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7554 UNSPEC_VBPERMQ. Sort other unspecs.
7556 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7558 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7560 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7563 * predict.c (drop_profile): Do not push/pop cfun; update also
7565 (handle_missing_profiles): Fix logic looking for zero profiles.
7567 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7570 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7571 on functions with #pragma omp declare simd or functions with simd
7573 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7574 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7575 Remove trailing \n from warning_at calls.
7577 2018-01-25 Tom de Vries <tom@codesourcery.com>
7580 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7581 for neutered workers.
7583 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7586 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7587 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7589 2018-01-24 Jeff Law <law@redhat.com>
7592 * i386.c (get_probe_interval): Move to earlier point.
7593 (ix86_compute_frame_layout): If -fstack-clash-protection and
7594 the frame is larger than the probe interval, then use pushes
7595 to save registers rather than reg->mem moves.
7596 (ix86_expand_prologue): Remove conditional for int_registers_saved
7599 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7602 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7603 min/max for never referenced object.
7605 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7608 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7610 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7611 attributes from DECL_ATTRIBUTES (decl) without affecting
7612 DECL_ATTRIBUTES (current_function_decl).
7613 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7614 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7616 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7618 PR tree-optimization/83979
7619 * fold-const.c (fold_comparison): Use constant_boolean_node
7620 instead of boolean_{true,false}_node.
7622 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7624 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7627 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7629 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7630 Simplify the clause that sets the length attribute.
7631 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7632 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7633 clause that sets the length attribute.
7634 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7636 2018-01-24 Tom de Vries <tom@codesourcery.com>
7639 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7640 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7641 Add strict parameter.
7642 (prevent_branch_around_nothing): Insert dummy insn between branch to
7643 label and label with no ptx insn inbetween.
7644 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7646 2018-01-24 Tom de Vries <tom@codesourcery.com>
7649 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7650 for neutered threads in warp.
7651 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7653 2018-01-24 Richard Biener <rguenther@suse.de>
7655 PR tree-optimization/83176
7656 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7659 2018-01-24 Richard Biener <rguenther@suse.de>
7661 PR tree-optimization/82819
7662 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7663 code generating pluses that are no-ops in the target precision.
7665 2018-01-24 Richard Biener <rguenther@suse.de>
7668 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7670 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7672 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7673 to merge probabilities.
7674 * predict.c (probably_never_executed): Also mark as cold functions
7675 with global 0 profile and guessed local profile.
7676 * profile-count.c (profile_probability::combine_with_count): New
7678 * profile-count.h (profile_probability::operator*,
7679 profile_probability::operator*=, profile_probability::operator/,
7680 profile_probability::operator/=): Reduce precision to adjusted
7681 and set value to guessed on contradictory divisions.
7682 (profile_probability::combine_with_freq): Remove.
7683 (profile_probability::combine_wiht_count): Declare.
7684 (profile_count::force_nonzero):: Set to adjusted.
7685 (profile_count::probability_in):: Set quality to adjusted.
7686 * tree-ssa-tail-merge.c (replace_block_by): Use
7689 2018-01-23 Andrew Waterman <andrew@sifive.com>
7690 Jim Wilson <jimw@sifive.com>
7692 * config/riscv/riscv.c (riscv_stack_boundary): New.
7693 (riscv_option_override): Set riscv_stack_boundary. Handle
7694 riscv_preferred_stack_boundary_arg.
7695 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7696 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7697 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7698 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7699 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7700 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7702 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7705 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7706 of struct ix86_frame.
7707 (ix86_expand_epilogue): Likewise. Add a local variable for
7708 the reg_save_offset field in struct ix86_frame.
7710 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7712 PR tree-optimization/82604
7713 * tree-loop-distribution.c (enum partition_kind): New enum item
7714 PKIND_PARTIAL_MEMSET.
7715 (partition_builtin_p): Support above new enum item.
7716 (generate_code_for_partition): Ditto.
7717 (compute_access_range): Differentiate cases that equality can be
7718 proven at all loops, the innermost loops or no loops.
7719 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7720 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7721 (finalize_partitions, distribute_loop): Don't fuse partition of
7722 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7723 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7726 2018-01-23 Martin Liska <mliska@suse.cz>
7728 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7729 order to ignore the predictor.
7730 (PRED_POLYMORPHIC_CALL): Likewise.
7731 (PRED_RECURSIVE_CALL): Likewise.
7733 2018-01-23 Martin Liska <mliska@suse.cz>
7735 * tree-profile.c (tree_profiling): Print function header to
7736 aware reader which function we are working on.
7737 * value-prof.c (gimple_find_values_to_profile): Do not print
7738 not interesting value histograms.
7740 2018-01-23 Martin Liska <mliska@suse.cz>
7742 * profile-count.h (enum profile_quality): Add
7743 profile_uninitialized as the first value. Do not number values
7744 as they are zero based.
7745 (profile_count::verify): Update sanity check.
7746 (profile_probability::verify): Likewise.
7748 2018-01-23 Nathan Sidwell <nathan@acm.org>
7750 * doc/invoke.texi (ffor-scope): Deprecate.
7752 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7754 PR tree-optimization/83510
7755 * domwalk.c (set_all_edges_as_executable): New function.
7756 (dom_walker::dom_walker): Convert bool param
7757 "skip_unreachable_blocks" to enum reachability. Move setup of
7758 edge flags to set_all_edges_as_executable and only do it when
7759 reachability is REACHABLE_BLOCKS.
7760 * domwalk.h (enum dom_walker::reachability): New enum.
7761 (dom_walker::dom_walker): Convert bool param
7762 "skip_unreachable_blocks" to enum reachability.
7763 (set_all_edges_as_executable): New decl.
7764 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7765 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7767 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7768 but converting true to REACHABLE_BLOCKS.
7769 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7771 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7772 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7773 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7775 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7776 if check_all_array_refs will be called.
7778 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7780 * tree.c (selftest::test_location_wrappers): Add more test
7783 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7785 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7786 (selftest::test_bit_in_range): Likewise.
7788 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7791 * doc/sourcebuild.texi (vect_float): Say that the selector
7792 only describes the situation when -funsafe-math-optimizations is on.
7793 (vect_float_strict): Document.
7795 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7797 PR tree-optimization/83965
7798 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7799 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7800 instead of checking only for a reduction.
7801 (vect_recog_widen_sum_pattern): Likewise.
7803 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7805 * predict.c (probably_never_executed): Only use precise profile info.
7806 (compute_function_frequency): Skip after inlining hack since we now
7807 have quality checking.
7809 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7811 * profile-count.h (profile_probability::very_unlikely,
7812 profile_probability::unlikely, profile_probability::even): Set
7813 precision to guessed.
7815 2018-01-23 Richard Biener <rguenther@suse.de>
7817 PR tree-optimization/83963
7818 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7819 Properly terminate dominator walk when crossing the exit edge not
7820 when visiting its source block.
7822 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7825 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7826 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7828 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7830 PR tree-optimization/83957
7831 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7832 semicolon after for body surrounded by braces.
7834 PR tree-optimization/83081
7835 * profile-count.h (profile_probability::split): New method.
7836 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7837 Use profile_probability::split.
7838 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7839 when splitting a single conditional jump into 2.
7841 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7843 PR tree-optimization/69452
7844 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7847 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7849 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7850 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7851 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7853 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7855 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7856 * config/rl78/rl78.md (movdi): New define_expand.
7857 * config/rl78/rl78.c (rl78_split_movdi): New function.
7859 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7862 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7864 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7865 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7866 128-bit to produce an UNSPEC move to get the double word with the
7867 signbit and then a shift directly to do signbit.
7868 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7869 implementation with a new version that just does either a direct
7870 move or a regular move. Move memory interface to separate insns.
7871 Move insns so they are next to the expander.
7872 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7873 with signbit move. Split big and little endian case.
7874 (signbit<mode>2_dm_mem_le): Likewise.
7875 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7876 (signbit<mode>2_dm2): Likewise.
7878 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7880 * config/rl78/rl78.md (anddi3): New define_expand.
7882 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7884 * config/rl78/rl78.md (umindi3): New define_expand.
7886 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7888 * config/rl78/rl78.md (smindi3): New define_expand.
7890 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7892 * config/rl78/rl78.md (smaxdi3): New define_expand.
7894 2018-01-22 Carl Love <cel@us.ibm.com>
7896 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7897 LVX_V1TI): Add macro expansion.
7898 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7899 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7900 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7901 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7902 Change check to determine if the instruction is a byte reversing
7903 entry. Fix typo in comment.
7904 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7905 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7906 Add def_builtin calls for new builtins.
7907 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7908 Add define_insn expansion.
7910 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7912 * config/rl78/rl78.md (umaxdi3): New define_expand.
7914 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7916 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7917 for non-QImode registers.
7919 2018-01-22 Richard Biener <rguenther@suse.de>
7921 PR tree-optimization/83963
7922 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7923 including the loop exit block.
7924 (scop_detection::merge_sese): Likewise.
7925 (scop_detection::add_scop): Do it here instead.
7927 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7929 * doc/sourcebuild.texi (arm_softfloat): Document.
7931 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7934 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7935 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7936 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7938 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7939 David Edelsohn <dje.gcc@gmail.com>
7942 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7943 Change "crset eq" to "crset 2".
7944 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7945 (*call_indirect_aix<mode>_nospec): Likewise.
7946 (*call_value_indirect_aix<mode>_nospec): Likewise.
7947 (*call_indirect_elfv2<mode>_nospec): Likewise.
7948 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7949 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7950 change assembly output from . to $.
7951 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7952 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7953 (*tablejump<mode>_internal1_nospec): Likewise.
7955 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7958 * config/sh/sh_optimize_sett_clrt.cc:
7959 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7961 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7963 PR tree-optimization/83940
7964 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7965 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7966 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7967 use it to pass back the definition type.
7968 (vect_check_store_rhs): Likewise rhs_dt_out.
7969 (vect_build_gather_load_calls): Add a mask_dt argument and use
7970 it instead of a call to vect_is_simple_use.
7971 (vectorizable_store): Update calls to vect_check_load_store_mask
7972 and vect_check_store_rhs. Use the dt returned by the latter instead
7973 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7974 instead of calls to vect_is_simple_use. Pass the scalar rather
7975 than the vector operand to vect_is_simple_use when handling
7976 second and subsequent copies of an rhs value.
7977 (vectorizable_load): Update calls to vect_check_load_store_mask
7978 and vect_build_gather_load_calls. Use the cached mask_dt and
7979 gs_info.offset_dt instead of calls to vect_is_simple_use.
7981 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7984 * tree-emutls.c: Include gimplify.h.
7985 (lower_emutls_2): New function.
7986 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7987 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7988 it before further processing.
7991 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7992 UINTVAL (trueop1) instead of INTVAL (op1).
7994 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7998 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7999 INCOMING_FRAME_SP_OFFSET if not defined.
8000 (scan_trace): Add ENTRY argument. If true and
8001 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8002 emit a note to adjust the CFA offset.
8003 (create_cfi_notes): Adjust scan_trace callers.
8004 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8005 INCOMING_FRAME_SP_OFFSET in the CIE.
8006 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8007 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8009 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8010 * doc/tm.texi: Regenerated.
8012 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8014 PR rtl-optimization/83147
8015 * lra-constraints.c (remove_inheritance_pseudos): Use
8016 lra_substitute_pseudo_within_insn.
8018 2018-01-19 Tom de Vries <tom@codesourcery.com>
8019 Cesar Philippidis <cesar@codesourcery.com>
8022 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8024 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8027 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8028 spaces for function labels.
8030 2018-01-19 Martin Liska <mliska@suse.cz>
8032 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8033 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8034 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8035 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8036 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8037 (PRED_CONST_RETURN): Change from 69 to 65.
8038 (PRED_NULL_RETURN): Change from 91 to 71.
8039 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8040 (PRED_LOOP_GUARD): Change from 66 to 73.
8042 2018-01-19 Martin Liska <mliska@suse.cz>
8044 * predict.c (predict_insn_def): Add new assert.
8045 (struct branch_predictor): Change type to signed integer.
8046 (test_prediction_value_range): Amend test to cover
8048 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8049 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8050 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8051 (PRED_LOOP_IV_COMPARE): Likewise.
8052 * predict.h (PROB_UNINITIALIZED): Define new constant.
8054 2018-01-19 Martin Liska <mliska@suse.cz>
8056 * predict.c (dump_prediction): Add new format for
8057 analyze_brprob.py script which is enabled with -details
8059 * profile-count.h (precise_p): New function.
8061 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8063 PR tree-optimization/83922
8064 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8065 there are no statements that need masking.
8066 (vect_active_double_reduction_p): New function.
8067 (vect_analyze_loop_operations): Use it when handling phis that
8068 are not in the loop header.
8070 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8072 PR tree-optimization/83914
8073 * tree-vect-loop.c (vectorizable_induction): Don't convert
8074 init_expr or apply the peeling adjustment for inductions
8075 that are nested within the vectorized loop.
8077 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8079 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8082 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8086 * function.h (gimplify_parameters): Add gimple_seq * argument.
8087 * function.c: Include gimple.h and options.h.
8088 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8089 for the added local temporaries if needed.
8090 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8091 if there are any parameter cleanups, wrap whole body into a
8092 try/finally with the cleanups.
8094 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8097 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8098 Use GET_MODE_CLASS for scalar floating point.
8100 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8104 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8105 Fix call of call_cgraph_insertion_hooks.
8107 2018-01-18 Martin Sebor <msebor@redhat.com>
8109 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8111 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8114 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8117 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8120 * common.opt: (-ffile-prefix-map): New option.
8121 * opts.c (common_handle_option): Defer it.
8122 * opts-global.c (handle_common_deferred_options): Handle it.
8123 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8124 * file-prefix-map.h: New file.
8125 (remap_debug_filename, add_debug_prefix_map): ...here.
8126 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8127 * final.c (debug_prefix_map, add_debug_prefix_map
8128 remap_debug_filename): Move to...
8129 * file-prefix-map.c: New file.
8130 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8131 generalize, get rid of alloca(), use strrchr() instead of strchr().
8132 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8133 Implement in terms of add_prefix_map().
8134 (remap_macro_filename, remap_debug_filename): Implement in term of
8136 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8137 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8138 * dbxout.c: Include file-prefix-map.h.
8139 * varasm.c: Likewise.
8140 * vmsdbgout.c: Likewise.
8141 * xcoffout.c: Likewise.
8142 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8143 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8144 * doc/invoke.texi (-ffile-prefix-map): Document.
8145 (-fdebug-prefix-map): Update description.
8147 2018-01-18 Martin Liska <mliska@suse.cz>
8149 * config/i386/i386.c (indirect_thunk_name): Document that also
8151 (output_indirect_thunk): Document why both instructions
8152 (pause and lfence) are generated.
8154 2018-01-18 Richard Biener <rguenther@suse.de>
8156 PR tree-optimization/83887
8157 * graphite-scop-detection.c
8158 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8159 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8160 (scop_detection::merge_sese): Re-implement with a flood-fill
8161 algorithm that properly finds a SESE region if it exists.
8163 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8166 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8167 pointer_diff optimizations use view_convert instead of convert.
8169 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8171 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8172 Generate different code for -mno-speculate-indirect-jumps.
8173 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8174 (*call_indirect_aix<mode>): Disable for
8175 -mno-speculate-indirect-jumps.
8176 (*call_indirect_aix<mode>_nospec): New define_insn.
8177 (*call_value_indirect_aix<mode>): Disable for
8178 -mno-speculate-indirect-jumps.
8179 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8180 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8181 -mno-speculate-indirect-jumps.
8182 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8184 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8186 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8187 long double type, set the flags for noting the default long double
8188 type, even if we don't pass or return a long double type.
8190 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8193 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8196 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8198 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8199 support for merge[hl].
8200 (fold_mergehl_helper): New helper function.
8201 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8202 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8203 (altivec_vmrglw_direct): Add xxmrglw insn.
8205 2018-01-17 Andrew Waterman <andrew@sifive.com>
8207 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8208 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8210 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8213 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8214 call the lto_location_cache before reading the
8215 DECL_SOURCE_LOCATION of the types.
8217 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8218 Richard Sandiford <richard.sandiford@linaro.org>
8220 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8221 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8222 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8223 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8224 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8226 * config/aarch64/constraints.md (aarch64_movti_operand):
8228 * config/aarch64/predicates.md (Uti): Add new constraint.
8230 2018-01-17 Carl Love <cel@us.ibm.com>
8232 * config/rs6000/vsx.md (define_expand xl_len_r,
8233 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8234 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8236 (define_expand, define_insn): Move the shift left from the
8237 define_insn to the define_expand for lxvl and stxvl instructions.
8238 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8239 and XL_LEN_R definitions to PURE.
8241 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8243 * config/i386/i386.c (indirect_thunk_name): Declare regno
8244 as unsigned int. Compare regno with INVALID_REGNUM.
8245 (output_indirect_thunk): Ditto.
8246 (output_indirect_thunk_function): Ditto.
8247 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8248 in the call to output_indirect_thunk_function.
8250 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8253 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8254 rather than the size of inner_type to determine the stack slot size
8255 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8257 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8260 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8263 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8265 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8266 endian Linux systems to optionally enable multilibs for selecting
8267 the long double type if the user configured an explicit type.
8268 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8269 have no long double multilibs if not defined.
8270 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8271 warn if the user used -mabi={ieee,ibm}longdouble and we built
8272 multilibs for long double.
8273 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8274 appropriate multilib option.
8275 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8277 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8278 for building long double multilibs.
8279 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8281 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8283 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8286 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8288 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8291 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8294 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8297 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8299 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8300 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8302 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8304 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8305 different rtl trees depending on TARGET_64BIT.
8306 (rs6000_gen_lvx): Likewise.
8308 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8310 * config/visium/visium.md (nop): Tweak comment.
8311 (hazard_nop): Likewise.
8313 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8315 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8316 -mspeculate-indirect-jumps.
8317 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8318 for -mno-speculate-indirect-jumps.
8319 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8320 (*call_value_indirect_elfv2<mode>): Disable for
8321 -mno-speculate-indirect-jumps.
8322 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8323 (indirect_jump): Emit different RTL for
8324 -mno-speculate-indirect-jumps.
8325 (*indirect_jump<mode>): Disable for
8326 -mno-speculate-indirect-jumps.
8327 (*indirect_jump<mode>_nospec): New define_insn.
8328 (tablejump): Emit different RTL for
8329 -mno-speculate-indirect-jumps.
8330 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8331 (tablejumpsi_nospec): New define_expand.
8332 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8333 (tablejumpdi_nospec): New define_expand.
8334 (*tablejump<mode>_internal1): Disable for
8335 -mno-speculate-indirect-jumps.
8336 (*tablejump<mode>_internal1_nospec): New define_insn.
8337 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8340 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8342 * caller-save.c (insert_save): Drop unnecessary parameter. All
8345 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8346 Richard Biener <rguenth@suse.de>
8349 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8350 return early, inline manually is_gimple_sizepos. Make sure if we
8351 call gimplify_expr we don't end up with a gimple constant.
8352 * tree.c (variably_modified_type_p): Don't return true for
8353 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8354 * gimplify.h (is_gimple_sizepos): Remove.
8356 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8358 PR tree-optimization/83857
8359 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8360 vectorizable_live_operation for pure SLP statements.
8361 (vectorizable_live_operation): Handle PHIs.
8363 2018-01-16 Richard Biener <rguenther@suse.de>
8365 PR tree-optimization/83867
8366 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8367 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8369 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8372 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8373 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8374 If off is not INTEGER_CST, issue a may not be aligned warning
8375 rather than isn't aligned. Use isn%'t rather than isn't.
8376 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8378 <case MULT_EXPR>: Improve the case when bottom and one of the
8379 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8380 operand, in that case check if the other operand is multiple of
8381 bottom divided by the INTEGER_CST operand.
8383 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8386 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8387 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8388 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8389 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8390 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8391 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8392 (pa_function_arg_size): New function.
8394 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8396 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8397 in a separate statement.
8399 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8401 PR tree-optimization/83847
8402 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8403 group gathers and scatters.
8405 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8407 PR rtl-optimization/86620
8408 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8410 PR rtl-optimization/83213
8411 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8412 to last if both are JUMP_INSNs.
8414 PR tree-optimization/83843
8415 * gimple-ssa-store-merging.c
8416 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8417 store_immediate_info for bswap/nop orig_stores.
8419 2018-01-15 Andrew Waterman <andrew@sifive.com>
8421 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8423 <UDIV>: Increase cost if !TARGET_DIV.
8425 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8427 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8428 (define_attr "cr_logical_3op"): New.
8429 (cceq_ior_compare): Adjust.
8430 (cceq_ior_compare_complement): Adjust.
8431 (*cceq_rev_compare): Adjust.
8432 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8433 (is_cracked_insn): Adjust.
8434 (insn_must_be_first_in_group): Adjust.
8435 * config/rs6000/40x.md: Adjust.
8436 * config/rs6000/440.md: Adjust.
8437 * config/rs6000/476.md: Adjust.
8438 * config/rs6000/601.md: Adjust.
8439 * config/rs6000/603.md: Adjust.
8440 * config/rs6000/6xx.md: Adjust.
8441 * config/rs6000/7450.md: Adjust.
8442 * config/rs6000/7xx.md: Adjust.
8443 * config/rs6000/8540.md: Adjust.
8444 * config/rs6000/cell.md: Adjust.
8445 * config/rs6000/e300c2c3.md: Adjust.
8446 * config/rs6000/e500mc.md: Adjust.
8447 * config/rs6000/e500mc64.md: Adjust.
8448 * config/rs6000/e5500.md: Adjust.
8449 * config/rs6000/e6500.md: Adjust.
8450 * config/rs6000/mpc.md: Adjust.
8451 * config/rs6000/power4.md: Adjust.
8452 * config/rs6000/power5.md: Adjust.
8453 * config/rs6000/power6.md: Adjust.
8454 * config/rs6000/power7.md: Adjust.
8455 * config/rs6000/power8.md: Adjust.
8456 * config/rs6000/power9.md: Adjust.
8457 * config/rs6000/rs64.md: Adjust.
8458 * config/rs6000/titan.md: Adjust.
8460 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8462 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8463 ix86_indirect_branch_register logic.
8465 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8467 * config/i386/constraints.md (Bs): Update
8468 ix86_indirect_branch_register check. Don't check
8469 ix86_indirect_branch_register with GOT_memory_operand.
8471 * config/i386/predicates.md (GOT_memory_operand): Don't check
8472 ix86_indirect_branch_register here.
8473 (GOT32_symbol_operand): Likewise.
8475 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8477 * config/i386/predicates.md (constant_call_address_operand):
8478 Rewrite ix86_indirect_branch_register logic.
8479 (sibcall_insn_operand): Likewise.
8481 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8483 * config/i386/constraints.md (Bs): Replace
8484 ix86_indirect_branch_thunk_register with
8485 ix86_indirect_branch_register.
8487 * config/i386/i386.md (indirect_jump): Likewise.
8488 (tablejump): Likewise.
8489 (*sibcall_memory): Likewise.
8490 (*sibcall_value_memory): Likewise.
8491 Peepholes of indirect call and jump via memory: Likewise.
8492 * config/i386/i386.opt: Likewise.
8493 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8494 (GOT_memory_operand): Likewise.
8495 (call_insn_operand): Likewise.
8496 (sibcall_insn_operand): Likewise.
8497 (GOT32_symbol_operand): Likewise.
8499 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8502 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8503 type rather than type addr's type points to.
8504 (expand_omp_atomic_mutex): Likewise.
8505 (expand_omp_atomic): Likewise.
8507 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8510 * config/i386/i386.c (output_indirect_thunk_function): Use
8511 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8512 for __x86_return_thunk.
8514 2018-01-15 Richard Biener <rguenther@suse.de>
8517 * expmed.c (extract_bit_field_1): Fix typo.
8519 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8522 * config/arm/iterators.md (VF): New mode iterator.
8523 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8524 Remove integer-related logic from pattern.
8525 (neon_vabd<mode>_3): Likewise.
8527 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8530 * common.opt (fstrict-overflow): No longer an alias.
8531 (fwrapv-pointer): New option.
8532 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8533 also for pointer types based on flag_wrapv_pointer.
8534 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8535 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8536 opts->x_flag_wrapv got set.
8537 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8538 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8539 POINTER_TYPE_OVERFLOW_UNDEFINED.
8540 * match.pd: Likewise in address comparison pattern.
8541 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8543 2018-01-15 Richard Biener <rguenther@suse.de>
8546 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8547 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8548 Reset type names to their identifier if their TYPE_DECL doesn't
8549 have linkage (and thus is used for ODR and devirt).
8550 (save_debug_info_for_decl): Remove.
8551 (save_debug_info_for_type): Likewise.
8552 (add_tree_to_fld_list): Adjust.
8553 * tree-pretty-print.c (dump_generic_node): Make dumping of
8554 type names more robust.
8556 2018-01-15 Richard Biener <rguenther@suse.de>
8558 * BASE-VER: Bump to 8.0.1.
8560 2018-01-14 Martin Sebor <msebor@redhat.com>
8563 * builtins.c (check_access): Avoid warning when the no-warning bit
8566 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8568 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8569 * ira-color (allocno_hard_regs_compare): Likewise.
8571 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8574 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8575 Use .pushsection/.popsection.
8577 2018-01-14 Martin Sebor <msebor@redhat.com>
8580 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8582 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8584 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8585 entry from extra_headers.
8586 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8587 extra_headers, make the list bitwise identical to the i?86-*-* one.
8589 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8591 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8592 -mcmodel=large with -mindirect-branch=thunk,
8593 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8594 -mfunction-return=thunk-extern.
8595 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8596 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8597 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8599 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8601 * config/i386/i386.c (print_reg): Print the name of the full
8602 integer register without '%'.
8603 (ix86_print_operand): Handle 'V'.
8604 * doc/extend.texi: Document 'V' modifier.
8606 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8608 * config/i386/constraints.md (Bs): Disallow memory operand for
8609 -mindirect-branch-register.
8611 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8612 (GOT_memory_operand): Likewise.
8613 (call_insn_operand): Likewise.
8614 (sibcall_insn_operand): Likewise.
8615 (GOT32_symbol_operand): Likewise.
8616 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8617 for -mindirect-branch-register.
8618 (tablejump): Likewise.
8619 (*sibcall_memory): Likewise.
8620 (*sibcall_value_memory): Likewise.
8621 Disallow peepholes of indirect call and jump via memory for
8622 -mindirect-branch-register.
8623 (*call_pop): Replace m with Bw.
8624 (*call_value_pop): Likewise.
8625 (*sibcall_pop_memory): Replace m with Bs.
8626 * config/i386/i386.opt (mindirect-branch-register): New option.
8627 * doc/invoke.texi: Document -mindirect-branch-register option.
8629 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8631 * config/i386/i386-protos.h (ix86_output_function_return): New.
8632 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8633 set function_return_type.
8634 (indirect_thunk_name): Add ret_p to indicate thunk for function
8636 (output_indirect_thunk_function): Pass false to
8637 indirect_thunk_name.
8638 (ix86_output_indirect_branch_via_reg): Likewise.
8639 (ix86_output_indirect_branch_via_push): Likewise.
8640 (output_indirect_thunk_function): Create alias for function
8641 return thunk if regno < 0.
8642 (ix86_output_function_return): New function.
8643 (ix86_handle_fndecl_attribute): Handle function_return.
8644 (ix86_attribute_table): Add function_return.
8645 * config/i386/i386.h (machine_function): Add
8646 function_return_type.
8647 * config/i386/i386.md (simple_return_internal): Use
8648 ix86_output_function_return.
8649 (simple_return_internal_long): Likewise.
8650 * config/i386/i386.opt (mfunction-return=): New option.
8651 (indirect_branch): Mention -mfunction-return=.
8652 * doc/extend.texi: Document function_return function attribute.
8653 * doc/invoke.texi: Document -mfunction-return= option.
8655 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8657 * config/i386/i386-opts.h (indirect_branch): New.
8658 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8659 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8660 with local indirect jump when converting indirect call and jump.
8661 (ix86_set_indirect_branch_type): New.
8662 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8663 (indirectlabelno): New.
8664 (indirect_thunk_needed): Likewise.
8665 (indirect_thunk_bnd_needed): Likewise.
8666 (indirect_thunks_used): Likewise.
8667 (indirect_thunks_bnd_used): Likewise.
8668 (INDIRECT_LABEL): Likewise.
8669 (indirect_thunk_name): Likewise.
8670 (output_indirect_thunk): Likewise.
8671 (output_indirect_thunk_function): Likewise.
8672 (ix86_output_indirect_branch_via_reg): Likewise.
8673 (ix86_output_indirect_branch_via_push): Likewise.
8674 (ix86_output_indirect_branch): Likewise.
8675 (ix86_output_indirect_jmp): Likewise.
8676 (ix86_code_end): Call output_indirect_thunk_function if needed.
8677 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8679 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8680 (ix86_attribute_table): Add indirect_branch.
8681 * config/i386/i386.h (machine_function): Add indirect_branch_type
8682 and has_local_indirect_jump.
8683 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8685 (tablejump): Likewise.
8686 (*indirect_jump): Use ix86_output_indirect_jmp.
8687 (*tablejump_1): Likewise.
8688 (simple_return_indirect_internal): Likewise.
8689 * config/i386/i386.opt (mindirect-branch=): New option.
8690 (indirect_branch): New.
8693 (thunk-inline): Likewise.
8694 (thunk-extern): Likewise.
8695 * doc/extend.texi: Document indirect_branch function attribute.
8696 * doc/invoke.texi: Document -mindirect-branch= option.
8698 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8701 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8703 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8705 * ipa-inline.c (want_inline_small_function_p): Return false if
8706 inlining has already failed with CIF_FINAL_ERROR.
8707 (update_caller_keys): Call want_inline_small_function_p before
8709 (update_callee_keys): Likewise.
8711 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8713 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8715 (rs6000_quadword_masked_address_p): Likewise.
8716 (quad_aligned_load_p): Likewise.
8717 (quad_aligned_store_p): Likewise.
8718 (const_load_sequence_p): Add comment to describe the outer-most loop.
8719 (mimic_memory_attributes_and_flags): New function.
8720 (rs6000_gen_stvx): Likewise.
8721 (replace_swapped_aligned_store): Likewise.
8722 (rs6000_gen_lvx): Likewise.
8723 (replace_swapped_aligned_load): Likewise.
8724 (replace_swapped_load_constant): Capitalize argument name in
8725 comment describing this function.
8726 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8727 and stores that access quad-word aligned addresses and replace
8728 with stvx or lvx instructions when appropriate.
8729 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8730 New function prototype.
8731 (rs6000_quadword_masked_address_p): Likewise.
8732 (rs6000_gen_lvx): Likewise.
8733 (rs6000_gen_stvx): Likewise.
8734 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8735 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8736 when memory address is aligned.
8737 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8738 this split to select lvx instruction when memory address is aligned.
8739 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8740 instruction when memory address is aligned.
8741 (*vsx_le_perm_load_v16qi): Likewise.
8742 (four unnamed splitters): Modify to select the stvx instruction
8743 when memory is aligned.
8745 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8747 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8748 which appears in the queue multiple times.
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 * tree-vectorizer.h (vec_lower_bound): New structure.
8755 (_loop_vec_info): Add check_nonzero and lower_bounds.
8756 (LOOP_VINFO_CHECK_NONZERO): New macro.
8757 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8758 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8759 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8760 fields. Make seg_len the distance travelled, not including the
8762 (dr_direction_indicator): Declare.
8763 (dr_zero_step_indicator): Likewise.
8764 (dr_known_forward_stride_p): Likewise.
8765 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8767 (runtime_alias_check_p): Allow runtime alias checks with
8769 (operator ==): Compare access_size and align.
8770 (prune_runtime_alias_test_list): Rework for new distinction between
8771 the access_size and seg_len.
8772 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8774 (get_segment_min_max): New function.
8775 (create_intersect_range_checks): Use it.
8776 (dr_step_indicator): New function.
8777 (dr_direction_indicator): Likewise.
8778 (dr_zero_step_indicator): Likewise.
8779 (dr_known_forward_stride_p): Likewise.
8780 * tree-loop-distribution.c (data_ref_segment_size): Return
8781 DR_STEP * (niters - 1).
8782 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8784 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8785 (vect_preserves_scalar_order_p): New function, split out from...
8786 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8787 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8788 (vect_vfa_access_size): New function.
8789 (vect_vfa_align): Likewise.
8790 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8791 (dump_lower_bound): New function.
8792 (vect_check_lower_bound): Likewise.
8793 (vect_small_gap_p): Likewise.
8794 (vectorizable_with_step_bound_p): Likewise.
8795 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8796 depencies if the vectorization factor is 1. Convert the checks
8797 for nonzero steps into checks on the bounds of DR_STEP. Try using
8798 a bunds check for variable steps if the minimum required step is
8799 relatively small. Update calls to the dr_with_seg_len
8800 constructor and to vect_compile_time_alias.
8801 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8803 (vect_loop_versioning): Call it.
8804 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8806 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8809 Alan Hayward <alan.hayward@arm.com>
8810 David Sherwood <david.sherwood@arm.com>
8812 * doc/sourcebuild.texi (vect_scatter_store): Document.
8813 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8815 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8817 * genopinit.c (main): Add supports_vec_scatter_store and
8818 supports_vec_scatter_store_cached to target_optabs.
8819 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8820 IFN_MASK_SCATTER_STORE.
8821 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8823 * internal-fn.h (internal_store_fn_p): Declare.
8824 (internal_fn_stored_value_index): Likewise.
8825 * internal-fn.c (scatter_store_direct): New macro.
8826 (expand_scatter_store_optab_fn): New function.
8827 (direct_scatter_store_optab_supported_p): New macro.
8828 (internal_store_fn_p): New function.
8829 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8830 IFN_MASK_SCATTER_STORE.
8831 (internal_fn_mask_index): Likewise.
8832 (internal_fn_stored_value_index): New function.
8833 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8835 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8836 * optabs-query.c (supports_vec_scatter_store_p): New function.
8837 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8838 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8839 true for scatter stores.
8840 (vect_gather_scatter_fn_p): Handle scatter stores too.
8841 (vect_check_gather_scatter): Consider using scatter stores if
8842 supports_vec_scatter_store_p.
8843 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8845 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8846 internal_fn_stored_value_index.
8847 (check_load_store_masking): Handle scatter stores too.
8848 (vect_get_store_rhs): Make public.
8849 (vectorizable_call): Use internal_store_fn_p.
8850 (vectorizable_store): Handle scatter store internal functions.
8851 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8852 when deciding whether the end of the group has been reached.
8853 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8854 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8855 (mask_scatter_store<mode>): New insns.
8857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8858 Alan Hayward <alan.hayward@arm.com>
8859 David Sherwood <david.sherwood@arm.com>
8861 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8862 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8863 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8865 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8866 Use vect_truncate_gather_scatter_offset if we can't treat the
8867 operation as a normal gather load or scatter store.
8868 (get_group_load_store_type): Take the gather_scatter_info
8869 as argument. Try using a gather load or scatter store for
8870 single-element groups.
8871 (get_load_store_type): Update calls to get_group_load_store_type
8872 and vect_use_strided_gather_scatters_p.
8874 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8875 Alan Hayward <alan.hayward@arm.com>
8876 David Sherwood <david.sherwood@arm.com>
8878 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8879 optional tree argument.
8880 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8882 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8883 but continue to use the current value as a fallback.
8884 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8885 to compare the updates.
8886 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8887 (get_load_store_type): Use it when handling a strided access.
8888 (vect_get_strided_load_store_ops): New function.
8889 (vect_get_data_ptr_increment): Likewise.
8890 (vectorizable_load): Handle strided gather loads. Always pass
8891 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8893 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8894 Alan Hayward <alan.hayward@arm.com>
8895 David Sherwood <david.sherwood@arm.com>
8897 * doc/md.texi (gather_load@var{m}): Document.
8898 (mask_gather_load@var{m}): Likewise.
8899 * genopinit.c (main): Add supports_vec_gather_load and
8900 supports_vec_gather_load_cached to target_optabs.
8901 * optabs-tree.c (init_tree_optimization_optabs): Use
8902 ggc_cleared_alloc to allocate target_optabs.
8903 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8904 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8906 * internal-fn.h (internal_load_fn_p): Declare.
8907 (internal_gather_scatter_fn_p): Likewise.
8908 (internal_fn_mask_index): Likewise.
8909 (internal_gather_scatter_fn_supported_p): Likewise.
8910 * internal-fn.c (gather_load_direct): New macro.
8911 (expand_gather_load_optab_fn): New function.
8912 (direct_gather_load_optab_supported_p): New macro.
8913 (direct_internal_fn_optab): New function.
8914 (internal_load_fn_p): Likewise.
8915 (internal_gather_scatter_fn_p): Likewise.
8916 (internal_fn_mask_index): Likewise.
8917 (internal_gather_scatter_fn_supported_p): Likewise.
8918 * optabs-query.c (supports_at_least_one_mode_p): New function.
8919 (supports_vec_gather_load_p): Likewise.
8920 * optabs-query.h (supports_vec_gather_load_p): Declare.
8921 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8922 and memory_type field.
8923 (NUM_PATTERNS): Bump to 15.
8924 * tree-vect-data-refs.c: Include internal-fn.h.
8925 (vect_gather_scatter_fn_p): New function.
8926 (vect_describe_gather_scatter_call): Likewise.
8927 (vect_check_gather_scatter): Try using internal functions for
8928 gather loads. Recognize existing calls to a gather load function.
8929 (vect_analyze_data_refs): Consider using gather loads if
8930 supports_vec_gather_load_p.
8931 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8932 (vect_get_gather_scatter_offset_type): Likewise.
8933 (vect_convert_mask_for_vectype): Likewise.
8934 (vect_add_conversion_to_patterm): Likewise.
8935 (vect_try_gather_scatter_pattern): Likewise.
8936 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8937 (vect_vect_recog_func_ptrs): Add it.
8938 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8939 internal_fn_mask_index and internal_gather_scatter_fn_p.
8940 (check_load_store_masking): Take the gather_scatter_info as an
8941 argument and handle gather loads.
8942 (vect_get_gather_scatter_ops): New function.
8943 (vectorizable_call): Check internal_load_fn_p.
8944 (vectorizable_load): Likewise. Handle gather load internal
8946 (vectorizable_store): Update call to check_load_store_masking.
8947 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8948 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8949 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8950 (aarch64_gather_scale_operand_d): New predicates.
8951 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8952 (mask_gather_load<mode>): New insns.
8954 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8955 Alan Hayward <alan.hayward@arm.com>
8956 David Sherwood <david.sherwood@arm.com>
8958 * optabs.def (fold_left_plus_optab): New optab.
8959 * doc/md.texi (fold_left_plus_@var{m}): Document.
8960 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8961 * internal-fn.c (fold_left_direct): Define.
8962 (expand_fold_left_optab_fn): Likewise.
8963 (direct_fold_left_optab_supported_p): Likewise.
8964 * fold-const-call.c (fold_const_fold_left): New function.
8965 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8966 * tree-parloops.c (valid_reduction_p): New function.
8967 (gather_scalar_reductions): Use it.
8968 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8969 (vect_finish_replace_stmt): Declare.
8970 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8971 (needs_fold_left_reduction_p): New function, split out from...
8972 (vect_is_simple_reduction): ...here. Accept reductions that
8973 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8974 (vect_force_simple_reduction): Also store the reduction type in
8975 the assignment's STMT_VINFO_REDUC_TYPE.
8976 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8977 (merge_with_identity): New function.
8978 (vect_expand_fold_left): Likewise.
8979 (vectorize_fold_left_reduction): Likewise.
8980 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8981 scalar phi in place for it. Check for target support and reject
8982 cases that would reassociate the operation. Defer the transform
8983 phase to vectorize_fold_left_reduction.
8984 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8985 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8986 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8988 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8990 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8991 call to ifc_temp_var.
8993 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8994 Alan Hayward <alan.hayward@arm.com>
8995 David Sherwood <david.sherwood@arm.com>
8997 * target.def (legitimize_address_displacement): Take the original
8998 offset as a poly_int.
8999 * targhooks.h (default_legitimize_address_displacement): Update
9001 * targhooks.c (default_legitimize_address_displacement): Likewise.
9002 * doc/tm.texi: Regenerate.
9003 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9004 as an argument, moving assert of ad->disp == ad->disp_term to...
9005 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9006 Try calling targetm.legitimize_address_displacement before expanding
9007 the address rather than afterwards, and adjust for the new interface.
9008 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9009 Match the new hook interface. Handle SVE addresses.
9010 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9015 * Makefile.in (OBJS): Add early-remat.o.
9016 * target.def (select_early_remat_modes): New hook.
9017 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9018 * doc/tm.texi: Regenerate.
9019 * targhooks.h (default_select_early_remat_modes): Declare.
9020 * targhooks.c (default_select_early_remat_modes): New function.
9021 * timevar.def (TV_EARLY_REMAT): New timevar.
9022 * passes.def (pass_early_remat): New pass.
9023 * tree-pass.h (make_pass_early_remat): Declare.
9024 * early-remat.c: New file.
9025 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9027 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9030 Alan Hayward <alan.hayward@arm.com>
9031 David Sherwood <david.sherwood@arm.com>
9033 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9034 vfm1 with a bound_epilog parameter.
9035 (vect_do_peeling): Update calls accordingly, and move the prologue
9036 call earlier in the function. Treat the base bound_epilog as 0 for
9037 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9038 this base when peeling for gaps.
9039 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9040 with fully-masked loops.
9041 (vect_estimate_min_profitable_iters): Handle the single peeled
9042 iteration in that case.
9044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9045 Alan Hayward <alan.hayward@arm.com>
9046 David Sherwood <david.sherwood@arm.com>
9048 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9049 single-element interleaving even if the size is not a power of 2.
9050 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9051 accesses for single-element interleaving if the group size is
9054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9055 Alan Hayward <alan.hayward@arm.com>
9056 David Sherwood <david.sherwood@arm.com>
9058 * doc/md.texi (fold_extract_last_@var{m}): Document.
9059 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9060 * optabs.def (fold_extract_last_optab): New optab.
9061 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9062 * internal-fn.c (fold_extract_direct): New macro.
9063 (expand_fold_extract_optab_fn): Likewise.
9064 (direct_fold_extract_optab_supported_p): Likewise.
9065 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9066 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9067 EXTRACT_LAST_REDUCTION.
9068 (get_initial_def_for_reduction): Do not create an initial vector
9069 for EXTRACT_LAST_REDUCTION reductions.
9070 (vectorizable_reduction): Leave the scalar phi in place for
9071 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9072 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9073 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9074 transform phase to vectorizable_condition.
9075 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9077 (vect_finish_stmt_generation): ...here.
9078 (vect_finish_replace_stmt): New function.
9079 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9080 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9082 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9084 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9085 Alan Hayward <alan.hayward@arm.com>
9086 David Sherwood <david.sherwood@arm.com>
9088 * doc/md.texi (extract_last_@var{m}): Document.
9089 * optabs.def (extract_last_optab): New optab.
9090 * internal-fn.def (EXTRACT_LAST): New internal function.
9091 * internal-fn.c (cond_unary_direct): New macro.
9092 (expand_cond_unary_optab_fn): Likewise.
9093 (direct_cond_unary_optab_supported_p): Likewise.
9094 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9095 loops using EXTRACT_LAST.
9096 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9097 (extract_last_<mode>): ...this optab.
9098 (vec_extract<mode><Vel>): Update accordingly.
9100 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9101 Alan Hayward <alan.hayward@arm.com>
9102 David Sherwood <david.sherwood@arm.com>
9104 * target.def (empty_mask_is_expensive): New hook.
9105 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9106 * doc/tm.texi: Regenerate.
9107 * targhooks.h (default_empty_mask_is_expensive): Declare.
9108 * targhooks.c (default_empty_mask_is_expensive): New function.
9109 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9110 if the target says that empty masks are expensive.
9111 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9113 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9116 Alan Hayward <alan.hayward@arm.com>
9117 David Sherwood <david.sherwood@arm.com>
9119 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9120 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9121 (vect_use_loop_mask_for_alignment_p): New function.
9122 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9123 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9124 niters_skip argument. Make sure that the first niters_skip elements
9125 of the first iteration are inactive.
9126 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9127 Update call to vect_set_loop_masks_directly.
9128 (get_misalign_in_elems): New function, split out from...
9129 (vect_gen_prolog_loop_niters): ...here.
9130 (vect_update_init_of_dr): Take a code argument that specifies whether
9131 the adjustment should be added or subtracted.
9132 (vect_update_init_of_drs): Likewise.
9133 (vect_prepare_for_masked_peels): New function.
9134 (vect_do_peeling): Skip prologue peeling if we're using a mask
9135 instead. Update call to vect_update_inits_of_drs.
9136 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9138 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9139 alignment. Do not include the number of peeled iterations in
9140 the minimum threshold in that case.
9141 (vectorizable_induction): Adjust the start value down by
9142 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9143 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9144 Take the number of skipped iterations into account when calculating
9146 * tree-vect-stmts.c (vect_gen_while_not): New function.
9148 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9149 Alan Hayward <alan.hayward@arm.com>
9150 David Sherwood <david.sherwood@arm.com>
9152 * doc/sourcebuild.texi (vect_fully_masked): Document.
9153 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9155 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9157 (vect_analyze_loop_2): ...here. Don't check the vectorization
9158 factor against the number of loop iterations if the loop is
9161 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9162 Alan Hayward <alan.hayward@arm.com>
9163 David Sherwood <david.sherwood@arm.com>
9165 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9166 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9167 (dump_groups): Update accordingly.
9168 (iv_use::mem_type): New member variable.
9169 (address_p): New function.
9170 (record_use): Add a mem_type argument and initialize the new
9172 (record_group_use): Add a mem_type argument. Use address_p.
9173 Remove obsolete null checks of base_object. Update call to record_use.
9174 (find_interesting_uses_op): Update call to record_group_use.
9175 (find_interesting_uses_cond): Likewise.
9176 (find_interesting_uses_address): Likewise.
9177 (get_mem_type_for_internal_fn): New function.
9178 (find_address_like_use): Likewise.
9179 (find_interesting_uses_stmt): Try find_address_like_use before
9180 calling find_interesting_uses_op.
9181 (addr_offset_valid_p): Use the iv mem_type field as the type
9182 of the addressed memory.
9183 (add_autoinc_candidates): Likewise.
9184 (get_address_cost): Likewise.
9185 (split_small_address_groups_p): Use address_p.
9186 (split_address_groups): Likewise.
9187 (add_iv_candidate_for_use): Likewise.
9188 (autoinc_possible_for_pair): Likewise.
9189 (rewrite_groups): Likewise.
9190 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9191 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9192 (get_alias_ptr_type_for_ptr_address): New function.
9193 (rewrite_use_address): Rewrite address uses in calls that were
9194 identified by find_address_like_use.
9196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9197 Alan Hayward <alan.hayward@arm.com>
9198 David Sherwood <david.sherwood@arm.com>
9200 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9202 * gimple-expr.h (is_gimple_addressable: Likewise.
9203 * gimple-expr.c (is_gimple_address): Likewise.
9204 * internal-fn.c (expand_call_mem_ref): New function.
9205 (expand_mask_load_optab_fn): Use it.
9206 (expand_mask_store_optab_fn): Likewise.
9208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9209 Alan Hayward <alan.hayward@arm.com>
9210 David Sherwood <david.sherwood@arm.com>
9212 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9213 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9214 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9215 (cond_umax@var{mode}): Document.
9216 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9217 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9218 (cond_umin_optab, cond_umax_optab): New optabs.
9219 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9220 (COND_IOR, COND_XOR): New internal functions.
9221 * internal-fn.h (get_conditional_internal_fn): Declare.
9222 * internal-fn.c (cond_binary_direct): New macro.
9223 (expand_cond_binary_optab_fn): Likewise.
9224 (direct_cond_binary_optab_supported_p): Likewise.
9225 (get_conditional_internal_fn): New function.
9226 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9227 Cope with reduction statements that are vectorized as calls rather
9229 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9230 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9231 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9232 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9233 (UNSPEC_COND_EOR): New unspecs.
9234 (optab): Add mappings for them.
9235 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9236 (sve_int_op, sve_fp_op): New int attributes.
9238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9239 Alan Hayward <alan.hayward@arm.com>
9240 David Sherwood <david.sherwood@arm.com>
9242 * optabs.def (while_ult_optab): New optab.
9243 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9244 * internal-fn.def (WHILE_ULT): New internal function.
9245 * internal-fn.h (direct_internal_fn_supported_p): New override
9246 that takes two types as argument.
9247 * internal-fn.c (while_direct): New macro.
9248 (expand_while_optab_fn): New function.
9249 (convert_optab_supported_p): Likewise.
9250 (direct_while_optab_supported_p): New macro.
9251 * wide-int.h (wi::udiv_ceil): New function.
9252 * tree-vectorizer.h (rgroup_masks): New structure.
9253 (vec_loop_masks): New typedef.
9254 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9256 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9257 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9258 (vect_max_vf): New function.
9259 (slpeel_make_loop_iterate_ntimes): Delete.
9260 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9261 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9262 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9263 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9264 internal-fn.h, stor-layout.h and optabs-query.h.
9265 (vect_set_loop_mask): New function.
9266 (add_preheader_seq): Likewise.
9267 (add_header_seq): Likewise.
9268 (interleave_supported_p): Likewise.
9269 (vect_maybe_permute_loop_masks): Likewise.
9270 (vect_set_loop_masks_directly): Likewise.
9271 (vect_set_loop_condition_masked): Likewise.
9272 (vect_set_loop_condition_unmasked): New function, split out from
9273 slpeel_make_loop_iterate_ntimes.
9274 (slpeel_make_loop_iterate_ntimes): Rename to..
9275 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9276 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9277 (vect_do_peeling): Update call accordingly.
9278 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9280 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9281 mask_compare_type, can_fully_mask_p and fully_masked_p.
9282 (release_vec_loop_masks): New function.
9283 (_loop_vec_info): Use it to free the loop masks.
9284 (can_produce_all_loop_masks_p): New function.
9285 (vect_get_max_nscalars_per_iter): Likewise.
9286 (vect_verify_full_masking): Likewise.
9287 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9288 retries, and free the mask rgroups before retrying. Check loop-wide
9289 reasons for disallowing fully-masked loops. Make the final decision
9290 about whether use a fully-masked loop or not.
9291 (vect_estimate_min_profitable_iters): Do not assume that peeling
9292 for the number of iterations will be needed for fully-masked loops.
9293 (vectorizable_reduction): Disable fully-masked loops.
9294 (vectorizable_live_operation): Likewise.
9295 (vect_halve_mask_nunits): New function.
9296 (vect_double_mask_nunits): Likewise.
9297 (vect_record_loop_mask): Likewise.
9298 (vect_get_loop_mask): Likewise.
9299 (vect_transform_loop): Handle the case in which the final loop
9300 iteration might handle a partial vector. Call vect_set_loop_condition
9301 instead of slpeel_make_loop_iterate_ntimes.
9302 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9303 (check_load_store_masking): New function.
9304 (prepare_load_store_mask): Likewise.
9305 (vectorizable_store): Handle fully-masked loops.
9306 (vectorizable_load): Likewise.
9307 (supportable_widening_operation): Use vect_halve_mask_nunits for
9309 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9310 (vect_gen_while): New function.
9311 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9312 (aarch64_uqdec<mode>): New insn.
9314 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9315 Alan Hayward <alan.hayward@arm.com>
9316 David Sherwood <david.sherwood@arm.com>
9318 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9319 (reduc_xor_scal_optab): New optabs.
9320 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9321 (reduc_xor_scal_@var{m}): Document.
9322 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9323 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9325 * fold-const-call.c (fold_const_call): Handle them.
9326 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9327 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9328 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9329 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9330 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9331 (UNSPEC_XORV): New unspecs.
9332 (optab): Add entries for them.
9333 (BITWISEV): New int iterator.
9334 (bit_reduc_op): New int attributes.
9336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9337 Alan Hayward <alan.hayward@arm.com>
9338 David Sherwood <david.sherwood@arm.com>
9340 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9341 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9342 * optabs.def (vec_shl_insert_optab): New optab.
9343 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9344 (duplicate_and_interleave): Likewise.
9345 * tree-vect-loop.c: Include internal-fn.h.
9346 (neutral_op_for_slp_reduction): New function, split out from
9347 get_initial_defs_for_reduction.
9348 (get_initial_def_for_reduction): Handle option 2 for variable-length
9349 vectors by loading the neutral value into a vector and then shifting
9350 the initial value into element 0.
9351 (get_initial_defs_for_reduction): Replace the code argument with
9352 the neutral value calculated by neutral_op_for_slp_reduction.
9353 Use gimple_build_vector for constant-length vectors.
9354 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9355 but the first group_size elements have a neutral value.
9356 Use duplicate_and_interleave otherwise.
9357 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9358 Update call to get_initial_defs_for_reduction. Handle SLP
9359 reductions for variable-length vectors by creating one vector
9360 result for each scalar result, with the elements associated
9361 with other scalar results stubbed out with the neutral value.
9362 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9363 Require IFN_VEC_SHL_INSERT for double reductions on
9364 variable-length vectors, or SLP reductions that have
9365 a neutral value. Require can_duplicate_and_interleave_p
9366 support for variable-length unchained SLP reductions if there
9367 is no neutral value, such as for MIN/MAX reductions. Also require
9368 the number of vector elements to be a multiple of the number of
9369 SLP statements when doing variable-length unchained SLP reductions.
9370 Update call to vect_create_epilog_for_reduction.
9371 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9372 and remove initial values.
9373 (duplicate_and_interleave): Make public.
9374 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9375 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9377 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9378 Alan Hayward <alan.hayward@arm.com>
9379 David Sherwood <david.sherwood@arm.com>
9381 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9382 (can_duplicate_and_interleave_p): New function.
9383 (vect_get_and_check_slp_defs): Take the vector of statements
9384 rather than just the current one. Remove excess parentheses.
9385 Restriction rejectinon of vect_constant_def and vect_external_def
9386 for variable-length vectors to boolean types, or types for which
9387 can_duplicate_and_interleave_p is false.
9388 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9389 (duplicate_and_interleave): New function.
9390 (vect_get_constant_vectors): Use gimple_build_vector for
9391 constant-length vectors and suitable variable-length constant
9392 vectors. Use duplicate_and_interleave for other variable-length
9393 vectors. Don't defer the update when inserting new statements.
9395 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9396 Alan Hayward <alan.hayward@arm.com>
9397 David Sherwood <david.sherwood@arm.com>
9399 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9400 min_profitable_iters doesn't go negative.
9402 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9403 Alan Hayward <alan.hayward@arm.com>
9404 David Sherwood <david.sherwood@arm.com>
9406 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9407 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9408 * optabs.def (vec_mask_load_lanes_optab): New optab.
9409 (vec_mask_store_lanes_optab): Likewise.
9410 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9411 (MASK_STORE_LANES): Likewise.
9412 * internal-fn.c (mask_load_lanes_direct): New macro.
9413 (mask_store_lanes_direct): Likewise.
9414 (expand_mask_load_optab_fn): Handle masked operations.
9415 (expand_mask_load_lanes_optab_fn): New macro.
9416 (expand_mask_store_optab_fn): Handle masked operations.
9417 (expand_mask_store_lanes_optab_fn): New macro.
9418 (direct_mask_load_lanes_optab_supported_p): Likewise.
9419 (direct_mask_store_lanes_optab_supported_p): Likewise.
9420 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9422 (vect_load_lanes_supported): Likewise.
9423 * tree-vect-data-refs.c (strip_conversion): New function.
9424 (can_group_stmts_p): Likewise.
9425 (vect_analyze_data_ref_accesses): Use it instead of checking
9426 for a pair of assignments.
9427 (vect_store_lanes_supported): Take a masked_p parameter.
9428 (vect_load_lanes_supported): Likewise.
9429 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9430 vect_store_lanes_supported and vect_load_lanes_supported.
9431 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9432 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9433 parameter. Don't allow gaps for masked accesses.
9434 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9435 and vect_load_lanes_supported.
9436 (get_load_store_type): Take a masked_p parameter and update
9437 call to get_group_load_store_type.
9438 (vectorizable_store): Update call to get_load_store_type.
9439 Handle IFN_MASK_STORE_LANES.
9440 (vectorizable_load): Update call to get_load_store_type.
9441 Handle IFN_MASK_LOAD_LANES.
9443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9444 Alan Hayward <alan.hayward@arm.com>
9445 David Sherwood <david.sherwood@arm.com>
9447 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9449 * config/aarch64/aarch64-protos.h
9450 (aarch64_sve_struct_memory_operand_p): Declare.
9451 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9452 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9453 (VPRED, vpred): Handle SVE structure modes.
9454 * config/aarch64/constraints.md (Utx): New constraint.
9455 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9456 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9457 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9458 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9459 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9460 structure modes. Split into pieces after RA.
9461 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9462 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9464 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9465 SVE structure modes.
9466 (aarch64_classify_address): Likewise.
9467 (sizetochar): Move earlier in file.
9468 (aarch64_print_operand): Handle SVE register lists.
9469 (aarch64_array_mode): New function.
9470 (aarch64_sve_struct_memory_operand_p): Likewise.
9471 (TARGET_ARRAY_MODE): Redefine.
9473 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9474 Alan Hayward <alan.hayward@arm.com>
9475 David Sherwood <david.sherwood@arm.com>
9477 * target.def (array_mode): New target hook.
9478 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9479 * doc/tm.texi: Regenerate.
9480 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9481 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9482 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9484 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9488 Alan Hayward <alan.hayward@arm.com>
9489 David Sherwood <david.sherwood@arm.com>
9491 * fold-const.c (fold_binary_loc): Check the argument types
9492 rather than the result type when testing for a vector operation.
9494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9496 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9497 * doc/tm.texi: Regenerate.
9499 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9500 Alan Hayward <alan.hayward@arm.com>
9501 David Sherwood <david.sherwood@arm.com>
9503 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9504 (sve): Document new AArch64 extension.
9505 * doc/md.texi (w): Extend the description of the AArch64
9506 constraint to include SVE vectors.
9507 (Upl, Upa): Document new AArch64 predicate constraints.
9508 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9510 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9511 (msve-vector-bits=): New option.
9512 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9513 SVE when these are disabled.
9514 (sve): New extension.
9515 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9516 modes. Adjust their number of units based on aarch64_sve_vg.
9517 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9518 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9519 aarch64_addr_query_type.
9520 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9521 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9522 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9523 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9524 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9525 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9526 (aarch64_simd_imm_zero_p): Delete.
9527 (aarch64_check_zero_based_sve_index_immediate): Declare.
9528 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9529 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9530 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9531 (aarch64_sve_float_mul_immediate_p): Likewise.
9532 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9534 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9535 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9536 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9537 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9538 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9539 (aarch64_regmode_natural_size): Likewise.
9540 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9541 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9543 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9544 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9545 for VG and the SVE predicate registers.
9546 (V_ALIASES): Add a "z"-prefixed alias.
9547 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9548 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9549 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9550 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9551 (REG_CLASS_NAMES): Add entries for them.
9552 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9553 and the predicate registers.
9554 (aarch64_sve_vg): Declare.
9555 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9556 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9557 (REGMODE_NATURAL_SIZE): Define.
9558 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9560 * config/aarch64/aarch64.c: Include cfgrtl.h.
9561 (simd_immediate_info): Add a constructor for series vectors,
9562 and an associated step field.
9563 (aarch64_sve_vg): New variable.
9564 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9565 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9566 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9567 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9568 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9569 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9570 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9571 (aarch64_get_mask_mode): New functions.
9572 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9573 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9574 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9575 predicate modes and predicate registers. Explicitly restrict
9576 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9577 to store a vector mode if it is recognized by
9578 aarch64_classify_vector_mode.
9579 (aarch64_regmode_natural_size): New function.
9580 (aarch64_hard_regno_caller_save_mode): Return the original mode
9582 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9583 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9584 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9585 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9587 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9588 does not overlap dest if the function is frame-related. Handle
9590 (aarch64_split_add_offset): New function.
9591 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9592 them aarch64_add_offset.
9593 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9594 and update call to aarch64_sub_sp.
9595 (aarch64_add_cfa_expression): New function.
9596 (aarch64_expand_prologue): Pass extra temporary registers to the
9597 functions above. Handle the case in which we need to emit new
9598 DW_CFA_expressions for registers that were originally saved
9599 relative to the stack pointer, but now have to be expressed
9600 relative to the frame pointer.
9601 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9603 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9604 IP0 and IP1 values for SVE frames.
9605 (aarch64_expand_vec_series): New function.
9606 (aarch64_expand_sve_widened_duplicate): Likewise.
9607 (aarch64_expand_sve_const_vector): Likewise.
9608 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9609 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9610 into the register, rather than emitting a SET directly.
9611 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9612 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9613 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9614 (offset_9bit_signed_scaled_p): New functions.
9615 (aarch64_replicate_bitmask_imm): New function.
9616 (aarch64_bitmask_imm): Use it.
9617 (aarch64_cannot_force_const_mem): Reject expressions involving
9618 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9619 (aarch64_classify_index): Handle SVE indices, by requiring
9620 a plain register index with a scale that matches the element size.
9621 (aarch64_classify_address): Handle SVE addresses. Assert that
9622 the mode of the address is VOIDmode or an integer mode.
9623 Update call to aarch64_classify_symbol.
9624 (aarch64_classify_symbolic_expression): Update call to
9625 aarch64_classify_symbol.
9626 (aarch64_const_vec_all_in_range_p): New function.
9627 (aarch64_print_vector_float_operand): Likewise.
9628 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9629 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9630 and the FP immediates 1.0 and 0.5.
9631 (aarch64_print_address_internal): Handle SVE addresses.
9632 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9633 (aarch64_regno_regclass): Handle predicate registers.
9634 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9636 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9637 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9638 (aarch64_convert_sve_vector_bits): New function.
9639 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9640 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9642 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9643 Handle SVE vector and predicate modes. Accept VL-based constants
9644 that need only one temporary register, and VL offsets that require
9645 no temporary registers.
9646 (aarch64_conditional_register_usage): Mark the predicate registers
9647 as fixed if SVE isn't available.
9648 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9649 Return true for SVE vector and predicate modes.
9650 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9651 rather than an unsigned int. Handle SVE modes.
9652 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9654 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9656 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9657 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9658 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9659 (aarch64_sve_float_mul_immediate_p): New functions.
9660 (aarch64_sve_valid_immediate): New function.
9661 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9662 Explicitly reject structure modes. Check for INDEX constants.
9663 Handle PTRUE and PFALSE constants.
9664 (aarch64_check_zero_based_sve_index_immediate): New function.
9665 (aarch64_simd_imm_zero_p): Delete.
9666 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9667 vector modes. Accept constants in the range of CNT[BHWD].
9668 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9669 ask for an Advanced SIMD mode.
9670 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9671 (aarch64_simd_vector_alignment): Handle SVE predicates.
9672 (aarch64_vectorize_preferred_vector_alignment): New function.
9673 (aarch64_simd_vector_alignment_reachable): Use it instead of
9675 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9676 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9678 (MAX_VECT_LEN): Delete.
9679 (expand_vec_perm_d): Add a vec_flags field.
9680 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9681 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9682 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9684 (aarch64_evpc_rev): Rename to...
9685 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9686 (aarch64_evpc_rev_global): New function.
9687 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9688 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9690 (aarch64_evpc_sve_tbl): New function.
9691 (aarch64_expand_vec_perm_const_1): Update after rename of
9692 aarch64_evpc_rev. Handle SVE permutes too, trying
9693 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9694 than aarch64_evpc_tbl.
9695 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9696 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9697 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9698 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9699 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9700 (aarch64_expand_sve_vcond): New functions.
9701 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9702 of aarch64_vector_mode_p.
9703 (aarch64_dwarf_poly_indeterminate_value): New function.
9704 (aarch64_compute_pressure_classes): Likewise.
9705 (aarch64_can_change_mode_class): Likewise.
9706 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9707 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9708 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9709 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9710 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9711 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9712 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9713 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9715 (Dn, Dl, Dr): Accept const as well as const_vector.
9716 (Dz): Likewise. Compare against CONST0_RTX.
9717 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9718 of "vector" where appropriate.
9719 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9720 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9721 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9722 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9723 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9724 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9725 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9726 (v_int_equiv): Extend to SVE modes.
9727 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9729 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9730 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9731 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9732 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9733 (SVE_COND_FP_CMP): New int iterators.
9734 (perm_hilo): Handle the new unpack unspecs.
9735 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9737 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9738 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9739 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9740 (aarch64_equality_operator, aarch64_constant_vector_operand)
9741 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9742 (aarch64_sve_nonimmediate_operand): Likewise.
9743 (aarch64_sve_general_operand): Likewise.
9744 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9745 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9746 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9747 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9748 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9749 (aarch64_sve_float_arith_immediate): Likewise.
9750 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9751 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9752 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9753 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9754 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9755 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9756 (aarch64_sve_float_arith_operand): Likewise.
9757 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9758 (aarch64_sve_float_mul_operand): Likewise.
9759 (aarch64_sve_vec_perm_operand): Likewise.
9760 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9761 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9762 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9763 as well as const_vector.
9764 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9765 in file. Use CONST0_RTX and CONSTM1_RTX.
9766 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9767 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9768 Use aarch64_simd_imm_zero.
9769 * config/aarch64/aarch64-sve.md: New file.
9770 * config/aarch64/aarch64.md: Include it.
9771 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9772 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9773 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9774 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9775 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9776 (sve): New attribute.
9777 (enabled): Disable instructions with the sve attribute unless
9779 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9780 aarch64_expand_mov_immediate.
9781 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9782 CNT[BHSD] immediates.
9783 (movti): Split CONST_POLY_INT moves into two halves.
9784 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9785 Split additions that need a temporary here if the destination
9786 is the stack pointer.
9787 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9788 (*add<mode>3_poly_1): New instruction.
9789 (set_clobber_cc): New expander.
9791 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9793 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9794 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9795 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9796 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9797 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9798 Change innermode from fixed_mode_size to machine_mode.
9799 (simplify_subreg): Update call accordingly. Handle a constant-sized
9800 subreg of a variable-length CONST_VECTOR.
9802 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9803 Alan Hayward <alan.hayward@arm.com>
9804 David Sherwood <david.sherwood@arm.com>
9806 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9807 (add_offset_to_base): New function, split out from...
9808 (create_mem_ref): ...here. When handling a scale other than 1,
9809 check first whether the address is valid without the offset.
9810 Add it into the base if so, leaving the index and scale as-is.
9812 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9815 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9816 fold_for_warn before checking if arg2 is INTEGER_CST.
9818 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9820 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9821 (store_multiple_operation): Delete.
9822 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9823 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9824 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9825 guarded by TARGET_STRING.
9826 (rs6000_output_load_multiple): Delete.
9827 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9828 OPTION_MASK_STRING / TARGET_STRING handling.
9829 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9830 (const rs6000_opt_masks) <"string">: Change mask to 0.
9831 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9832 (MASK_STRING): Delete.
9833 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9835 (load_multiple): Delete.
9842 (store_multiple): Delete.
9849 (movmemsi_8reg): Delete.
9850 (corresponding unnamed define_insn): Delete.
9851 (movmemsi_6reg): Delete.
9852 (corresponding unnamed define_insn): Delete.
9853 (movmemsi_4reg): Delete.
9854 (corresponding unnamed define_insn): Delete.
9855 (movmemsi_2reg): Delete.
9856 (corresponding unnamed define_insn): Delete.
9857 (movmemsi_1reg): Delete.
9858 (corresponding unnamed define_insn): Delete.
9859 * config/rs6000/rs6000.opt (mno-string): New.
9860 (mstring): Replace by deprecation warning stub.
9861 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9863 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9865 * regrename.c (regrename_do_replace): If replacing the same
9866 reg multiple times, try to reuse last created gen_raw_REG.
9869 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9870 main to workaround a bug in GDB.
9872 2018-01-12 Tom de Vries <tom@codesourcery.com>
9875 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9877 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9879 PR rtl-optimization/80481
9880 * ira-color.c (get_cap_member): New function.
9881 (allocnos_conflict_by_live_ranges_p): Use it.
9882 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9883 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9885 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9888 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9889 (*saddl_se_1): Ditto.
9891 (*ssubl_se_1): Ditto.
9893 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9895 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9896 rather than wi::to_widest for DR_INITs.
9897 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9898 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9899 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9901 (vect_analyze_group_access_1): Note that here.
9903 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9905 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9906 polynomial type sizes.
9908 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9910 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9911 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9912 (gimple_add_tmp_var): Likewise.
9914 2018-01-12 Martin Liska <mliska@suse.cz>
9916 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9917 (gimple_alloc_sizes): Likewise.
9918 (dump_gimple_statistics): Use PRIu64 in printf format.
9919 * gimple.h: Change uint64_t to int.
9921 2018-01-12 Martin Liska <mliska@suse.cz>
9923 * tree-core.h: Use uint64_t instead of int.
9924 * tree.c (tree_node_counts): Likewise.
9925 (tree_node_sizes): Likewise.
9926 (dump_tree_statistics): Use PRIu64 in printf format.
9928 2018-01-12 Martin Liska <mliska@suse.cz>
9930 * Makefile.in: As qsort_chk is implemented in vec.c, add
9931 vec.o to linkage of gencfn-macros.
9932 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9933 passing the info to record_node_allocation_statistics.
9934 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9936 * ggc-common.c (struct ggc_usage): Add operator== and use
9937 it in operator< and compare function.
9938 * mem-stats.h (struct mem_usage): Likewise.
9939 * vec.c (struct vec_usage): Remove operator< and compare
9940 function. Can be simply inherited.
9942 2018-01-12 Martin Jambor <mjambor@suse.cz>
9945 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9946 * tree-ssa-math-opts.c: Include domwalk.h.
9947 (convert_mult_to_fma_1): New function.
9948 (fma_transformation_info): New type.
9949 (fma_deferring_state): Likewise.
9950 (cancel_fma_deferring): New function.
9951 (result_of_phi): Likewise.
9952 (last_fma_candidate_feeds_initial_phi): Likewise.
9953 (convert_mult_to_fma): Added deferring logic, split actual
9954 transformation to convert_mult_to_fma_1.
9955 (math_opts_dom_walker): New type.
9956 (math_opts_dom_walker::after_dom_children): New method, body moved
9957 here from pass_optimize_widening_mul::execute, added deferring logic
9959 (pass_optimize_widening_mul::execute): Moved most of code to
9960 math_opts_dom_walker::after_dom_children.
9961 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9962 * config/i386/i386.c (ix86_option_override_internal): Added
9963 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9965 2018-01-12 Richard Biener <rguenther@suse.de>
9968 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9969 inline instance vars.
9971 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9974 * config/rx/rx.c (rx_is_restricted_memory_address):
9977 2018-01-12 Richard Biener <rguenther@suse.de>
9979 PR tree-optimization/80846
9980 * target.def (split_reduction): New target hook.
9981 * targhooks.c (default_split_reduction): New function.
9982 * targhooks.h (default_split_reduction): Declare.
9983 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9984 target requests first reduce vectors by combining low and high
9986 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9987 (get_vectype_for_scalar_type_and_size): Export.
9988 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9989 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9990 * doc/tm.texi: Regenerate.
9991 * config/i386/i386.c (ix86_split_reduction): Implement
9992 TARGET_VECTORIZE_SPLIT_REDUCTION.
9994 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9997 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9998 in PIC mode except for TARGET_VXWORKS_RTP.
9999 * config/sparc/sparc.c: Include cfgrtl.h.
10000 (TARGET_INIT_PIC_REG): Define.
10001 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10002 (sparc_pic_register_p): New predicate.
10003 (sparc_legitimate_address_p): Use it.
10004 (sparc_legitimize_pic_address): Likewise.
10005 (sparc_delegitimize_address): Likewise.
10006 (sparc_mode_dependent_address_p): Likewise.
10007 (gen_load_pcrel_sym): Remove 4th parameter.
10008 (load_got_register): Adjust call to above. Remove obsolete stuff.
10009 (sparc_expand_prologue): Do not call load_got_register here.
10010 (sparc_flat_expand_prologue): Likewise.
10011 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10012 (sparc_use_pseudo_pic_reg): New function.
10013 (sparc_init_pic_reg): Likewise.
10014 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10015 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10017 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10019 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10020 Add item for branch_cost.
10022 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10024 PR rtl-optimization/83565
10025 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10026 not extend the result to a larger mode for rotate operations.
10027 (num_sign_bit_copies1): Likewise.
10029 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10032 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10034 Use values-Xc.o for -pedantic.
10035 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10037 2018-01-12 Martin Liska <mliska@suse.cz>
10040 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10042 (possible_polymorphic_call_targets): Use it.
10043 (ipa_devirt): Likewise.
10045 2018-01-12 Martin Liska <mliska@suse.cz>
10047 * profile-count.h (enum profile_quality): Use 0 as invalid
10048 enum value of profile_quality.
10050 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10052 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10053 -mext-string options.
10055 2018-01-12 Richard Biener <rguenther@suse.de>
10057 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10058 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10059 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10061 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10063 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10065 * configure.ac (--with-long-double-format): Add support for the
10066 configuration option to change the default long double format on
10068 * config.gcc (powerpc*-linux*-*): Likewise.
10069 * configure: Regenerate.
10070 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10071 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10072 used without modification.
10074 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10076 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10077 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10078 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10079 MISC_BUILTIN_SPEC_BARRIER.
10080 (rs6000_init_builtins): Likewise.
10081 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10083 (speculation_barrier): New define_insn.
10084 * doc/extend.texi: Document __builtin_speculation_barrier.
10086 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10089 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10090 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10091 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10093 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10094 integral modes instead of "ss" and "sd".
10095 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10096 vectors with 32-bit and 64-bit elements.
10097 (vecdupssescalarmodesuffix): New mode attribute.
10098 (vec_dup<mode>): Use it.
10100 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10103 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10104 frame if argument is passed on stack.
10106 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10109 * ree.c (combine_reaching_defs): Optimize also
10110 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10111 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10113 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10115 PR middle-end/83189
10116 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10118 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10120 PR middle-end/83718
10121 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10122 after they are computed.
10124 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10126 PR tree-optimization/83695
10127 * gimple-loop-linterchange.cc
10128 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10129 reset cached scev information after interchange.
10130 (pass_linterchange::execute): Remove call to scev_reset_htab.
10132 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10134 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10135 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10136 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10137 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10138 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10139 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10140 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10141 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10142 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10143 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10144 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10145 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10146 (V_lane_reg): Likewise.
10147 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10149 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10150 (vfmal_lane_low<mode>_intrinsic,
10151 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10152 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10153 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10154 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10155 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10156 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10158 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10160 * config/arm/arm-cpus.in (fp16fml): New feature.
10161 (ALL_SIMD): Add fp16fml.
10162 (armv8.2-a): Add fp16fml as an option.
10163 (armv8.3-a): Likewise.
10164 (armv8.4-a): Add fp16fml as part of fp16.
10165 * config/arm/arm.h (TARGET_FP16FML): Define.
10166 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10168 * config/arm/arm-modes.def (V2HF): Define.
10169 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10170 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10171 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10172 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10173 vfmsl_low, vfmsl_high): New set of builtins.
10174 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10175 (vfml_op): New code attribute.
10176 (VFMLHALVES): New int iterator.
10177 (VFML, VFMLSEL): New mode attributes.
10178 (V_reg): Define mapping for V2HF.
10179 (V_hi, V_lo): New mode attributes.
10180 (VF_constraint): Likewise.
10181 (vfml_half, vfml_half_selector): New int attributes.
10182 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10184 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10185 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10187 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10188 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10189 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10190 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10192 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10193 Document new effective target and option set.
10195 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10197 * config/arm/arm-cpus.in (armv8_4): New feature.
10198 (ARMv8_4a): New fgroup.
10199 (armv8.4-a): New arch.
10200 * config/arm/arm-tables.opt: Regenerate.
10201 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10202 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10203 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10204 Add matching rules for -march=armv8.4-a and extensions.
10205 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10207 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10210 * config/rx/rx.md (BW): New mode attribute.
10211 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10213 2018-01-11 Richard Biener <rguenther@suse.de>
10215 PR tree-optimization/83435
10216 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10217 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10218 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10220 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10221 Alan Hayward <alan.hayward@arm.com>
10222 David Sherwood <david.sherwood@arm.com>
10224 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10226 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10227 (aarch64_print_address_internal): Use it to check for a zero offset.
10229 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10230 Alan Hayward <alan.hayward@arm.com>
10231 David Sherwood <david.sherwood@arm.com>
10233 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10234 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10235 Return a poly_int64 rather than a HOST_WIDE_INT.
10236 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10237 rather than a HOST_WIDE_INT.
10238 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10239 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10240 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10241 final_offset from HOST_WIDE_INT to poly_int64.
10242 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10243 to_constant when getting the number of units in an Advanced SIMD
10245 (aarch64_builtin_vectorized_function): Check for a constant number
10247 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10249 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10250 attribute instead of GET_MODE_NUNITS.
10251 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10252 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10253 GET_MODE_SIZE for fixed-size registers.
10254 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10255 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10256 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10257 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10258 (aarch64_print_operand, aarch64_print_address_internal)
10259 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10260 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10261 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10262 Handle polynomial GET_MODE_SIZE.
10263 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10264 wider than SImode without modification.
10265 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10266 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10267 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10268 passing and returning SVE modes.
10269 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10270 rather than GEN_INT.
10271 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10272 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10273 (aarch64_allocate_and_probe_stack_space): Likewise.
10274 (aarch64_layout_frame): Cope with polynomial offsets.
10275 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10276 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10277 polynomial offsets.
10278 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10279 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10280 poly_int64 rather than a HOST_WIDE_INT.
10281 (aarch64_get_separate_components, aarch64_process_components)
10282 (aarch64_expand_prologue, aarch64_expand_epilogue)
10283 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10284 (aarch64_anchor_offset): New function, split out from...
10285 (aarch64_legitimize_address): ...here.
10286 (aarch64_builtin_vectorization_cost): Handle polynomial
10287 TYPE_VECTOR_SUBPARTS.
10288 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10290 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10291 number of elements from the PARALLEL rather than the mode.
10292 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10293 rather than GET_MODE_BITSIZE.
10294 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10295 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10296 (aarch64_expand_vec_perm_const_1): Handle polynomial
10297 d->perm.length () and d->perm elements.
10298 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10299 Apply to_constant to d->perm elements.
10300 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10301 polynomial CONST_VECTOR_NUNITS.
10302 (aarch64_move_pointer): Take amount as a poly_int64 rather
10304 (aarch64_progress_pointer): Avoid temporary variable.
10305 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10306 the mode attribute instead of GET_MODE.
10308 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10309 Alan Hayward <alan.hayward@arm.com>
10310 David Sherwood <david.sherwood@arm.com>
10312 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10313 x exists before using it.
10314 (aarch64_add_constant_internal): Rename to...
10315 (aarch64_add_offset_1): ...this. Replace regnum with separate
10316 src and dest rtxes. Handle the case in which they're different,
10317 including when the offset is zero. Replace scratchreg with an rtx.
10318 Use 2 additions if there is no spare register into which we can
10319 move a 16-bit constant.
10320 (aarch64_add_constant): Delete.
10321 (aarch64_add_offset): Replace reg with separate src and dest
10322 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10323 Use aarch64_add_offset_1.
10324 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10325 an rtx rather than an int. Take the delta as a poly_int64
10326 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10327 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10328 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10329 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10330 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10331 and aarch64_add_sp.
10332 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10333 aarch64_add_constant.
10335 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10337 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10338 Use scalar_float_mode.
10340 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10342 * config/aarch64/aarch64-simd.md
10343 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10344 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10345 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10346 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10347 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10348 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10349 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10350 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10351 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10352 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10354 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10357 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10358 targ_options->x_arm_arch_string is non NULL.
10360 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10362 * config/aarch64/aarch64.h
10363 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10365 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10368 * expmed.c (emit_store_flag_force): Swap if const op0
10369 and change VOIDmode to mode of op0.
10371 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10373 PR rtl-optimization/83761
10374 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10375 than bytes to mode_for_size.
10377 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10379 PR middle-end/83189
10380 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10381 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10384 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10386 PR middle-end/83575
10387 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10388 when in layout mode.
10389 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10390 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10393 2018-01-10 Michael Collison <michael.collison@arm.com>
10395 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10396 * config/aarch64/aarch64-option-extension.def: Add
10397 AARCH64_OPT_EXTENSION of 'fp16fml'.
10398 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10399 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10400 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10401 * config/aarch64/constraints.md (Ui7): New constraint.
10402 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10403 (VFMLA_SEL_W): Ditto.
10406 (VFMLA16_LOW): New int iterator.
10407 (VFMLA16_HIGH): Ditto.
10408 (UNSPEC_FMLAL): New unspec.
10409 (UNSPEC_FMLSL): Ditto.
10410 (UNSPEC_FMLAL2): Ditto.
10411 (UNSPEC_FMLSL2): Ditto.
10412 (f16mac): New code attribute.
10413 * config/aarch64/aarch64-simd-builtins.def
10414 (aarch64_fmlal_lowv2sf): Ditto.
10415 (aarch64_fmlsl_lowv2sf): Ditto.
10416 (aarch64_fmlalq_lowv4sf): Ditto.
10417 (aarch64_fmlslq_lowv4sf): Ditto.
10418 (aarch64_fmlal_highv2sf): Ditto.
10419 (aarch64_fmlsl_highv2sf): Ditto.
10420 (aarch64_fmlalq_highv4sf): Ditto.
10421 (aarch64_fmlslq_highv4sf): Ditto.
10422 (aarch64_fmlal_lane_lowv2sf): Ditto.
10423 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10424 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10425 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10426 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10427 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10428 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10429 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10430 (aarch64_fmlal_lane_highv2sf): Ditto.
10431 (aarch64_fmlsl_lane_highv2sf): Ditto.
10432 (aarch64_fmlal_laneq_highv2sf): Ditto.
10433 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10434 (aarch64_fmlalq_lane_highv4sf): Ditto.
10435 (aarch64_fmlsl_lane_highv4sf): Ditto.
10436 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10437 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10438 * config/aarch64/aarch64-simd.md:
10439 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10440 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10441 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10442 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10443 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10444 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10445 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10446 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10447 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10448 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10449 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10450 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10451 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10452 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10453 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10454 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10455 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10456 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10457 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10458 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10459 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10460 (vfmlsl_low_u32): Ditto.
10461 (vfmlalq_low_u32): Ditto.
10462 (vfmlslq_low_u32): Ditto.
10463 (vfmlal_high_u32): Ditto.
10464 (vfmlsl_high_u32): Ditto.
10465 (vfmlalq_high_u32): Ditto.
10466 (vfmlslq_high_u32): Ditto.
10467 (vfmlal_lane_low_u32): Ditto.
10468 (vfmlsl_lane_low_u32): Ditto.
10469 (vfmlal_laneq_low_u32): Ditto.
10470 (vfmlsl_laneq_low_u32): Ditto.
10471 (vfmlalq_lane_low_u32): Ditto.
10472 (vfmlslq_lane_low_u32): Ditto.
10473 (vfmlalq_laneq_low_u32): Ditto.
10474 (vfmlslq_laneq_low_u32): Ditto.
10475 (vfmlal_lane_high_u32): Ditto.
10476 (vfmlsl_lane_high_u32): Ditto.
10477 (vfmlal_laneq_high_u32): Ditto.
10478 (vfmlsl_laneq_high_u32): Ditto.
10479 (vfmlalq_lane_high_u32): Ditto.
10480 (vfmlslq_lane_high_u32): Ditto.
10481 (vfmlalq_laneq_high_u32): Ditto.
10482 (vfmlslq_laneq_high_u32): Ditto.
10483 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10484 (AARCH64_FL_FOR_ARCH8_4): New.
10485 (AARCH64_ISA_F16FML): New ISA flag.
10486 (TARGET_F16FML): New feature flag for fp16fml.
10487 (doc/invoke.texi): Document new fp16fml option.
10489 2018-01-10 Michael Collison <michael.collison@arm.com>
10491 * config/aarch64/aarch64-builtins.c:
10492 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10493 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10494 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10495 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10496 (AARCH64_ISA_SHA3): New ISA flag.
10497 (TARGET_SHA3): New feature flag for sha3.
10498 * config/aarch64/iterators.md (sha512_op): New int attribute.
10499 (CRYPTO_SHA512): New int iterator.
10500 (UNSPEC_SHA512H): New unspec.
10501 (UNSPEC_SHA512H2): Ditto.
10502 (UNSPEC_SHA512SU0): Ditto.
10503 (UNSPEC_SHA512SU1): Ditto.
10504 * config/aarch64/aarch64-simd-builtins.def
10505 (aarch64_crypto_sha512hqv2di): New builtin.
10506 (aarch64_crypto_sha512h2qv2di): Ditto.
10507 (aarch64_crypto_sha512su0qv2di): Ditto.
10508 (aarch64_crypto_sha512su1qv2di): Ditto.
10509 (aarch64_eor3qv8hi): Ditto.
10510 (aarch64_rax1qv2di): Ditto.
10511 (aarch64_xarqv2di): Ditto.
10512 (aarch64_bcaxqv8hi): Ditto.
10513 * config/aarch64/aarch64-simd.md:
10514 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10515 (aarch64_crypto_sha512su0qv2di): Ditto.
10516 (aarch64_crypto_sha512su1qv2di): Ditto.
10517 (aarch64_eor3qv8hi): Ditto.
10518 (aarch64_rax1qv2di): Ditto.
10519 (aarch64_xarqv2di): Ditto.
10520 (aarch64_bcaxqv8hi): Ditto.
10521 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10522 (vsha512h2q_u64): Ditto.
10523 (vsha512su0q_u64): Ditto.
10524 (vsha512su1q_u64): Ditto.
10525 (veor3q_u16): Ditto.
10526 (vrax1q_u64): Ditto.
10527 (vxarq_u64): Ditto.
10528 (vbcaxq_u16): Ditto.
10529 * config/arm/types.md (crypto_sha512): New type attribute.
10530 (crypto_sha3): Ditto.
10531 (doc/invoke.texi): Document new sha3 option.
10533 2018-01-10 Michael Collison <michael.collison@arm.com>
10535 * config/aarch64/aarch64-builtins.c:
10536 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10537 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10538 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10539 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10540 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10541 (AARCH64_ISA_SM4): New ISA flag.
10542 (TARGET_SM4): New feature flag for sm4.
10543 * config/aarch64/aarch64-simd-builtins.def
10544 (aarch64_sm3ss1qv4si): Ditto.
10545 (aarch64_sm3tt1aq4si): Ditto.
10546 (aarch64_sm3tt1bq4si): Ditto.
10547 (aarch64_sm3tt2aq4si): Ditto.
10548 (aarch64_sm3tt2bq4si): Ditto.
10549 (aarch64_sm3partw1qv4si): Ditto.
10550 (aarch64_sm3partw2qv4si): Ditto.
10551 (aarch64_sm4eqv4si): Ditto.
10552 (aarch64_sm4ekeyqv4si): Ditto.
10553 * config/aarch64/aarch64-simd.md:
10554 (aarch64_sm3ss1qv4si): Ditto.
10555 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10556 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10557 (aarch64_sm4eqv4si): Ditto.
10558 (aarch64_sm4ekeyqv4si): Ditto.
10559 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10560 (sm3part_op): Ditto.
10561 (CRYPTO_SM3TT): Ditto.
10562 (CRYPTO_SM3PART): Ditto.
10563 (UNSPEC_SM3SS1): New unspec.
10564 (UNSPEC_SM3TT1A): Ditto.
10565 (UNSPEC_SM3TT1B): Ditto.
10566 (UNSPEC_SM3TT2A): Ditto.
10567 (UNSPEC_SM3TT2B): Ditto.
10568 (UNSPEC_SM3PARTW1): Ditto.
10569 (UNSPEC_SM3PARTW2): Ditto.
10570 (UNSPEC_SM4E): Ditto.
10571 (UNSPEC_SM4EKEY): Ditto.
10572 * config/aarch64/constraints.md (Ui2): New constraint.
10573 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10574 * config/arm/types.md (crypto_sm3): New type attribute.
10575 (crypto_sm4): Ditto.
10576 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10577 (vsm3tt1aq_u32): Ditto.
10578 (vsm3tt1bq_u32): Ditto.
10579 (vsm3tt2aq_u32): Ditto.
10580 (vsm3tt2bq_u32): Ditto.
10581 (vsm3partw1q_u32): Ditto.
10582 (vsm3partw2q_u32): Ditto.
10583 (vsm4eq_u32): Ditto.
10584 (vsm4ekeyq_u32): Ditto.
10585 (doc/invoke.texi): Document new sm4 option.
10587 2018-01-10 Michael Collison <michael.collison@arm.com>
10589 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10590 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10591 (AARCH64_FL_FOR_ARCH8_4): New.
10592 (AARCH64_FL_V8_4): New flag.
10593 (doc/invoke.texi): Document new armv8.4-a option.
10595 2018-01-10 Michael Collison <michael.collison@arm.com>
10597 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10598 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10599 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10600 * config/aarch64/aarch64-option-extension.def: Add
10601 AARCH64_OPT_EXTENSION of 'sha2'.
10602 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10603 (crypto): Disable sha2 and aes if crypto disabled.
10604 (crypto): Enable aes and sha2 if enabled.
10605 (simd): Disable sha2 and aes if simd disabled.
10606 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10608 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10609 (TARGET_SHA2): New feature flag for sha2.
10610 (TARGET_AES): New feature flag for aes.
10611 * config/aarch64/aarch64-simd.md:
10612 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10613 conditional on TARGET_AES.
10614 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10615 (aarch64_crypto_sha1hsi): Make pattern conditional
10617 (aarch64_crypto_sha1hv4si): Ditto.
10618 (aarch64_be_crypto_sha1hv4si): Ditto.
10619 (aarch64_crypto_sha1su1v4si): Ditto.
10620 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10621 (aarch64_crypto_sha1su0v4si): Ditto.
10622 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10623 (aarch64_crypto_sha256su0v4si): Ditto.
10624 (aarch64_crypto_sha256su1v4si): Ditto.
10625 (doc/invoke.texi): Document new aes and sha2 options.
10627 2018-01-10 Martin Sebor <msebor@redhat.com>
10629 PR tree-optimization/83781
10630 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10633 2018-01-11 Martin Sebor <msebor@gmail.com>
10634 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10636 PR tree-optimization/83501
10637 PR tree-optimization/81703
10639 * tree-ssa-strlen.c (get_string_cst): Rename...
10640 (get_string_len): ...to this. Handle global constants.
10641 (handle_char_store): Adjust.
10643 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10644 Jim Wilson <jimw@sifive.com>
10646 * config/riscv/riscv-protos.h (riscv_output_return): New.
10647 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10648 (riscv_attribute_table, riscv_output_return),
10649 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10650 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10651 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10653 (riscv_expand_prologue): Add early return for naked function.
10654 (riscv_expand_epilogue): Likewise.
10655 (riscv_function_ok_for_sibcall): Return false for naked function.
10656 (riscv_set_current_function): New.
10657 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10658 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10659 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10660 * doc/extend.texi (RISC-V Function Attributes): New.
10662 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10664 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10665 check for 128-bit long double before checking TCmode.
10666 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10667 128-bit long doubles before checking TFmode or TCmode.
10668 (FLOAT128_IBM_P): Likewise.
10670 2018-01-10 Martin Sebor <msebor@redhat.com>
10672 PR tree-optimization/83671
10673 * builtins.c (c_strlen): Unconditionally return zero for the empty
10675 Use -Warray-bounds for warnings.
10676 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10677 for non-constant array indices with COMPONENT_REF, arrays of
10678 arrays, and pointers to arrays.
10679 (gimple_fold_builtin_strlen): Determine and set length range for
10680 non-constant character arrays.
10682 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10684 PR middle-end/81897
10685 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10688 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10690 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10692 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10695 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10696 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10697 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10698 indexed_or_indirect_operand predicate.
10699 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10700 (*vsx_le_perm_load_v8hi): Likewise.
10701 (*vsx_le_perm_load_v16qi): Likewise.
10702 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10703 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10704 (*vsx_le_perm_store_v8hi): Likewise.
10705 (*vsx_le_perm_store_v16qi): Likewise.
10706 (eight unnamed splitters): Likewise.
10708 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10710 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10711 * config/rs6000/emmintrin.h: Likewise.
10712 * config/rs6000/mmintrin.h: Likewise.
10713 * config/rs6000/xmmintrin.h: Likewise.
10715 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10718 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10720 * tree.c (tree_nop_conversion): Return true for location wrapper
10722 (maybe_wrap_with_location): New function.
10723 (selftest::check_strip_nops): New function.
10724 (selftest::test_location_wrappers): New function.
10725 (selftest::tree_c_tests): Call it.
10726 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10727 (maybe_wrap_with_location): New decl.
10728 (EXPR_LOCATION_WRAPPER_P): New macro.
10729 (location_wrapper_p): New inline function.
10730 (tree_strip_any_location_wrapper): New inline function.
10732 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10735 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10736 stack_realign_offset for the largest alignment of stack slot
10738 (ix86_find_max_used_stack_alignment): New function.
10739 (ix86_finalize_stack_frame_flags): Use it. Set
10740 max_used_stack_alignment if we don't realign stack.
10741 * config/i386/i386.h (machine_function): Add
10742 max_used_stack_alignment.
10744 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10746 * config/arm/arm.opt (-mbranch-cost): New option.
10747 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10750 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10753 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10754 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10756 2018-01-10 Richard Biener <rguenther@suse.de>
10759 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10760 early out so it also covers the case where we have a non-NULL
10763 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10765 PR tree-optimization/83753
10766 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10767 for non-strided grouped accesses if the number of elements is 1.
10769 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10772 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10773 * i386.h (TARGET_USE_GATHER): Define.
10774 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10776 2018-01-10 Martin Liska <mliska@suse.cz>
10779 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10780 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10782 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10783 CLEANUP_NO_PARTITIONING is not set.
10785 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10787 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10788 for vectors, as a partial revert of r254296.
10789 * rtl.h (const_vec_p): Delete.
10790 (const_vec_duplicate_p): Don't test for vector CONSTs.
10791 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10792 * expmed.c (make_tree): Likewise.
10795 * common.md (E, F): Use CONSTANT_P instead of checking for
10797 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10798 checking for CONST_VECTOR.
10800 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10802 PR middle-end/83575
10803 * predict.c (force_edge_cold): Handle in more sane way edges
10804 with no prediction.
10806 2018-01-09 Carl Love <cel@us.ibm.com>
10808 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10810 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10811 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10812 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10813 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10814 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10815 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10816 * config/rs6000/rs6000-protos.h: Add extern defition for
10817 rs6000_generate_float2_double_code.
10818 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10820 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10821 (float2_v2df): Add define_expand.
10823 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10826 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10827 op_mode in the force_to_mode call.
10829 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10831 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10832 instead of checking each element individually.
10833 (aarch64_evpc_uzp): Likewise.
10834 (aarch64_evpc_zip): Likewise.
10835 (aarch64_evpc_ext): Likewise.
10836 (aarch64_evpc_rev): Likewise.
10837 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10838 instead of checking each element individually. Return true without
10840 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10841 whether all selected elements come from the same input, instead of
10842 checking each element individually. Remove calls to gen_rtx_REG,
10843 start_sequence and end_sequence and instead assert that no rtl is
10846 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10848 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10849 order of HIGH and CONST checks.
10851 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10853 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10854 if the destination isn't an SSA_NAME.
10856 2018-01-09 Richard Biener <rguenther@suse.de>
10858 PR tree-optimization/83668
10859 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10861 (canonicalize_loop_form): ... here, renamed from ...
10862 (canonicalize_loop_closed_ssa_form): ... this and amended to
10863 swap successor edges for loop exit blocks to make us use
10864 the RPO order we need for initial schedule generation.
10866 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10868 PR tree-optimization/64811
10869 * match.pd: When optimizing comparisons with Inf, avoid
10870 introducing or losing exceptions from comparisons with NaN.
10872 2018-01-09 Martin Liska <mliska@suse.cz>
10875 * asan.c (shadow_mem_size): Add gcc_assert.
10877 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10879 Don't save registers in main().
10882 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10883 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10884 * config/avr/avr.c (avr_set_current_function): Don't error if
10885 naked, OS_task or OS_main are specified at the same time.
10886 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10888 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10890 * common/config/avr/avr-common.c (avr_option_optimization_table):
10891 Switch on -mmain-is-OS_task for optimizing compilations.
10893 2018-01-09 Richard Biener <rguenther@suse.de>
10895 PR tree-optimization/83572
10896 * graphite.c: Include cfganal.h.
10897 (graphite_transform_loops): Connect infinite loops to exit
10898 and remove fake edges at the end.
10900 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10902 * ipa-inline.c (edge_badness): Revert accidental checkin.
10904 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10907 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10908 symbols; not inline clones.
10910 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10913 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10914 hard registers. Formatting fixes.
10916 PR preprocessor/83722
10917 * gcc.c (try_generate_repro): Pass
10918 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10919 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10922 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10923 Kito Cheng <kito.cheng@gmail.com>
10925 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10926 (riscv_leaf_function_p): Delete.
10927 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10929 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10931 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10933 (do_ifelse): New function.
10934 (do_isel): New function.
10935 (do_sub3): New function.
10936 (do_add3): New function.
10937 (do_load_mask_compare): New function.
10938 (do_overlap_load_compare): New function.
10939 (expand_compare_loop): New function.
10940 (expand_block_compare): Call expand_compare_loop() when appropriate.
10941 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10942 option description.
10943 (-mblock-compare-inline-loop-limit): New option.
10945 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10948 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10949 Reverse order of second and third operands in first alternative.
10950 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10951 of first and second elements in UNSPEC_VPERMR vector.
10952 (altivec_expand_vec_perm_le): Likewise.
10954 2018-01-08 Jeff Law <law@redhat.com>
10956 PR rtl-optimizatin/81308
10957 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10958 (process_switch): If group_case_labels makes a change, then set
10960 (pass_convert_switch::execute): If a switch is converted, then
10961 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10963 PR rtl-optimization/81308
10964 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10967 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10969 PR target/83663 - Revert r255946
10970 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10971 generation for cases where splatting a value is not useful.
10972 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10973 across a vec_duplicate and a paradoxical subreg forming a vector
10974 mode to a vec_concat.
10976 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10978 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10979 -march=armv8.3-a variants.
10980 * config/arm/t-multilib: Likewise.
10981 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10983 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10985 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10987 (cceq_ior_compare_complement): Give it a name so I can use it, and
10988 change boolean_or_operator predicate to boolean_operator so it can
10989 be used to generate a crand.
10990 (eqne): New code iterator.
10991 (bd/bd_neg): New code_attrs.
10992 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10993 a single define_insn.
10994 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10995 decrement (bdnzt/bdnzf/bdzt/bdzf).
10996 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10997 with the new names of the branch decrement patterns, and added the
10998 names of the branch decrement conditional patterns.
11000 2018-01-08 Richard Biener <rguenther@suse.de>
11002 PR tree-optimization/83563
11003 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11006 2018-01-08 Richard Biener <rguenther@suse.de>
11008 PR middle-end/83713
11009 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11011 2018-01-08 Richard Biener <rguenther@suse.de>
11013 PR tree-optimization/83685
11014 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11015 references to abnormals.
11017 2018-01-08 Richard Biener <rguenther@suse.de>
11020 * dwarf2out.c (output_indirect_strings): Handle empty
11021 skeleton_debug_str_hash.
11022 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11024 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11026 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11027 (emit_store_direct): Likewise.
11028 (arc_trampoline_adjust_address): Likewise.
11029 (arc_asm_trampoline_template): New function.
11030 (arc_initialize_trampoline): Use asm_trampoline_template.
11031 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11032 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11033 * config/arc/arc.md (flush_icache): Delete pattern.
11035 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11037 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11038 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11041 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11044 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11045 by not USED_FOR_TARGET.
11046 (make_pass_resolve_sw_modes): Likewise.
11048 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11050 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11053 2018-01-08 Richard Biener <rguenther@suse.de>
11055 PR middle-end/83580
11056 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11058 2018-01-08 Richard Biener <rguenther@suse.de>
11060 PR middle-end/83517
11061 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11063 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11065 PR middle-end/81897
11066 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11067 basic blocks with a small number of successors.
11068 (convert_control_dep_chain_into_preds): Improve handling of
11070 (dump_predicates): Split apart into...
11071 (dump_pred_chain): ...here...
11072 (dump_pred_info): ...and here.
11073 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11074 (can_chain_union_be_invalidated_p): Improve check for invalidation
11076 (uninit_uses_cannot_happen): Avoid unnecessary if
11077 convert_control_dep_chain_into_preds yielded nothing.
11079 2018-01-06 Martin Sebor <msebor@redhat.com>
11081 PR tree-optimization/83640
11082 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11083 subtracting negative offset from size.
11084 (builtin_access::overlap): Adjust offset bounds of the access to fall
11085 within the size of the object if possible.
11087 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11089 PR rtl-optimization/83699
11090 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11091 extract_bit_field_as_subreg to cases in which the extracted
11092 value is also a vector.
11094 * lra-constraints.c (process_alt_operands): Test for the equivalence
11095 substitutions when detecting a possible reload cycle.
11097 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11100 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11101 by default if flag_selective_schedling{,2}. Formatting fixes.
11103 PR rtl-optimization/83682
11104 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11105 if it has non-VECTOR_MODE element mode.
11106 (vec_duplicate_p): Likewise.
11108 PR middle-end/83694
11109 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11110 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11112 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11115 * config/i386/i386-builtin.def
11116 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11117 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11118 Require also OPTION_MASK_ISA_AVX512F in addition to
11119 OPTION_MASK_ISA_GFNI.
11120 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11121 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11122 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11123 to OPTION_MASK_ISA_GFNI.
11124 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11125 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11126 OPTION_MASK_ISA_AVX512BW.
11127 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11128 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11129 addition to OPTION_MASK_ISA_GFNI.
11130 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11131 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11132 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11133 to OPTION_MASK_ISA_GFNI.
11134 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11135 a requirement for all ISAs rather than any of them with a few
11137 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11139 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11140 bitmasks to be enabled with 3 exceptions, instead of requiring any
11141 enabled ISA with lots of exceptions.
11142 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11143 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11144 Change avx512bw in isa attribute to avx512f.
11145 * config/i386/sgxintrin.h: Add license boilerplate.
11146 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11147 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11148 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11149 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11151 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11152 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11153 temporarily sse2 rather than sse if not enabled already.
11156 * config/i386/sse.md (VI248_VLBW): Rename to ...
11157 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11158 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11159 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11160 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11161 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11162 mode iterator instead of VI248_VLBW.
11164 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11166 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11167 (record_modified): Skip clobbers; add debug output.
11168 (param_change_prob): Use sreal frequencies.
11170 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11172 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11173 punt for user-aligned variables.
11175 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11177 * tree-chrec.c (chrec_contains_symbols): Return true for
11180 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11183 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11184 of (x|y) == x for BICS pattern.
11186 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11188 PR tree-optimization/83605
11189 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11190 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11193 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11195 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11196 * config/epiphany/rtems.h: New file.
11198 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11199 Uros Bizjak <ubizjak@gmail.com>
11202 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11203 QIreg_operand instead of register_operand predicate.
11204 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11205 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11206 comments instead of -fmitigate[-_]rop.
11208 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11211 * cgraphunit.c (symbol_table::compile): Switch to text_section
11212 before calling assembly_start debug hook.
11213 * run-rtl-passes.c (run_rtl_passes): Likewise.
11216 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11218 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11219 range_int_cst_p rather than !symbolic_range_p before calling
11220 extract_range_from_multiplicative_op_1.
11222 2018-01-04 Jeff Law <law@redhat.com>
11224 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11225 redundant test in assertion.
11227 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11229 * doc/rtl.texi: Document machine_mode wrapper classes.
11231 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11233 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11234 using tree_to_uhwi.
11236 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11238 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11239 the VEC_PERM_EXPR fold to fail.
11241 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11244 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11245 to switched_sections.
11247 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11250 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11251 test for d.testing.
11253 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11256 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11257 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11259 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11262 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11263 is BLKmode and bitpos not zero or mode change is needed.
11265 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11268 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11271 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11274 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11275 instead of MULT rtx. Update all corresponding splitters.
11276 (*saddl_se): Ditto.
11277 (*ssub<modesuffix>): Ditto.
11278 (*ssubl_se): Ditto.
11279 (*cmp_sadd_di): Update split patterns.
11280 (*cmp_sadd_si): Ditto.
11281 (*cmp_sadd_sidi): Ditto.
11282 (*cmp_ssub_di): Ditto.
11283 (*cmp_ssub_si): Ditto.
11284 (*cmp_ssub_sidi): Ditto.
11285 * config/alpha/predicates.md (const23_operand): New predicate.
11286 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11287 Look for ASHIFT, not MULT inner operand.
11288 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11290 2018-01-04 Martin Liska <mliska@suse.cz>
11292 PR gcov-profile/83669
11293 * gcov.c (output_intermediate_file): Add version to intermediate
11295 * doc/gcov.texi: Document new field 'version' in intermediate
11296 file format. Fix location of '-k' option of gcov command.
11298 2018-01-04 Martin Liska <mliska@suse.cz>
11301 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11303 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11305 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11307 2018-01-03 Martin Sebor <msebor@redhat.com>
11309 PR tree-optimization/83655
11310 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11311 checking calls with invalid arguments.
11313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11315 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11316 (vectorizable_mask_load_store): Delete.
11317 (vectorizable_call): Return false for masked loads and stores.
11318 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11319 instead of gimple_assign_rhs1.
11320 (vectorizable_load): Handle IFN_MASK_LOAD.
11321 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11325 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11327 (vectorizable_mask_load_store): ...here.
11328 (vectorizable_load): ...and here.
11330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11332 * tree-vect-stmts.c (vect_build_all_ones_mask)
11333 (vect_build_zero_merge_argument): New functions, split out from...
11334 (vectorizable_load): ...here.
11336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11338 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11340 (vectorizable_mask_load_store): ...here.
11341 (vectorizable_store): ...and here.
11343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11345 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11347 (vectorizable_mask_load_store): ...here.
11349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11351 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11352 (vect_model_store_cost): Take a vec_load_store_type instead of a
11354 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11355 (vect_model_store_cost): Take a vec_load_store_type instead of a
11357 (vectorizable_mask_load_store): Update accordingly.
11358 (vectorizable_store): Likewise.
11359 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11363 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11364 IFN_MASK_LOAD calls here rather than...
11365 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11368 Alan Hayward <alan.hayward@arm.com>
11369 David Sherwood <david.sherwood@arm.com>
11371 * expmed.c (extract_bit_field_1): For vector extracts,
11372 fall back to extract_bit_field_as_subreg if vec_extract
11375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11376 Alan Hayward <alan.hayward@arm.com>
11377 David Sherwood <david.sherwood@arm.com>
11379 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11380 they are variable or constant sized.
11381 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11382 slots for constant-sized data.
11384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11385 Alan Hayward <alan.hayward@arm.com>
11386 David Sherwood <david.sherwood@arm.com>
11388 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11389 handling COND_EXPRs with boolean comparisons, try to find a better
11390 basis for the mask type than the boolean itself.
11392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11394 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11395 is calculated and how it can be overridden.
11396 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11397 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11399 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11403 Alan Hayward <alan.hayward@arm.com>
11404 David Sherwood <david.sherwood@arm.com>
11406 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11407 Remove the mode argument.
11408 (aarch64_simd_valid_immediate): Remove the mode and inverse
11410 * config/aarch64/iterators.md (bitsize): New iterator.
11411 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11412 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11413 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11414 aarch64_simd_valid_immediate.
11415 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11416 (aarch64_reg_or_bic_imm): Likewise.
11417 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11418 with an insn_type enum and msl with a modifier_type enum.
11419 Replace element_width with a scalar_mode. Change the shift
11420 to unsigned int. Add constructors for scalar_float_mode and
11421 scalar_int_mode elements.
11422 (aarch64_vect_float_const_representable_p): Delete.
11423 (aarch64_can_const_movi_rtx_p)
11424 (aarch64_simd_scalar_immediate_valid_for_move)
11425 (aarch64_simd_make_constant): Update call to
11426 aarch64_simd_valid_immediate.
11427 (aarch64_advsimd_valid_immediate_hs): New function.
11428 (aarch64_advsimd_valid_immediate): Likewise.
11429 (aarch64_simd_valid_immediate): Remove mode and inverse
11430 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11431 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11432 and aarch64_float_const_representable_p on the result.
11433 (aarch64_output_simd_mov_immediate): Remove mode argument.
11434 Update call to aarch64_simd_valid_immediate and use of
11435 simd_immediate_info.
11436 (aarch64_output_scalar_simd_mov_immediate): Update call
11439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11440 Alan Hayward <alan.hayward@arm.com>
11441 David Sherwood <david.sherwood@arm.com>
11443 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11444 (mode_nunits): Likewise CONST_MODE_NUNITS.
11445 * machmode.def (ADJUST_NUNITS): Document.
11446 * genmodes.c (mode_data::need_nunits_adj): New field.
11447 (blank_mode): Update accordingly.
11448 (adj_nunits): New variable.
11449 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11451 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11452 listed in adj_nunits.
11453 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11454 listed in adj_nunits. Don't emit case statements for such modes.
11455 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11456 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11457 nothing if adj_nunits is nonnull.
11458 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11459 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11460 (emit_mode_fbit): Update use of print_maybe_const_decl.
11461 (emit_move_size): Likewise. Treat the array as non-const
11463 (emit_mode_adjustments): Handle adj_nunits.
11465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11467 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11468 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11469 (VECTOR_MODES): Use it.
11470 (make_vector_modes): Take the prefix as an argument.
11472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11473 Alan Hayward <alan.hayward@arm.com>
11474 David Sherwood <david.sherwood@arm.com>
11476 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11477 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11478 for MODE_VECTOR_BOOL.
11479 * machmode.def (VECTOR_BOOL_MODE): Document.
11480 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11481 (make_vector_bool_mode): New function.
11482 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11484 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11485 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11487 * stor-layout.c (int_mode_for_mode): Likewise.
11488 * tree.c (build_vector_type_for_mode): Likewise.
11489 * varasm.c (output_constant_pool_2): Likewise.
11490 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11491 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11492 for MODE_VECTOR_BOOL.
11493 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11494 of mode class checks.
11495 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11496 instead of a list of mode class checks.
11497 (expand_vector_scalar_condition): Likewise.
11498 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11501 Alan Hayward <alan.hayward@arm.com>
11502 David Sherwood <david.sherwood@arm.com>
11504 * machmode.h (mode_size): Change from unsigned short to
11506 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11507 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11508 or if measurement_type is not polynomial.
11509 (fixed_size_mode::includes_p): Check for constant-sized modes.
11510 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11511 return a poly_uint16 rather than an unsigned short.
11512 (emit_mode_size): Change the type of mode_size from unsigned short
11513 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11514 (emit_mode_adjustments): Cope with polynomial vector sizes.
11515 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11517 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11519 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11520 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11521 * caller-save.c (setup_save_areas): Likewise.
11522 (replace_reg_with_saved_mem): Likewise.
11523 * calls.c (emit_library_call_value_1): Likewise.
11524 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11525 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11526 (gen_lowpart_for_combine): Likewise.
11527 * convert.c (convert_to_integer_1): Likewise.
11528 * cse.c (equiv_constant, cse_insn): Likewise.
11529 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11530 (cselib_subst_to_values): Likewise.
11531 * dce.c (word_dce_process_block): Likewise.
11532 * df-problems.c (df_word_lr_mark_ref): Likewise.
11533 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11534 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11535 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11536 (rtl_for_decl_location): Likewise.
11537 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11538 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11539 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11540 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11541 (expand_expr_real_1): Likewise.
11542 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11543 (pad_below): Likewise.
11544 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11545 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11546 * ira.c (get_subreg_tracking_sizes): Likewise.
11547 * ira-build.c (ira_create_allocno_objects): Likewise.
11548 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11549 (ira_sort_regnos_for_alter_reg): Likewise.
11550 * ira-costs.c (record_operand_costs): Likewise.
11551 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11552 (resolve_simple_move): Likewise.
11553 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11554 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11555 (lra_constraints): Likewise.
11556 (CONST_POOL_OK_P): Reject variable-sized modes.
11557 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11558 (add_pseudo_to_slot, lra_spill): Likewise.
11559 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11560 * optabs-query.c (get_best_extraction_insn): Likewise.
11561 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11562 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11563 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11564 * recog.c (offsettable_address_addr_space_p): Likewise.
11565 * regcprop.c (maybe_mode_change): Likewise.
11566 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11567 * regrename.c (build_def_use): Likewise.
11568 * regstat.c (dump_reg_info): Likewise.
11569 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11570 (find_reloads, find_reloads_subreg_address): Likewise.
11571 * reload1.c (eliminate_regs_1): Likewise.
11572 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11573 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11574 (simplify_binary_operation_1, simplify_subreg): Likewise.
11575 * targhooks.c (default_function_arg_padding): Likewise.
11576 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11577 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11578 (verify_gimple_assign_ternary): Likewise.
11579 * tree-inline.c (estimate_move_cost): Likewise.
11580 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11581 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11582 (get_address_cost_ainc): Likewise.
11583 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11584 (vect_supportable_dr_alignment): Likewise.
11585 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11586 (vectorizable_reduction): Likewise.
11587 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11588 (vectorizable_operation, vectorizable_load): Likewise.
11589 * tree.c (build_same_sized_truth_vector_type): Likewise.
11590 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11591 * var-tracking.c (emit_note_insn_var_location): Likewise.
11592 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11593 (ADDR_VEC_ALIGN): Likewise.
11595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11596 Alan Hayward <alan.hayward@arm.com>
11597 David Sherwood <david.sherwood@arm.com>
11599 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11601 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11602 or if measurement_type is polynomial.
11603 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11604 * combine.c (make_extraction): Likewise.
11605 * dse.c (find_shift_sequence): Likewise.
11606 * dwarf2out.c (mem_loc_descriptor): Likewise.
11607 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11608 (extract_bit_field, extract_low_bits): Likewise.
11609 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11610 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11611 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11612 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11613 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11614 * reload.c (find_reloads): Likewise.
11615 * reload1.c (alter_reg): Likewise.
11616 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11617 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11618 * tree-if-conv.c (predicate_mem_writes): Likewise.
11619 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11620 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11621 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11622 * valtrack.c (dead_debug_insert_temp): Likewise.
11623 * varasm.c (mergeable_constant_section): Likewise.
11624 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11627 Alan Hayward <alan.hayward@arm.com>
11628 David Sherwood <david.sherwood@arm.com>
11630 * expr.c (expand_assignment): Cope with polynomial mode sizes
11631 when assigning to a CONCAT.
11633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11634 Alan Hayward <alan.hayward@arm.com>
11635 David Sherwood <david.sherwood@arm.com>
11637 * machmode.h (mode_precision): Change from unsigned short to
11639 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11641 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11642 or if measurement_type is not polynomial.
11643 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11644 in which the mode is already known to be a scalar_int_mode.
11645 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11646 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11648 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11649 for GET_MODE_PRECISION.
11650 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11651 for GET_MODE_PRECISION.
11652 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11654 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11655 (expand_field_assignment, make_extraction): Likewise.
11656 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11657 (get_last_value): Likewise.
11658 * convert.c (convert_to_integer_1): Likewise.
11659 * cse.c (cse_insn): Likewise.
11660 * expr.c (expand_expr_real_1): Likewise.
11661 * lra-constraints.c (simplify_operand_subreg): Likewise.
11662 * optabs-query.c (can_atomic_load_p): Likewise.
11663 * optabs.c (expand_atomic_load): Likewise.
11664 (expand_atomic_store): Likewise.
11665 * ree.c (combine_reaching_defs): Likewise.
11666 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11667 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11668 * tree.h (type_has_mode_precision_p): Likewise.
11669 * ubsan.c (instrument_si_overflow): Likewise.
11671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11672 Alan Hayward <alan.hayward@arm.com>
11673 David Sherwood <david.sherwood@arm.com>
11675 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11676 polynomial numbers of units.
11677 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11678 (valid_vector_subparts_p): New function.
11679 (build_vector_type): Remove temporary shim and take the number
11680 of units as a poly_uint64 rather than an int.
11681 (build_opaque_vector_type): Take the number of units as a
11682 poly_uint64 rather than an int.
11683 * tree.c (build_vector_from_ctor): Handle polynomial
11684 TYPE_VECTOR_SUBPARTS.
11685 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11686 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11687 (build_vector_from_val): If the number of units is variable,
11688 use build_vec_duplicate_cst for constant operands and
11689 VEC_DUPLICATE_EXPR otherwise.
11690 (make_vector_type): Remove temporary is_constant ().
11691 (build_vector_type, build_opaque_vector_type): Take the number of
11692 units as a poly_uint64 rather than an int.
11693 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11695 * cfgexpand.c (expand_debug_expr): Likewise.
11696 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11697 (store_constructor, expand_expr_real_1): Likewise.
11698 (const_scalar_mask_from_tree): Likewise.
11699 * fold-const-call.c (fold_const_reduction): Likewise.
11700 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11701 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11702 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11703 (fold_relational_const): Likewise.
11704 (native_interpret_vector): Likewise. Change the size from an
11705 int to an unsigned int.
11706 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11707 TYPE_VECTOR_SUBPARTS.
11708 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11709 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11710 duplicating a non-constant operand into a variable-length vector.
11711 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11712 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11713 * ipa-icf.c (sem_variable::equals): Likewise.
11714 * match.pd: Likewise.
11715 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11716 * print-tree.c (print_node): Likewise.
11717 * stor-layout.c (layout_type): Likewise.
11718 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11719 * tree-cfg.c (verify_gimple_comparison): Likewise.
11720 (verify_gimple_assign_binary): Likewise.
11721 (verify_gimple_assign_ternary): Likewise.
11722 (verify_gimple_assign_single): Likewise.
11723 * tree-pretty-print.c (dump_generic_node): Likewise.
11724 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11725 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11726 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11727 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11728 (vect_shift_permute_load_chain): Likewise.
11729 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11730 (expand_vector_condition, optimize_vector_constructor): Likewise.
11731 (lower_vec_perm, get_compute_type): Likewise.
11732 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11733 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11734 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11735 (vect_recog_mask_conversion_pattern): Likewise.
11736 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11737 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11738 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11739 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11740 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11741 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11742 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11743 (supportable_widening_operation): Likewise.
11744 (supportable_narrowing_operation): Likewise.
11745 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11747 * varasm.c (output_constant): Likewise.
11749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11750 Alan Hayward <alan.hayward@arm.com>
11751 David Sherwood <david.sherwood@arm.com>
11753 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11754 so that both the length == 3 and length != 3 cases set up their
11755 own permute vectors. Add comments explaining why we know the
11756 number of elements is constant.
11757 (vect_permute_load_chain): Likewise.
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 * machmode.h (mode_nunits): Change from unsigned char to
11765 (ONLY_FIXED_SIZE_MODES): New macro.
11766 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11767 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11768 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11770 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11771 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11772 or if measurement_type is not polynomial.
11773 * genmodes.c (ZERO_COEFFS): New macro.
11774 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11776 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11777 Use ZERO_COEFFS when emitting initializers.
11778 * data-streamer.h (bp_pack_poly_value): New function.
11779 (bp_unpack_poly_value): Likewise.
11780 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11781 for GET_MODE_NUNITS.
11782 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11783 for GET_MODE_NUNITS.
11784 * tree.c (make_vector_type): Remove temporary shim and make
11785 the real function take the number of units as a poly_uint64
11786 rather than an int.
11787 (build_vector_type_for_mode): Handle polynomial nunits.
11788 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11789 * emit-rtl.c (const_vec_series_p_1): Likewise.
11790 (gen_rtx_CONST_VECTOR): Likewise.
11791 * fold-const.c (test_vec_duplicate_folding): Likewise.
11792 * genrecog.c (validate_pattern): Likewise.
11793 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11794 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11795 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11796 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11797 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11798 * rtlanal.c (subreg_get_info): Likewise.
11799 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11800 (vect_grouped_load_supported): Likewise.
11801 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11802 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11803 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11804 (simplify_const_unary_operation, simplify_binary_operation_1)
11805 (simplify_const_binary_operation, simplify_ternary_operation)
11806 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11807 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11808 instead of CONST_VECTOR_NUNITS.
11809 * varasm.c (output_constant_pool_2): Likewise.
11810 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11811 explicit-encoded elements in the XVEC for variable-length vectors.
11813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11815 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11818 Alan Hayward <alan.hayward@arm.com>
11819 David Sherwood <david.sherwood@arm.com>
11821 * coretypes.h (fixed_size_mode): Declare.
11822 (fixed_size_mode_pod): New typedef.
11823 * builtins.h (target_builtins::x_apply_args_mode)
11824 (target_builtins::x_apply_result_mode): Change type to
11825 fixed_size_mode_pod.
11826 * builtins.c (apply_args_size, apply_result_size, result_vector)
11827 (expand_builtin_apply_args_1, expand_builtin_apply)
11828 (expand_builtin_return): Update accordingly.
11830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11832 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11833 * cselib.c (cselib_hash_rtx): Likewise.
11834 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11835 CONST_VECTOR encoding.
11837 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11838 Jeff Law <law@redhat.com>
11841 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11842 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11843 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11844 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11847 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11848 explicitly probe *sp in a noreturn function if there were any callee
11849 register saves or frame pointer is needed.
11851 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11854 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11855 BLKmode for ternary, binary or unary expressions.
11858 * var-tracking.c (delete_vta_debug_insn): New inline function.
11859 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11860 insns from get_insns () to NULL instead of each bb separately.
11861 Use delete_vta_debug_insn. No longer static.
11862 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11863 delete_vta_debug_insns callers.
11864 * rtl.h (delete_vta_debug_insns): Declare.
11865 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11866 instead of variable_tracking_main.
11868 2018-01-03 Martin Sebor <msebor@redhat.com>
11870 PR tree-optimization/83603
11871 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11872 arguments past the endof the argument list in functions declared
11873 without a prototype.
11874 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11875 Avoid checking when arguments are null.
11877 2018-01-03 Martin Sebor <msebor@redhat.com>
11880 * doc/extend.texi (attribute const): Fix a typo.
11881 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11882 issuing -Wsuggest-attribute for void functions.
11884 2018-01-03 Martin Sebor <msebor@redhat.com>
11886 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11887 offset_int::from instead of wide_int::to_shwi.
11888 (maybe_diag_overlap): Remove assertion.
11889 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11890 * gimple-ssa-sprintf.c (format_directive): Same.
11891 (parse_directive): Same.
11892 (sprintf_dom_walker::compute_format_length): Same.
11893 (try_substitute_return_value): Same.
11895 2018-01-03 Jeff Law <law@redhat.com>
11897 PR middle-end/83654
11898 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11899 non-constant residual for zero at runtime and avoid probing in
11900 that case. Reorganize code for trailing problem to mirror handling
11903 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11905 PR tree-optimization/83501
11906 * tree-ssa-strlen.c (get_string_cst): New.
11907 (handle_char_store): Call get_string_cst.
11909 2018-01-03 Martin Liska <mliska@suse.cz>
11911 PR tree-optimization/83593
11912 * tree-ssa-strlen.c: Include tree-cfg.h.
11913 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11914 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11915 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11917 (strlen_dom_walker::before_dom_children): Call
11918 gimple_purge_dead_eh_edges. Dump tranformation with details
11920 (strlen_dom_walker::before_dom_children): Update call by adding
11921 new argument cleanup_eh.
11922 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11924 2018-01-03 Martin Liska <mliska@suse.cz>
11927 * cif-code.def (VARIADIC_THUNK): New enum value.
11928 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11931 2018-01-03 Jan Beulich <jbeulich@suse.com>
11933 * sse.md (mov<mode>_internal): Tighten condition for when to use
11934 vmovdqu<ssescalarsize> for TI and OI modes.
11936 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11938 Update copyright years.
11940 2018-01-03 Martin Liska <mliska@suse.cz>
11943 * ipa-visibility.c (function_and_variable_visibility): Skip
11944 functions with noipa attribure.
11946 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11948 * gcc.c (process_command): Update copyright notice dates.
11949 * gcov-dump.c (print_version): Ditto.
11950 * gcov.c (print_version): Ditto.
11951 * gcov-tool.c (print_version): Ditto.
11952 * gengtype.c (create_file): Ditto.
11953 * doc/cpp.texi: Bump @copying's copyright year.
11954 * doc/cppinternals.texi: Ditto.
11955 * doc/gcc.texi: Ditto.
11956 * doc/gccint.texi: Ditto.
11957 * doc/gcov.texi: Ditto.
11958 * doc/install.texi: Ditto.
11959 * doc/invoke.texi: Ditto.
11961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11963 * vector-builder.h (vector_builder::m_full_nelts): Change from
11964 unsigned int to poly_uint64.
11965 (vector_builder::full_nelts): Update prototype accordingly.
11966 (vector_builder::new_vector): Likewise.
11967 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11968 (vector_builder::operator ==): Likewise.
11969 (vector_builder::finalize): Likewise.
11970 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11971 Take the number of elements as a poly_uint64 rather than an
11973 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11974 from unsigned int to poly_uint64.
11975 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11976 (vec_perm_indices::new_vector): Likewise.
11977 (vec_perm_indices::length): Likewise.
11978 (vec_perm_indices::nelts_per_input): Likewise.
11979 (vec_perm_indices::input_nelts): Likewise.
11980 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11981 number of elements per input as a poly_uint64 rather than an
11982 unsigned int. Use the original encoding for variable-length
11983 vectors, rather than clamping each individual element.
11984 For the second and subsequent elements in each pattern,
11985 clamp the step and base before clamping their sum.
11986 (vec_perm_indices::series_p): Handle polynomial element counts.
11987 (vec_perm_indices::all_in_range_p): Likewise.
11988 (vec_perm_indices_to_tree): Likewise.
11989 (vec_perm_indices_to_rtx): Likewise.
11990 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11991 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11992 (tree_vector_builder::new_binary_operation): Handle polynomial
11993 element counts. Return false if we need to know the number
11994 of elements at compile time.
11995 * fold-const.c (fold_vec_perm): Punt if the number of elements
11996 isn't known at compile time.
11998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12000 * vec-perm-indices.h (vec_perm_builder): Change element type
12001 from HOST_WIDE_INT to poly_int64.
12002 (vec_perm_indices::element_type): Update accordingly.
12003 (vec_perm_indices::clamp): Handle polynomial element_types.
12004 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12005 (vec_perm_indices::all_in_range_p): Likewise.
12006 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12008 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12009 polynomial vec_perm_indices element types.
12010 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12011 * fold-const.c (fold_vec_perm): Likewise.
12012 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12013 * tree-vect-generic.c (lower_vec_perm): Likewise.
12014 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12015 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12016 element type to HOST_WIDE_INT.
12018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12019 Alan Hayward <alan.hayward@arm.com>
12020 David Sherwood <david.sherwood@arm.com>
12022 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12023 rather than an int. Use plus_constant.
12024 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12025 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12028 Alan Hayward <alan.hayward@arm.com>
12029 David Sherwood <david.sherwood@arm.com>
12031 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12032 a HOST_WIDE_INT to a poly_int64.
12034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12035 Alan Hayward <alan.hayward@arm.com>
12036 David Sherwood <david.sherwood@arm.com>
12038 * calls.c (load_register_parameters): Cope with polynomial
12039 mode sizes. Require a constant size for BLKmode parameters
12040 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12041 forces a parameter to be padded at the lsb end in order to
12042 fill a complete number of words, require the parameter size
12043 to be ordered wrt UNITS_PER_WORD.
12045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12046 Alan Hayward <alan.hayward@arm.com>
12047 David Sherwood <david.sherwood@arm.com>
12049 * reload1.c (spill_stack_slot_width): Change element type
12050 from unsigned int to poly_uint64_pod.
12051 (alter_reg): Treat mode sizes as polynomial.
12053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12054 Alan Hayward <alan.hayward@arm.com>
12055 David Sherwood <david.sherwood@arm.com>
12057 * reload.c (complex_word_subreg_p): New function.
12058 (reload_inner_reg_of_subreg, push_reload): Use it.
12060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12061 Alan Hayward <alan.hayward@arm.com>
12062 David Sherwood <david.sherwood@arm.com>
12064 * lra-constraints.c (process_alt_operands): Reject matched
12065 operands whose sizes aren't ordered.
12066 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12073 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12080 Use plus_constant instead of gen_rtx_PLUS.
12082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12083 Alan Hayward <alan.hayward@arm.com>
12084 David Sherwood <david.sherwood@arm.com>
12086 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12087 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12088 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12089 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12090 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12091 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12092 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12093 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12094 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12095 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12097 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12098 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12099 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12100 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12101 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12102 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12103 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12104 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12105 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12106 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12108 * expr.c (emit_move_resolve_push): Treat the input and result
12109 of PUSH_ROUNDING as a poly_int64.
12110 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12111 (emit_push_insn): Likewise.
12112 * lra-eliminations.c (mark_not_eliminable): Likewise.
12113 * recog.c (push_operand): Likewise.
12114 * reload1.c (elimination_effects): Likewise.
12115 * rtlanal.c (nonzero_bits1): Likewise.
12116 * calls.c (store_one_arg): Likewise. Require the padding to be
12117 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12124 Use plus_constant instead of gen_rtx_PLUS.
12126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12127 Alan Hayward <alan.hayward@arm.com>
12128 David Sherwood <david.sherwood@arm.com>
12130 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12131 rather than an int.
12133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12134 Alan Hayward <alan.hayward@arm.com>
12135 David Sherwood <david.sherwood@arm.com>
12137 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12138 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12139 via stack temporaries. Treat the mode size as polynomial too.
12141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12142 Alan Hayward <alan.hayward@arm.com>
12143 David Sherwood <david.sherwood@arm.com>
12145 * expr.c (expand_expr_real_2): When handling conversions involving
12146 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12147 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12148 as a poly_uint64 too.
12150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12151 Alan Hayward <alan.hayward@arm.com>
12152 David Sherwood <david.sherwood@arm.com>
12154 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12157 Alan Hayward <alan.hayward@arm.com>
12158 David Sherwood <david.sherwood@arm.com>
12160 * combine.c (can_change_dest_mode): Handle polynomial
12161 REGMODE_NATURAL_SIZE.
12162 * expmed.c (store_bit_field_1): Likewise.
12163 * expr.c (store_constructor): Likewise.
12164 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12165 and polynomial REGMODE_NATURAL_SIZE.
12166 (gen_lowpart_common): Likewise.
12167 * reginfo.c (record_subregs_of_mode): Likewise.
12168 * rtlanal.c (read_modify_subreg_p): Likewise.
12170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12171 Alan Hayward <alan.hayward@arm.com>
12172 David Sherwood <david.sherwood@arm.com>
12174 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12175 numbers of elements.
12177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12178 Alan Hayward <alan.hayward@arm.com>
12179 David Sherwood <david.sherwood@arm.com>
12181 * match.pd: Cope with polynomial numbers of vector elements.
12183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12184 Alan Hayward <alan.hayward@arm.com>
12185 David Sherwood <david.sherwood@arm.com>
12187 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12188 in a POINTER_PLUS_EXPR.
12190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12191 Alan Hayward <alan.hayward@arm.com>
12192 David Sherwood <david.sherwood@arm.com>
12194 * omp-simd-clone.c (simd_clone_subparts): New function.
12195 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12196 (ipa_simd_modify_function_body): Likewise.
12198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12199 Alan Hayward <alan.hayward@arm.com>
12200 David Sherwood <david.sherwood@arm.com>
12202 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12203 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12204 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12205 (expand_vector_condition, vector_element): Likewise.
12206 (subparts_gt): New function.
12207 (get_compute_type): Use subparts_gt.
12208 (count_type_subparts): Delete.
12209 (expand_vector_operations_1): Use subparts_gt instead of
12210 count_type_subparts.
12212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12213 Alan Hayward <alan.hayward@arm.com>
12214 David Sherwood <david.sherwood@arm.com>
12216 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12217 (vect_compile_time_alias): ...this new function. Do the calculation
12218 on poly_ints rather than trees.
12219 (vect_prune_runtime_alias_test_list): Update call accordingly.
12221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12222 Alan Hayward <alan.hayward@arm.com>
12223 David Sherwood <david.sherwood@arm.com>
12225 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12227 (vect_schedule_slp_instance): Likewise.
12229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12230 Alan Hayward <alan.hayward@arm.com>
12231 David Sherwood <david.sherwood@arm.com>
12233 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12234 constant and extern definitions for variable-length vectors.
12235 (vect_get_constant_vectors): Note that the number of units
12236 is known to be constant.
12238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12239 Alan Hayward <alan.hayward@arm.com>
12240 David Sherwood <david.sherwood@arm.com>
12242 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12243 of units as polynomial. Choose between WIDE and NARROW based
12246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12247 Alan Hayward <alan.hayward@arm.com>
12248 David Sherwood <david.sherwood@arm.com>
12250 * tree-vect-stmts.c (simd_clone_subparts): New function.
12251 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12254 Alan Hayward <alan.hayward@arm.com>
12255 David Sherwood <david.sherwood@arm.com>
12257 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12258 vectors as polynomial. Use build_index_vector for
12259 IFN_GOMP_SIMD_LANE.
12261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12262 Alan Hayward <alan.hayward@arm.com>
12263 David Sherwood <david.sherwood@arm.com>
12265 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12266 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12267 for variable-length vectors.
12268 (vectorizable_mask_load_store): Treat the number of units as
12269 polynomial, asserting that it is constant if the condition has
12270 already been enforced.
12271 (vectorizable_store, vectorizable_load): Likewise.
12273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12274 Alan Hayward <alan.hayward@arm.com>
12275 David Sherwood <david.sherwood@arm.com>
12277 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12278 of units as polynomial. Punt if we can't tell at compile time
12279 which vector contains the final result.
12281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12282 Alan Hayward <alan.hayward@arm.com>
12283 David Sherwood <david.sherwood@arm.com>
12285 * tree-vect-loop.c (vectorizable_induction): Treat the number
12286 of units as polynomial. Punt on SLP inductions. Use an integer
12287 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12288 cast of such a series for variable-length floating-point
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 * tree.h (build_index_vector): Declare.
12296 * tree.c (build_index_vector): New function.
12297 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12298 of units as polynomial, forcibly converting it to a constant if
12299 vectorizable_reduction has already enforced the condition.
12300 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12301 to create a {1,2,3,...} vector.
12302 (vectorizable_reduction): Treat the number of units as polynomial.
12303 Choose vectype_in based on the largest scalar element size rather
12304 than the smallest number of units. Enforce the restrictions
12307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12308 Alan Hayward <alan.hayward@arm.com>
12309 David Sherwood <david.sherwood@arm.com>
12311 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12312 number of units as polynomial.
12314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12315 Alan Hayward <alan.hayward@arm.com>
12316 David Sherwood <david.sherwood@arm.com>
12318 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12319 * target.def (autovectorize_vector_sizes): Return the vector sizes
12320 by pointer, using vector_sizes rather than a bitmask.
12321 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12322 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12323 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12325 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12326 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12327 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12328 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12329 * omp-general.c (omp_max_vf): Likewise.
12330 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12331 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12332 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12333 * tree-vect-slp.c (vect_slp_bb): Likewise.
12334 * doc/tm.texi: Regenerate.
12335 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12337 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12338 the vector size as a poly_uint64 rather than an unsigned int.
12339 (current_vector_size): Change from an unsigned int to a poly_uint64.
12340 (get_vectype_for_scalar_type): Update accordingly.
12341 * tree.h (build_truth_vector_type): Take the size and number of
12342 units as a poly_uint64 rather than an unsigned int.
12343 (build_vector_type): Add a temporary overload that takes
12344 the number of units as a poly_uint64 rather than an unsigned int.
12345 * tree.c (make_vector_type): Likewise.
12346 (build_truth_vector_type): Take the number of units as a poly_uint64
12347 rather than an unsigned int.
12349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12350 Alan Hayward <alan.hayward@arm.com>
12351 David Sherwood <david.sherwood@arm.com>
12353 * target.def (get_mask_mode): Take the number of units and length
12354 as poly_uint64s rather than unsigned ints.
12355 * targhooks.h (default_get_mask_mode): Update accordingly.
12356 * targhooks.c (default_get_mask_mode): Likewise.
12357 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12358 * doc/tm.texi: Regenerate.
12360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12361 Alan Hayward <alan.hayward@arm.com>
12362 David Sherwood <david.sherwood@arm.com>
12364 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12365 * omp-general.c (omp_max_vf): Likewise.
12366 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12367 (expand_omp_simd): Handle polynomial safelen.
12368 * omp-low.c (omplow_simd_context): Add a default constructor.
12369 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12370 (lower_rec_simd_input_clauses): Update accordingly.
12371 (lower_rec_input_clauses): Likewise.
12373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12374 Alan Hayward <alan.hayward@arm.com>
12375 David Sherwood <david.sherwood@arm.com>
12377 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12378 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12379 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12380 (vect_analyze_slp_cost): Likewise.
12381 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12382 (vect_model_load_cost): Likewise.
12384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12385 Alan Hayward <alan.hayward@arm.com>
12386 David Sherwood <david.sherwood@arm.com>
12388 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12389 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12390 from an unsigned int * to a poly_uint64_pod *.
12391 (calculate_unrolling_factor): New function.
12392 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12395 Alan Hayward <alan.hayward@arm.com>
12396 David Sherwood <david.sherwood@arm.com>
12398 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12399 from an unsigned int to a poly_uint64.
12400 (_loop_vec_info::slp_unrolling_factor): Likewise.
12401 (_loop_vec_info::vectorization_factor): Change from an int
12403 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12404 (vect_get_num_vectors): New function.
12405 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12406 (vect_get_num_copies): Use vect_get_num_vectors.
12407 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12408 to an unsigned int *.
12409 (vect_analyze_data_refs): Change min_vf from an int * to a
12411 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12412 than an unsigned HOST_WIDE_INT.
12413 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12414 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12415 to an unsigned int *.
12416 (vect_analyze_data_ref_dependences): Likewise.
12417 (vect_compute_data_ref_alignment): Handle polynomial vf.
12418 (vect_enhance_data_refs_alignment): Likewise.
12419 (vect_prune_runtime_alias_test_list): Likewise.
12420 (vect_shift_permute_load_chain): Likewise.
12421 (vect_supportable_dr_alignment): Likewise.
12422 (dependence_distance_ge_vf): Take the vectorization factor as a
12423 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12424 (vect_analyze_data_refs): Change min_vf from an int * to a
12426 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12427 vfm1 as a poly_uint64 rather than an int. Make the same change
12428 for the returned bound_scalar.
12429 (vect_gen_vector_loop_niters): Handle polynomial vf.
12430 (vect_do_peeling): Likewise. Update call to
12431 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12432 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12434 * tree-vect-loop.c (vect_determine_vectorization_factor)
12435 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12436 (vect_get_known_peeling_cost): Likewise.
12437 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12438 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12439 (vect_transform_loop): Likewise. Use the lowest possible VF when
12440 updating the upper bounds of the loop.
12441 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12442 rather than an int.
12443 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12444 polynomial unroll factors.
12445 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12446 (vect_make_slp_decision): Likewise.
12447 (vect_supported_load_permutation_p): Likewise, and polynomial
12449 (vect_analyze_slp_cost): Handle polynomial vf.
12450 (vect_slp_analyze_node_operations): Likewise.
12451 (vect_slp_analyze_bb_1): Likewise.
12452 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12453 than an unsigned HOST_WIDE_INT.
12454 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12455 (vectorizable_load): Handle polynomial vf.
12456 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12458 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12461 Alan Hayward <alan.hayward@arm.com>
12462 David Sherwood <david.sherwood@arm.com>
12464 * match.pd: Handle bit operations involving three constants
12465 and try to fold one pair.
12467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12469 * tree-vect-loop-manip.c: Include gimple-fold.h.
12470 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12471 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12472 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12473 Add a path that uses a step of VF instead of 1, but disable it
12475 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12476 and niters_no_overflow parameters. Update calls to
12477 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12478 Create a new SSA name if the latter choses to use a ste other
12479 than zero, and return it via niters_vector_mult_vf_var.
12480 * tree-vect-loop.c (vect_transform_loop): Update calls to
12481 vect_do_peeling, vect_gen_vector_loop_niters and
12482 slpeel_make_loop_iterate_ntimes.
12483 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12484 (vect_gen_vector_loop_niters): Update declarations after above changes.
12486 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12488 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12489 128-bit round to integer instructions.
12490 (ceil<mode>2): Likewise.
12491 (btrunc<mode>2): Likewise.
12492 (round<mode>2): Likewise.
12494 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12496 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12497 unaligned VSX load/store on P8/P9.
12498 (expand_block_clear): Allow the use of unaligned VSX
12499 load/store on P8/P9.
12501 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12503 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12505 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12506 swap associated with both a load and a store.
12508 2018-01-02 Andrew Waterman <andrew@sifive.com>
12510 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12511 * config/riscv/riscv.md (clear_cache): Use it.
12513 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12515 * web.c: Remove out-of-date comment.
12517 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12519 * expr.c (fixup_args_size_notes): Check that any existing
12520 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12521 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12522 (emit_single_push_insn): ...here.
12524 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12526 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12527 (const_vector_encoded_nelts): New function.
12528 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12529 (const_vector_int_elt, const_vector_elt): Declare.
12530 * emit-rtl.c (const_vector_int_elt_1): New function.
12531 (const_vector_elt): Likewise.
12532 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12533 of CONST_VECTOR_ELT.
12535 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12537 * expr.c: Include rtx-vector-builder.h.
12538 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12539 directly on the tree encoding.
12540 (const_vector_from_tree): Likewise.
12541 * optabs.c: Include rtx-vector-builder.h.
12542 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12543 sequence of "u" values.
12544 * vec-perm-indices.c: Include rtx-vector-builder.h.
12545 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12546 directly on the vec_perm_indices encoding.
12548 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12550 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12551 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12552 * rtx-vector-builder.h: New file.
12553 * rtx-vector-builder.c: Likewise.
12554 * rtl.h (rtx_def::u2): Add a const_vector field.
12555 (CONST_VECTOR_NPATTERNS): New macro.
12556 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12557 (CONST_VECTOR_DUPLICATE_P): Likewise.
12558 (CONST_VECTOR_STEPPED_P): Likewise.
12559 (CONST_VECTOR_ENCODED_ELT): Likewise.
12560 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12561 (unwrap_const_vec_duplicate): Likewise.
12562 (const_vec_series_p): Check for a non-duplicated vector encoding.
12563 Say that the function only returns true for integer vectors.
12564 * emit-rtl.c: Include rtx-vector-builder.h.
12565 (gen_const_vec_duplicate_1): Delete.
12566 (gen_const_vector): Call gen_const_vec_duplicate instead of
12567 gen_const_vec_duplicate_1.
12568 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12569 (gen_const_vec_duplicate): Use rtx_vector_builder.
12570 (gen_const_vec_series): Likewise.
12571 (gen_rtx_CONST_VECTOR): Likewise.
12572 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12573 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12574 Build a new vector rather than modifying a CONST_VECTOR in-place.
12575 (handle_special_swappables): Update call accordingly.
12576 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12577 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12578 Build a new vector rather than modifying a CONST_VECTOR in-place.
12579 (handle_special_swappables): Update call accordingly.
12581 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12583 * simplify-rtx.c (simplify_const_binary_operation): Use
12584 CONST_VECTOR_ELT instead of XVECEXP.
12586 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12588 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12589 the selector elements to be different from the data elements
12590 if the selector is a VECTOR_CST.
12591 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12592 ssizetype for the selector.
12594 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12596 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12597 before testing each element individually.
12598 * tree-vect-generic.c (lower_vec_perm): Likewise.
12600 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12602 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12603 * selftest-run-tests.c (selftest::run_tests): Call it.
12604 * vector-builder.h (vector_builder::operator ==): New function.
12605 (vector_builder::operator !=): Likewise.
12606 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12607 (vec_perm_indices::all_from_input_p): New function.
12608 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12609 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12610 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12611 instead of reading the VECTOR_CST directly. Detect whether both
12612 vector inputs are the same before constructing the vec_perm_indices,
12613 and update the number of inputs argument accordingly. Use the
12614 utility functions added above. Only construct sel2 if we need to.
12616 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12618 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12619 the broadcast of the low byte.
12620 (expand_mult_highpart): Use an explicit encoding for the permutes.
12621 * optabs-query.c (can_mult_highpart_p): Likewise.
12622 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12623 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12624 (vectorizable_bswap): Likewise.
12625 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12626 explicit encoding for the power-of-2 permutes.
12627 (vect_permute_store_chain): Likewise.
12628 (vect_grouped_load_supported): Likewise.
12629 (vect_permute_load_chain): Likewise.
12631 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12633 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12634 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12635 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12636 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12637 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12638 (vect_gen_perm_mask_any): Likewise.
12640 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12642 * int-vector-builder.h: New file.
12643 * vec-perm-indices.h: Include int-vector-builder.h.
12644 (vec_perm_indices): Redefine as an int_vector_builder.
12645 (auto_vec_perm_indices): Delete.
12646 (vec_perm_builder): Redefine as a stand-alone class.
12647 (vec_perm_indices::vec_perm_indices): New function.
12648 (vec_perm_indices::clamp): Likewise.
12649 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12650 (vec_perm_indices::new_vector): New function.
12651 (vec_perm_indices::new_expanded_vector): Update for new
12652 vec_perm_indices class.
12653 (vec_perm_indices::rotate_inputs): New function.
12654 (vec_perm_indices::all_in_range_p): Operate directly on the
12655 encoded form, without computing elided elements.
12656 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12657 encoding. Update for new vec_perm_indices class.
12658 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12659 the given vec_perm_builder.
12660 (expand_vec_perm_var): Update vec_perm_builder constructor.
12661 (expand_mult_highpart): Use vec_perm_builder instead of
12662 auto_vec_perm_indices.
12663 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12664 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12665 or double series encoding as appropriate.
12666 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12667 vec_perm_indices instead of auto_vec_perm_indices.
12668 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12669 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12670 (vect_permute_store_chain): Likewise.
12671 (vect_grouped_load_supported): Likewise.
12672 (vect_permute_load_chain): Likewise.
12673 (vect_shift_permute_load_chain): Likewise.
12674 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12675 (vect_transform_slp_perm_load): Likewise.
12676 (vect_schedule_slp_instance): Likewise.
12677 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12678 (vectorizable_mask_load_store): Likewise.
12679 (vectorizable_bswap): Likewise.
12680 (vectorizable_store): Likewise.
12681 (vectorizable_load): Likewise.
12682 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12683 vec_perm_indices instead of auto_vec_perm_indices. Use
12684 tree_to_vec_perm_builder to read the vector from a tree.
12685 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12686 vec_perm_builder instead of a vec_perm_indices.
12687 (have_whole_vector_shift): Use vec_perm_builder and
12688 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12689 truncation to calc_vec_perm_mask_for_shift.
12690 (vect_create_epilog_for_reduction): Likewise.
12691 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12692 from auto_vec_perm_indices to vec_perm_indices.
12693 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12694 instead of changing individual elements.
12695 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12696 the vector in d.perm.
12697 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12698 from auto_vec_perm_indices to vec_perm_indices.
12699 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12700 instead of changing individual elements.
12701 (arm_vectorize_vec_perm_const): Use new_vector to install
12702 the vector in d.perm.
12703 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12704 Update vec_perm_builder constructor.
12705 (rs6000_expand_interleave): Likewise.
12706 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12707 (rs6000_expand_interleave): Likewise.
12709 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12711 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12712 to qimode could truncate the indices.
12713 * optabs.c (expand_vec_perm_var): Likewise.
12715 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12717 * Makefile.in (OBJS): Add vec-perm-indices.o.
12718 * vec-perm-indices.h: New file.
12719 * vec-perm-indices.c: Likewise.
12720 * target.h (vec_perm_indices): Replace with a forward class
12722 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12723 * optabs.h: Include vec-perm-indices.h.
12724 (expand_vec_perm): Delete.
12725 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12726 (expand_vec_perm_const): Declare.
12727 * target.def (vec_perm_const_ok): Replace with...
12728 (vec_perm_const): ...this new hook.
12729 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12730 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12731 * doc/tm.texi: Regenerate.
12732 * optabs.def (vec_perm_const): Delete.
12733 * doc/md.texi (vec_perm_const): Likewise.
12734 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12735 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12736 expand_vec_perm for constant permutation vectors. Assert that
12737 the mode of variable permutation vectors is the integer equivalent
12738 of the mode that is being permuted.
12739 * optabs-query.h (selector_fits_mode_p): Declare.
12740 * optabs-query.c: Include vec-perm-indices.h.
12741 (selector_fits_mode_p): New function.
12742 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12743 is defined, instead of checking whether the vec_perm_const_optab
12744 exists. Use targetm.vectorize.vec_perm_const instead of
12745 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12746 fit in the vector mode before using a variable permute.
12747 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12748 vec_perm_indices instead of an rtx.
12749 (expand_vec_perm): Replace with...
12750 (expand_vec_perm_const): ...this new function. Take the selector
12751 as a vec_perm_indices rather than an rtx. Also take the mode of
12752 the selector. Update call to shift_amt_for_vec_perm_mask.
12753 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12754 Use vec_perm_indices::new_expanded_vector to expand the original
12755 selector into bytes. Check whether the indices fit in the vector
12756 mode before using a variable permute.
12757 (expand_vec_perm_var): Make global.
12758 (expand_mult_highpart): Use expand_vec_perm_const.
12759 * fold-const.c: Includes vec-perm-indices.h.
12760 * tree-ssa-forwprop.c: Likewise.
12761 * tree-vect-data-refs.c: Likewise.
12762 * tree-vect-generic.c: Likewise.
12763 * tree-vect-loop.c: Likewise.
12764 * tree-vect-slp.c: Likewise.
12765 * tree-vect-stmts.c: Likewise.
12766 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12768 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12769 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12770 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12771 (aarch64_vectorize_vec_perm_const): ...this new function.
12772 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12773 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12774 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12775 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12776 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12777 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12778 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12780 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12781 check for NEON modes.
12782 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12783 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12784 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12785 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12787 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12788 the old VEC_PERM_CONST conditions.
12789 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12790 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12791 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12792 (ia64_vectorize_vec_perm_const_ok): Merge into...
12793 (ia64_vectorize_vec_perm_const): ...this new function.
12794 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12795 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12796 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12797 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12798 * config/mips/mips.c (mips_expand_vec_perm_const)
12799 (mips_vectorize_vec_perm_const_ok): Merge into...
12800 (mips_vectorize_vec_perm_const): ...this new function.
12801 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12802 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12803 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12804 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12805 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12806 (rs6000_expand_vec_perm_const): Delete.
12807 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12809 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12810 (altivec_expand_vec_perm_const_le): Take each operand individually.
12811 Operate on constant selectors rather than rtxes.
12812 (altivec_expand_vec_perm_const): Likewise. Update call to
12813 altivec_expand_vec_perm_const_le.
12814 (rs6000_expand_vec_perm_const): Delete.
12815 (rs6000_vectorize_vec_perm_const_ok): Delete.
12816 (rs6000_vectorize_vec_perm_const): New function.
12817 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12818 an element count and rtx array.
12819 (rs6000_expand_extract_even): Update call accordingly.
12820 (rs6000_expand_interleave): Likewise.
12821 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12822 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12823 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12824 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12825 (rs6000_expand_vec_perm_const): Delete.
12826 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12827 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12828 (altivec_expand_vec_perm_const_le): Take each operand individually.
12829 Operate on constant selectors rather than rtxes.
12830 (altivec_expand_vec_perm_const): Likewise. Update call to
12831 altivec_expand_vec_perm_const_le.
12832 (rs6000_expand_vec_perm_const): Delete.
12833 (rs6000_vectorize_vec_perm_const_ok): Delete.
12834 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12835 reference to the SPE evmerge intructions.
12836 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12837 an element count and rtx array.
12838 (rs6000_expand_extract_even): Update call accordingly.
12839 (rs6000_expand_interleave): Likewise.
12840 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12841 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12843 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12845 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12847 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12848 vector mode and that that mode matches the mode of the data
12850 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12851 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12852 directly using expand_vec_perm_1 when forcing selectors into
12854 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12856 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12858 * optabs-query.h (can_vec_perm_p): Delete.
12859 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12860 * optabs-query.c (can_vec_perm_p): Split into...
12861 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12862 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12863 particular selector is valid.
12864 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12865 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12866 (vect_grouped_load_supported): Likewise.
12867 (vect_shift_permute_load_chain): Likewise.
12868 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12869 (vect_transform_slp_perm_load): Likewise.
12870 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12871 (vectorizable_bswap): Likewise.
12872 (vect_gen_perm_mask_checked): Likewise.
12873 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12874 implementations of variable permutation vectors into account
12875 when deciding which selector to use.
12876 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12877 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12878 with a false third argument.
12879 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12880 to test whether the constant selector is valid and can_vec_perm_var_p
12881 to test whether a variable selector is valid.
12883 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12885 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12886 * optabs-query.c (can_vec_perm_p): Likewise.
12887 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12888 instead of vec_perm_indices.
12889 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12890 (vect_gen_perm_mask_checked): Likewise,
12891 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12892 (vect_gen_perm_mask_checked): Likewise,
12894 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12896 * optabs-query.h (qimode_for_vec_perm): Declare.
12897 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12898 (qimode_for_vec_perm): ...this new function.
12899 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12901 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12903 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12904 does not have a conditional at the top.
12906 2018-01-02 Richard Biener <rguenther@suse.de>
12908 * ipa-inline.c (big_speedup_p): Fix expression.
12910 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12913 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12916 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12920 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12921 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12922 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12923 cond_taken_branch_cost 3->4.
12925 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12927 PR tree-optimization/83581
12928 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12929 TODO_cleanup_cfg if any changes have been made.
12931 PR middle-end/83608
12932 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12933 convert_modes if target mode has the right side, but different mode
12936 PR middle-end/83609
12937 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12938 last argument when extracting from CONCAT. If either from_real or
12939 from_imag is NULL, use expansion through memory. If result is not
12940 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12941 the parts directly to inner mode, if even that fails, use expansion
12944 PR middle-end/83623
12945 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12946 check for bswap in mode rather than HImode and use that in expand_unop
12949 Copyright (C) 2018 Free Software Foundation, Inc.
12951 Copying and distribution of this file, with or without modification,
12952 are permitted in any medium without royalty provided the copyright
12953 notice and this notice are preserved.