1 2018-05-17 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/85757
4 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5 remove defs that only feed that PHI from further processing.
7 2018-05-16 Jim Wilson <jimw@sifive.com>
9 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
11 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
13 2018-05-16 Mark Wielaard <mark@klomp.org>
15 * dwarf2out.c (count_index_strings): New function.
16 (output_indirect_strings): Call count_index_strings and generate
17 header for dwarf_version >= 5.
19 2018-05-16 Mark Wielaard <mark@klomp.org>
21 * dwarf2out.c (dwarf_FORM): New function.
22 (set_indirect_string): Use dwarf_FORM.
23 (reset_indirect_string): Likewise.
24 (size_of_die): Likewise.
25 (value_format): Likewise.
26 (output_die): Likewise.
27 (add_skeleton_AT_string): Likewise.
28 (output_macinfo_op): Likewise.
29 (index_string): Likewise.
30 (output_index_string_offset): Likewise.
31 (output_index_string): Likewise.
32 (count_index_strings): Likewise.
34 2018-05-16 Carl Love <cel@us.ibm.com>
36 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
37 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
39 2018-05-16 Martin Jambor <mjambor@suse.cz>
41 * ipa-prop.c (ipa_free_all_edge_args): Remove.
42 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
44 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
46 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
47 (fnma<mode>4): Likewise.
48 (fms<mode>4): Likewise.
49 (fnms<mode>4): Likewise.
50 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
51 (aarch64_fnma<mode>4): Likewise.
52 (aarch64_fms<mode>4): Likewise.
53 (aarch64_fnms<mode>4): Likewise.
54 (aarch64_fnmadd<mode>4): Likewise.
56 2018-05-16 Jason Merrill <jason@redhat.com>
58 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
60 2018-05-16 Richard Biener <rguenther@suse.de>
62 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
63 (dump_stmt_cost): Declare.
64 (add_stmt_cost): Dump cost we add.
65 (add_stmt_costs): New function.
66 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
68 (vect_analyze_stmt): Adjust prototype.
69 (vectorizable_condition): Likewise.
70 (vectorizable_live_operation): Likewise.
71 (vectorizable_reduction): Likewise.
72 (vectorizable_induction): Likewise.
73 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
74 cost vector to pass to vectorizable_ and record afterwards.
75 (vect_model_reduction_cost): Take cost vector argument and adjust.
76 (vect_model_induction_cost): Likewise.
77 (vectorizable_reduction): Likewise.
78 (vectorizable_induction): Likewise.
79 (vectorizable_live_operation): Likewise.
80 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
81 SLP_TREE_NUMBER_OF_VEC_STMTS.
82 (vect_analyze_slp_cost_1): Remove.
83 (vect_analyze_slp_cost): Likewise.
84 (vect_slp_analyze_node_operations): Take visited args and
85 a target cost vector. Avoid processing already visited stmt sets.
86 (vect_slp_analyze_operations): Use a local cost vector to gather
87 costs and register those of non-discarded instances.
88 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
89 (vect_schedule_slp_instance): Remove copying of
90 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
92 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
93 adding cost. Record cost entry location.
94 (vect_prologue_cost_for_slp_op): Function to compute cost of
95 a constant or invariant generated for SLP vect in the prologue,
96 split out from vect_analyze_slp_cost_1.
97 (vect_model_simple_cost): Make static. Adjust for SLP costing.
98 (vect_model_promotion_demotion_cost): Likewise.
99 (vect_model_store_cost): Likewise, make static.
100 (vect_model_load_cost): Likewise.
101 (vectorizable_bswap): Add cost vector arg and adjust.
102 (vectorizable_call): Likewise.
103 (vectorizable_simd_clone_call): Likewise.
104 (vectorizable_conversion): Likewise.
105 (vectorizable_assignment): Likewise.
106 (vectorizable_shift): Likewise.
107 (vectorizable_operation): Likewise.
108 (vectorizable_store): Likewise.
109 (vectorizable_load): Likewise.
110 (vectorizable_condition): Likewise.
111 (vectorizable_comparison): Likewise.
112 (can_vectorize_live_stmts): Likewise.
113 (vect_analyze_stmt): Likewise.
114 (vect_transform_stmt): Adjust calls to vectorizable_*.
115 * tree-vectorizer.c: Include gimple-pretty-print.h.
116 (dump_stmt_cost): New function.
118 2018-05-16 Richard Biener <rguenther@suse.de>
120 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
121 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
122 * tree-ssa-dse.c: Include tree-ssa-loop.h.
123 (check_name): New callback.
124 (dse_classify_store): Track cycles via a visited bitmap of PHI
125 defs and simplify handling of in-loop and across loop dead stores
126 and properly fail for loop-variant refs. Handle byte-tracking with
127 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
130 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
132 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
133 (vect_get_mask_type_for_stmt): Likewise.
134 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
136 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
137 to determine the statement's vector type and the vector type that
138 should be used for calculating nunits. Deal with cases in which
139 the type has to be deferred.
140 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
141 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
142 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
143 (vect_determine_vf_for_stmt): New functions, split out from...
144 (vect_determine_vectorization_factor): ...here.
145 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
146 (vect_get_mask_type_for_stmt): New functions, split out from
147 vect_determine_vectorization_factor.
149 2018-05-16 Richard Biener <rguenther@suse.de>
151 * tree-cfg.c (verify_gimple_assign_ternary): Properly
152 verify the [VEC_]COND_EXPR embedded comparison.
154 2018-05-15 Martin Sebor <msebor@redhat.com>
156 PR tree-optimization/85753
157 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
158 RECORD_TYPE in addition to ARRAY_TYPE.
160 2018-05-15 Martin Sebor <msebor@redhat.com>
163 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
165 2018-05-15 Richard Biener <rguenther@suse.de>
167 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
168 add by_clobber_p one. Change algorithm to collect all defs
169 representing uses we need to walk and try reducing them to
170 a single one before failing.
171 (dse_dom_walker::dse_optimize_stmt): Adjust.
173 2018-05-13 Mark Wielaard <mark@klomp.org>
175 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
176 (size_of_loc_descr): Likewise.
177 (output_loc_operands): Likewise.
178 (output_loc_operands_raw): Likewise.
179 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
180 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
181 (hash_loc_operands): Likewise.
182 (compare_loc_operands): Likewise.
184 2018-05-14 Mark Wielaard <mark@klomp.org>
186 * dwarf2out.c (count_index_addrs): New function.
187 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
189 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
191 PR tree-optimization/83648
192 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
193 return value as malloc candidate.
195 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
198 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
199 as true in call to suggest_attribute.
201 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
203 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
206 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
208 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
210 (aarch64_crypto_aesd_fused): Likewise.
212 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
214 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
215 (movsi_aarch64): Likewise.
216 (load_pairsi): Likewise.
217 (load_pairdi): Likewise.
218 (store_pairsi): Likewise.
219 (store_pairdi): Likewise.
220 (load_pairsf): Likewise.
221 (load_pairdf): Likewise.
222 (store_pairsf): Likewise.
223 (store_pairdf): Likewise.
224 (zero_extend): Likewise.
225 (trunc): Swap alternatives.
226 (fcvt_target): Add '?' to prefer w over r.
228 2018-05-14 Jakub Jelinek <jakub@redhat.com>
231 * config/i386/i386.md: Disallow non-commutative arithmetics in
232 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
233 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
234 in the peephole2 before it.
236 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
238 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
239 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
240 (ix86_handle_option): Handle -mcldemote.
241 * config.gcc: New header.
242 * config/i386/cldemoteintrin.h: New file.
243 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
244 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
246 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
247 OPTION_MASK_ISA_CLDEMOTE.
248 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
249 (ix86_valid_target_attribute_inner_p): Ditto.
250 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
251 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
252 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
253 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
254 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
256 * config/i386/i386.opt: Add -mcldemote.
257 * config/i386/x86intrin.h: New header.
258 * doc/invoke.texi: Add -mcldemote.
260 2018-05-14 Richard Biener <rguenther@suse.de>
262 * doc/match-and-simplify.texi: Adjust :s documentation.
264 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
266 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
267 intended memcpy size.
268 (REORDER_45): Likewise.
270 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
273 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
274 * vec.c (qsort_chk): Use gcc_qsort.
275 * Makefile.in (OBJS-libcommon): Add sort.o.
276 (build/sort.o): New target. Use it...
277 (BUILD_RTL): ... here, and...
278 (build/gencfn-macros): ... here, and...
279 (build/genmatch): ... here.
281 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
282 Chung-Ju Wu <jasonwucj@gmail.com>
284 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
285 * config/nds32/nds32-graywolf.md: New file.
286 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
287 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
289 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
290 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
291 * config/nds32/nds32.md (pipeline_model): Add graywolf.
292 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
293 * config/nds32/pipelines.md: Include n15 settings.
295 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
296 Chung-Ju Wu <jasonwucj@gmail.com>
298 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
299 * config/nds32/nds32-n13.md: New file.
300 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
301 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
303 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
304 * config/nds32/nds32.md (pipeline_model): Add n13.
305 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
306 * config/nds32/pipelines.md: Include n13 settings.
308 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
309 Chung-Ju Wu <jasonwucj@gmail.com>
311 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
312 * config/nds32/nds32-n10.md: New file.
313 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
314 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
316 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
317 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
318 * config/nds32/nds32.md (pipeline_model): Add n10.
319 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
320 * config/nds32/pipelines.md: Include n10 settings.
322 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
323 Kito Cheng <kito.cheng@gmail.com>
324 Chung-Ju Wu <jasonwucj@gmail.com>
326 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
327 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
328 Add enum values for DSP extension instructions.
329 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
331 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
332 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
334 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
335 * config/nds32/nds32-dspext.md: New file for DSP implementation.
336 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
337 * config/nds32/nds32-intrinsic.md: Likewise.
338 * config/nds32/nds32_intrinsic.h: Likewise.
339 * config/nds32/nds32-md-auxiliary.c: Likewise.
340 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
341 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
342 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
343 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
344 * config/nds32/nds32-protos.h: New declarations for DSP extension.
345 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
346 TYPE_DMAC in switch statement.
347 * config/nds32/nds32.c: New checking and implementation for DSP
348 extension instructions.
349 * config/nds32/nds32.h: Likewise.
350 * config/nds32/nds32.md: Likewise.
351 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
352 * config/nds32/predicates.md: Implement new predicates for DSP
355 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
357 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
358 Reformat alternatives and attributes so it is easier to identify
359 which constraints/attributes go with which instruction.
360 (mov<mode>_hardfloat32, FMOVE64): Likewise.
361 (mov<mode>_softfloat32, FMOVE64): Likewise.
362 (mov<mode>_hardfloat64, FMOVE64): Likewise.
363 (mov<mode>_softfloat64, FMOVE64): Likewise.
365 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
367 * doc/extend.texi (PowerPC Built-in Functions): Rename this
369 (Basic PowerPC Built-in Functions): The new name of the
370 subsection previously known as "PowerPC Built-in Functions".
371 (Basic PowerPC Built-in Functions Available on all Configurations):
373 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
374 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
375 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
376 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
378 2018-05-11 Martin Jambor <mjambor@suse.cz>
381 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
384 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
387 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
389 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
391 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
392 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
393 (ix86_handle_option): Handle -mwaitpkg.
394 * config.gcc: New header.
395 * config/i386/cpuid.h (bit_WAITPKG): New bit.
396 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
397 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
399 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
400 OPTION_MASK_ISA_WAITPKG.
401 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
402 (ix86_option_override_internal): Add PTA_WAITPKG.
403 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
404 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
406 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
407 __builtin_ia32_umwait and __builtin_ia32_tpause.
408 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
409 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
410 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
411 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
412 UNSPECV_TPAUSE): New.
413 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
414 * config/i386/i386.opt: Add -mwaitpkg.
415 * config/i386/waitpkgintrin.h: New file.
416 * config/i386/x86intrin.h: New header.
417 * doc/invoke.texi: Add -mwaitpkg.
419 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
422 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
424 (cortex-m0): Use armv6s-m isa.
425 (cortex-m0plus): Likewise.
426 (cortex-m1): Likewise.
427 (cortex-m0.small-multiply): Likewise.
428 (cortex-m0plus.small-multiply): Likewise.
429 (cortex-m1.small-multiply): Likewise.
431 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
432 Jakub Jelinek <jakub@redhat.com>
434 PR tree-optimization/85692
435 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
436 source permute as well.
438 2018-05-11 Martin Liska <mliska@suse.cz>
441 * doc/extend.texi: Document LLVM style format for no_sanitize
444 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
446 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
447 mode_supports_vsx_dform_quad to mode_supports_dq_form.
448 (mode_supports_vsx_dform_quad): Likewise.
449 (mode_supports_vmx_dform): Move these functions to be next to the
450 other mode_supports functions.
451 (mode_supports_dq_form): Likewise.
452 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
453 mode_supports_dq_form.
454 (reg_offset_addressing_ok_p): Likewise.
455 (offsettable_ok_by_alignment): Likewise.
456 (rs6000_legitimate_offset_address_p): Likewise.
457 (legitimate_lo_sum_address_p): Likewise.
458 (rs6000_legitimize_address): Likewise.
459 (rs6000_legitimize_reload_address): Likewise.
460 (rs6000_secondary_reload_inner): Likewise.
461 (rs6000_preferred_reload_class): Likewise.
462 (rs6000_output_move_128bit): Likewise.
464 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
466 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
467 Generate SImode target register for null target.
468 <case IX86_BUILTIN_XGETBV>: Ditto.
469 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
470 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
472 2018-05-10 Carl Love <cel@us.ibm.com>
474 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
475 dcbtt and dcbtstt if operands[2] is 0.
477 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
480 * config/i386/sse.md (usadv64qi): New expander.
482 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
484 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
485 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
486 -maltivec=be support.
487 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
488 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
489 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
490 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
491 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
492 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
493 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
494 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
495 altivec_vsumsws): Adjust.
496 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
497 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
498 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
500 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
501 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
502 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
503 (altivec_lve<VI_char>x): Delete expand.
504 (*altivec_lve<VI_char>x_internal): Rename to...
505 (altivec_lve<VI_char>x): ... this.
506 (altivec_lvxl_<mode>): Delete expand.
507 (*altivec_lvxl_<mode>_internal): Rename to ...
508 (altivec_lvxl_<mode>): ... this.
509 (altivec_stvxl_<mode>): Delete expand.
510 (*altivec_stvxl_<mode>_internal): Rename to ...
511 (altivec_stvxl_<mode>): ... this.
512 (altivec_stve<VI_char>x): Delete expand.
513 (*altivec_stve<VI_char>x_internal): Rename to ...
514 (altivec_stve<VI_char>x): ... this.
515 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
516 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
517 reduc_plus_scal_<mode>): Adjust.
518 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
520 (rs6000_cpu_cpp_builtins): Adjust.
521 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
522 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
523 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
524 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
525 -maltivec=be support.
526 (rs6000_split_vec_extract_var): Adjust.
527 (rs6000_split_v4si_init): Adjust.
528 (swap_selector_for_mode): Delete.
529 (altivec_expand_lvx_be, altivec_expand_stvx_be,
530 altivec_expand_stvex_be): Delete.
531 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
532 -maltivec=be support.
533 (rs6000_gimple_fold_builtin): Ditto.
534 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
536 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
537 (TARGET_DIRECT_MOVE_64BIT): Adjust.
538 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
539 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
540 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
541 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
542 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
543 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
544 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
545 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
546 anonymous split): Adjust.
547 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
548 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
550 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
552 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
553 when --with-gxx-include-dir is also specified.
554 * configure: Regenerate.
556 2018-05-09 Jim Wilson <jimw@sifive.com>
559 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
560 * config/riscv/t-withmultilib: New.
561 * config/riscv/withmultilib.h: New.
562 * doc/install.texi: Document RISC-V --with-multilib-list support.
564 2018-05-09 Richard Biener <rguenther@suse.de>
566 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
568 (vect_bb_vectorization_profitable_p): Adjust. Compute
569 actual scalar cost using the cost vector and the add_stmt_cost
572 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
574 PR rtl-optimization/85645
575 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
576 in the REG_CFA_REGISTER note for LR, don't leave it empty.
578 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
580 PR rtl-optimization/85645
581 * shrink-wrap.c (spread_components): Return a boolean saying if
582 anything was changed.
583 (try_shrink_wrapping_separate): Iterate spread_components until
584 nothing changes anymore.
586 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
588 PR rtl-optimization/85645
589 * regrename.c (build_def_use): Also kill the chains that include the
590 destination of a REG_CFA_REGISTER note.
592 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
594 PR rtl-optimization/85645
595 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
596 insn that has a REG_CFA_REGISTER note.
598 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
600 * cfgexpand.c (expand_clobber): New function.
601 (expand_gimple_stmt_1): Use it.
602 * tree-vect-stmts.c (vect_clobber_variable): New function,
604 (vectorizable_simd_clone_call): ...here.
605 (vectorizable_store): Emit a clobber either side of an
606 IFN_STORE_LANES sequence.
607 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
609 2018-05-09 Tom de Vries <tom@codesourcery.com>
612 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
613 (define_insn "trap_if_false"): Add exit after trap.
615 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
617 PR rtl-optimization/85638
618 * bb-reorder.c: Include common/common-target.h.
619 (create_forwarder_block): New function extracted from...
620 (fix_up_crossing_landing_pad): ...here. Rename into...
621 (dw2_fix_up_crossing_landing_pad): ...this.
622 (sjlj_fix_up_crossing_landing_pad): New function.
623 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
624 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
625 from both partitions and exit the loop after one iteration.
627 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
630 * doc/extend.texi (PowerPC Built-in Functions): Rename this
632 (Basic PowerPC Built-in Functions): The new name of the
633 subsection previously known as "PowerPC Built-in Functions".
634 (Basic PowerPC Built-in Functions Available on all Configurations):
636 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
638 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
640 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
642 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
645 2018-05-08 Jim Wilson <jimw@sifive.com>
647 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
648 (LD_EMUL_SUFFIX): New.
651 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
653 * doc/extend.texi (PowerPC Built-in Functions): Rename this
655 (Basic PowerPC Built-in Functions): The new name of the
656 subsection previously known as "PowerPC Built-in Functions".
657 (Basic PowerPC Built-in Functions Available on all Configurations):
659 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
661 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
663 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
665 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
668 2018-05-08 Jakub Jelinek <jakub@redhat.com>
671 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
672 after cmpelim optimization.
674 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
676 * config.gcc: Support "goldmont".
677 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
678 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
680 * config/i386/i386.c (m_GOLDMONT): Define.
681 (processor_target_table): Add "goldmont".
682 (PTA_GOLDMONT): Define.
683 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
684 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
685 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
686 (fold_builtin_cpu): Add "goldmont".
687 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
688 (ix86_option_override_internal): Add "goldmont".
689 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
690 (processor_type): Add PROCESSOR_GOLDMONT.
691 * config/i386/i386.md: Add CPU "glm".
692 * config/i386/glm.md: New file.
693 * config/i386/x86-tune.def: Add m_GOLDMONT.
694 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
696 2018-05-08 Jakub Jelinek <jakub@redhat.com>
699 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
701 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
702 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
703 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
706 * config/i386/i386.c (ix86_fold_builtin): Handle
707 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
710 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
711 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
713 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
716 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
717 (check_arch): Likewise.
718 (check_fpu): Return the result rather than printing it.
719 (end arch): Fix operator precedence.
721 (END): Print the result from check_fpu.
723 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
724 Alan Hayward <alan.hayward@arm.com>
725 David Sherwood <david.sherwood@arm.com>
727 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
728 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
729 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
730 (*fcmuo<mode>_and): New patterns.
732 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
734 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
735 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
736 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
737 (cmp_op, sve_imm_con): New code attributes.
738 (SVE_COND_INT_CMP, imm_con): Delete.
739 (cmp_op): Remove above unspecs from int attribute.
740 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
742 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
743 comparison-specific unspecs.
744 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
745 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
746 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
747 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
748 (*vec_fcm<cmp_op><mode>): Rename to...
749 (*fcm<cmp_op><mode>): ...this and adjust likewise.
750 (*vec_fcmuo<mode>): Rename to...
751 (*fcmuo<mode>): ...this and adjust likewise.
752 (*pred_fcm<cmp_op><mode>): New pattern.
753 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
754 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
756 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
758 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
759 (aarch64_emit_sve_predicated_cond): New function.
760 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
761 (aarch64_emit_unspec_cond_or): Replace with...
762 (aarch64_emit_sve_or_conds): ...this new function. Use
763 aarch64_emit_sve_ptrue_op for the individual comparisons and
764 aarch64_emit_binop to OR them together.
765 (aarch64_emit_inverted_unspec_cond): Replace with...
766 (aarch64_emit_sve_inverted_cond): ...this new function. Use
767 aarch64_emit_sve_ptrue_op for the comparison and
768 aarch64_emit_unop to invert the result.
769 (aarch64_expand_sve_vec_cmp_float): Update after the above
770 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
772 2018-05-07 Nathan Sidwell <nathan@acm.org>
774 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
775 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
776 (Backwards Compatibility): Likewise.
778 2018-05-07 Luis Machado <luis.machado@linaro.org>
782 2018-05-07 Luis Machado <luis.machado@linaro.org>
784 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
785 <prefetch_dynamic_strides>: New const bool field.
786 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
787 prefetch_dynamic_strides.
788 (exynosm1_prefetch_tune): Likewise.
789 (thunderxt88_prefetch_tune): Likewise.
790 (thunderx_prefetch_tune): Likewise.
791 (thunderx2t99_prefetch_tune): Likewise.
792 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
793 (aarch64_override_options_internal): Update to set
794 PARAM_PREFETCH_DYNAMIC_STRIDES.
795 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
796 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
797 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
798 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
799 prefetch-dynamic-strides setting.
801 2018-05-07 Luis Machado <luis.machado@linaro.org>
803 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
804 <minimum_stride>: New const int field.
805 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
806 minimum_stride field.
807 (exynosm1_prefetch_tune): Likewise.
808 (thunderxt88_prefetch_tune): Likewise.
809 (thunderx_prefetch_tune): Likewise.
810 (thunderx2t99_prefetch_tune): Likewise.
811 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
812 (aarch64_override_options_internal): Update to set
813 PARAM_PREFETCH_MINIMUM_STRIDE.
814 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
815 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
816 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
817 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
818 stride is constant and is below the minimum stride threshold.
820 2018-05-07 Luis Machado <luis.machado@linaro.org>
822 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
825 2018-05-07 Luis Machado <luis.machado@linaro.org>
827 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
828 <prefetch_dynamic_strides>: New const bool field.
829 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
830 prefetch_dynamic_strides.
831 (exynosm1_prefetch_tune): Likewise.
832 (thunderxt88_prefetch_tune): Likewise.
833 (thunderx_prefetch_tune): Likewise.
834 (thunderx2t99_prefetch_tune): Likewise.
835 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
836 (aarch64_override_options_internal): Update to set
837 PARAM_PREFETCH_DYNAMIC_STRIDES.
838 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
839 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
840 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
841 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
842 prefetch-dynamic-strides setting.
844 2018-05-07 Luis Machado <luis.machado@linaro.org>
846 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
847 <minimum_stride>: New const int field.
848 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
849 minimum_stride field.
850 (exynosm1_prefetch_tune): Likewise.
851 (thunderxt88_prefetch_tune): Likewise.
852 (thunderx_prefetch_tune): Likewise.
853 (thunderx2t99_prefetch_tune): Likewise.
854 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
855 (aarch64_override_options_internal): Update to set
856 PARAM_PREFETCH_MINIMUM_STRIDE.
857 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
858 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
859 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
860 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
861 stride is constant and is below the minimum stride threshold.
863 2018-05-06 Jakub Jelinek <jakub@redhat.com>
866 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
867 the type is addressable. Don't force op into register if it has
870 2018-05-05 Roland McGrath <mcgrathr@google.com>
873 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
874 any section for which we don't know a specific type it should have,
875 regardless of name. Previously this was done only for the exact
876 names ".init_array", ".fini_array", and ".preinit_array".
877 (default_elf_asm_named_section): Add comment about
878 relationship with default_section_type_flags and SECTION_NOTYPE.
879 (get_section): Don't consider it a type conflict if one side has
880 SECTION_NOTYPE and the other doesn't, as long as neither has the
881 SECTION_BSS et al used in the default_section_type_flags logic.
883 2018-05-05 Tom de Vries <tom@codesourcery.com>
886 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
887 (workaround_barsyncs): New function.
888 (nvptx_reorg): Use workaround_barsyncs.
889 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
890 (define_expand "nvptx_membar_cta"): New define_expand.
891 (define_insn "*nvptx_membar_cta"): New insn.
893 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
895 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
896 To improve optimization opportunities.
897 * builtin-types.def: The new needed builtin types for the above.
899 2018-05-04 Richard Biener <rguenther@suse.de>
901 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
902 * gimple-ssa-store-merging.c
903 (imm_store_chain_info::output_merged_store): Remove redundant create,
904 release split_store vector contents on failure.
905 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
906 scalar stmt vector on cache hit.
908 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
910 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
912 * config.gcc (powerpc-xilinx-eabi*): Remove.
913 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
915 (fusion_addis_mem_combo_load): Ditto.
916 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
918 (rs6000_cpu_cpp_builtins): Ditto.
919 * config/rs6000/rs6000-linux.c
920 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
921 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
922 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
924 (rs6000_setup_reg_addr_masks): Ditto.
925 (rs6000_init_hard_regno_mode_ok): Ditto.
926 (rs6000_option_override_internal): Ditto.
927 (legitimate_lo_sum_address_p): Ditto.
928 (rs6000_legitimize_address): Ditto.
929 (rs6000_legitimize_reload_address): Ditto.
930 (rs6000_legitimate_address_p): Ditto.
931 (abi_v4_pass_in_fpr): Ditto.
932 (setup_incoming_varargs): Ditto.
933 (rs6000_gimplify_va_arg): Ditto.
934 (rs6000_split_multireg_move): Ditto.
935 (rs6000_savres_strategy): Ditto.
936 (rs6000_emit_prologue_components): Ditto.
937 (rs6000_emit_epilogue_components): Ditto.
938 (rs6000_emit_prologue): Ditto.
939 (rs6000_emit_epilogue): Ditto.
940 (rs6000_elf_file_end): Ditto.
941 (rs6000_function_value): Ditto.
942 (rs6000_libcall_value): Ditto.
943 * config/rs6000/rs6000.h: Ditto.
944 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
945 (TARGET_MINMAX): ... this. New.
946 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
947 * config/rs6000/rs6000.md: Remove Xilinx FP support.
948 (*movsi_internal1_single): Delete.
949 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
950 mfpu=, mxilinx-fpu): Delete.
951 * config/rs6000/singlefp.h: Delete.
952 * config/rs6000/sysv4.h: Remove Xilinx FP support.
953 * config/rs6000/t-rs6000: Ditto.
954 * config/rs6000/t-xilinx: Delete.
955 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
956 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
957 (VStype_simple): Delete.
958 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
959 * config/rs6000/xfpu.h: Delete.
960 * config/rs6000/xfpu.md: Delete.
961 * config/rs6000/xilinx.h: Delete.
962 * config/rs6000/xilinx.opt: Delete.
963 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
964 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
966 2018-05-04 Tom de Vries <tom@codesourcery.com>
969 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
972 2018-05-04 Richard Biener <rguenther@suse.de>
975 * tree-complex.c (update_complex_assignment): We are always in SSA form.
976 (expand_complex_div_wide): Likewise.
977 (expand_complex_operations_1): Likewise.
978 (expand_complex_libcall): Preserve EH info of the original stmt.
979 (tree_lower_complex): Handle removed blocks.
980 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
981 on complex multiplication and division libcall builtins.
983 2018-05-04 Richard Biener <rguenther@suse.de>
986 * fold-const.c (negate_expr_p): Restrict negation of operand
987 zero of a division to when we know that can happen without
989 (fold_negate_expr_1): Likewise.
991 2018-05-04 Jakub Jelinek <jakub@redhat.com>
994 * real.h (real_nextafter): Declare.
995 * real.c (real_nextafter): New function.
996 * fold-const-call.c (fold_const_nextafter): New function.
997 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
999 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1000 even when arg1_mode is different from arg0_mode.
1002 2018-05-03 Nathan Sidwell <nathan@acm.org>
1004 * doc/extend.texi (Deprecated Features): Remove
1006 (Backwards Compatibility): Likewise.
1007 * doc/invoke.texi (C++ Language Options): Likewise.
1008 (C++ Dialect Options): Likewise.
1010 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1013 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1014 _mm512_mask_mullox_epi64): New intrinsics.
1016 2018-05-03 Tom de Vries <tom@codesourcery.com>
1019 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1020 dump files): Add offload-tree.
1022 2018-05-03 Richard Biener <rguenther@suse.de>
1024 PR tree-optimization/85615
1025 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1026 to loops not nested in BBs loop father to avoid creating multi-entry
1029 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1031 PR tree-optimization/70291
1032 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1033 arguments. Change return type to tree. Emit libcall as a new
1034 statement rather than replacing existing one when inplace_p is true.
1035 (expand_complex_multiplication_components): New function.
1036 (expand_complex_multiplication): Expand floating-point complex
1037 multiplication using the above.
1038 (expand_complex_division): Rename inner_type parameter to type.
1039 Update expand_complex_libcall call-site.
1040 (expand_complex_operations_1): Update expand_complex_multiplication
1041 and expand_complex_division call-sites.
1043 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1046 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1047 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1048 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1049 the highest significant bit of the shift count mask is clear. In
1050 check whether and[sq]i3 is needed verify that all significant bits
1051 of the shift count other than the highest are set.
1053 2018-05-02 Tom de Vries <tom@codesourcery.com>
1056 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1057 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1058 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1059 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1060 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1061 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1062 __builtin_goacc_parlevel_size.
1064 2018-05-02 Richard Biener <rguenther@suse.de>
1066 PR tree-optimization/85597
1067 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1068 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1071 2018-05-02 Tom de Vries <tom@codesourcery.com>
1074 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1075 dump files): Add ltrans-tree.
1077 2018-05-02 Tom de Vries <tom@codesourcery.com>
1080 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1081 dump files): Add wpa-ipa.
1083 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1085 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1086 powerpc*-*-linux*paired* target.
1087 * config/rs6000/750cl.h: Delete.
1088 * config/rs6000/paired.h: Delete.
1089 * config/rs6000/paired.md: Delete.
1090 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1092 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1093 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1094 comment. Remove paired float support.
1095 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1096 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1098 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1099 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1101 * config/rs6000/rs6000.c: Remove paired float support.
1102 (paired_expand_vector_init, paired_expand_vector_move,
1103 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1104 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1105 paired_expand_builtin, paired_expand_predicate_builtin,
1106 paired_init_builtins): Delete.
1107 * config/rs6000/rs6000.h: Remove paired float support.
1108 * config/rs6000/rs6000.md: Remove paired float support.
1109 (move_from_CR_ov_bit): Delete.
1110 * config/rs6000/rs6000.opt (mpaired): Delete.
1111 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1112 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1114 2018-05-02 Richard Biener <rguenther@suse.de>
1117 * gimplify.c (gimplify_save_expr): When in SSA form allow
1118 SAVE_EXPRs to compute to SSA vars.
1120 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1123 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1124 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1125 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1126 clobber operands[2], instead use a new pseudo. Formatting fixes.
1128 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1130 PR tree-optimization/85586
1131 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1132 exit early for statements in the same group if the accesses are
1135 2018-05-02 Tom de Vries <tom@codesourcery.com>
1138 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1141 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1143 PR tree-optimization/85143
1144 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1146 2018-05-01 Tom de Vries <tom@codesourcery.com>
1149 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1150 not found" error message.
1152 2018-05-01 Tom de Vries <tom@codesourcery.com>
1155 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1156 * vec.c (test_ordered_remove_if): New function.
1157 (vec_c_tests): Call test_ordered_remove_if.
1158 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1159 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1160 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1161 VEC_ORDERED_REMOVE_IF.
1163 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1165 PR tree-optimization/82665
1166 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1167 pointer subtraction where arguments come from a memchr call.
1169 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1171 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1172 --push-state --as-needed and --pop-state instead of --as-needed and
1173 --no-as-needed if ld supports it.
1174 * configure: Regenerated.
1177 * doc/install.texi2html: Replace _002d with - and _002a with * in
1178 generated html files using sed.
1180 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1183 * gcc-rich-location.c (blank_line_before_p): New function.
1184 (use_new_line): New function.
1185 (gcc_rich_location::add_fixit_insert_formatted): New function.
1186 * gcc-rich-location.h
1187 (gcc_rich_location::add_fixit_insert_formatted): New function.
1189 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1191 * selftest.c (assert_streq): Rename "expected" and "actual" to
1192 "val1" and "val2". Extend NULL-handling to cover both inputs
1193 symmetrically, while still requiring both to be non-NULL for a pass.
1194 * selftest.h (assert_streq): Rename "expected" and "actual" to
1196 (ASSERT_EQ): Likewise.
1197 (ASSERT_EQ_AT): Likewise.
1198 (ASSERT_KNOWN_EQ): Likewise.
1199 (ASSERT_KNOWN_EQ_AT): Likewise.
1200 (ASSERT_NE): Likewise.
1201 (ASSERT_MAYBE_NE): Likewise.
1202 (ASSERT_MAYBE_NE_AT): Likewise.
1203 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1204 the assertion to pass.
1205 (ASSERT_STREQ_AT): Likewise.
1207 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1209 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1210 interaction with -pie.
1212 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1214 * selftest.h: Fix alphabetization of per-source-file selftest
1217 2018-04-30 Jason Merrill <jason@redhat.com>
1219 PR c++/61982 - dead stores to destroyed objects.
1220 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1223 2018-04-30 Jason Merrill <jason@redhat.com>
1225 * tree.c (build_clobber): New.
1226 * tree.h: Declare it.
1227 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1229 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1231 * diagnostic-show-locus.c (layout::layout): Update for
1232 location_get_source_line returning a char_span.
1233 (struct char_span): Move to input.h.
1234 (struct correction): Update for fields in char_span becoming
1236 (struct source_line): Update for location_get_source_line
1237 returning a char_span.
1238 (layout::print_line): Likewise.
1239 * edit-context.c (edited_file::print_content): Likewise.
1240 (edited_file::print_diff_hunk): Likewise.
1241 (edited_file::print_run_of_changed_lines): Likewise.
1242 (edited_file::get_num_lines): Likewise.
1243 (edited_line::edited_line): Likewise.
1244 * final.c (asm_show_source): Likewise.
1245 * input.c (location_get_source_line): Convert return type
1246 from const char * to char_span, losing the final "line_len"
1248 (dump_location_info): Update for the above.
1249 (get_substring_ranges_for_loc): Likewise. Use a char_span
1250 when handling the literal within the line.
1251 (test_reading_source_line): Update for location_get_source_line
1252 returning a char_span.
1253 * input.h (class char_span): Move here from
1254 diagnostic-show-locus.c, converting from a struct to a class.
1255 Make data members private.
1256 (char_span::operator bool): New.
1257 (char_span::length): New.
1258 (char_span::get_buffer): New.
1259 (char_span::operator[]): New.
1260 (char_span::subspan): Make const.
1261 (char_span::xstrdup): New.
1262 (location_get_source_line): Convert return type from const char *
1263 to char_span, losing the final "line_size" param.
1265 2018-04-30 Jan Hubicka <jh@suse.cz>
1267 * lto-wrapper.c (ltrans_priorities): New static var.
1268 (cmp_priority): New.
1269 (run_gcc): Read priorities and if doing parallel build order
1270 the Makefile by them.
1272 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1274 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1276 2018-04-30 Richard Biener <rguenther@suse.de>
1278 * tree-cfg.c (verify_address): Remove base argument, add
1279 flag whether to check TREE_ADDRESSABLE and do that.
1280 (verify_expr): Remove.
1281 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1282 (verify_gimple_assign_single): Likewise.
1283 (verify_gimple_switch): Likewise.
1284 (verify_expr_location_1): Dereference tp once. Add (disabled)
1285 piece from verify_expr.
1286 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1288 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1290 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1292 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1294 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1295 (small_data_pattern): Likewise.
1296 (arc_rewrite_small_data): Likewise.
1297 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1298 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1299 (get_symbol_alignment): New function.
1300 (legitimate_small_data_address_p): Likewise.
1301 (legitimate_scaled_address): Update, call
1302 legitimate_small_data_address_p.
1303 (output_sdata): New static variable.
1304 (arc_print_operand): Update how we handle small data operands.
1305 (arc_print_operand_address): Likewise.
1306 (arc_legitimate_address_p): Update, use
1307 legitimate_small_data_address_p.
1308 (arc_rewrite_small_data_p): Remove.
1309 (arc_rewrite_small_data_1): Likewise.
1310 (arc_rewrite_small_data): Likewise.
1311 (small_data_pattern): Likewise.
1312 (compact_sda_memory_operand): Update to use
1313 legitimate_small_data_address_p and get_symbol_alignment.
1314 (prepare_move_operands): Don't rewite sdata pattern.
1315 (prepare_extend_operands): Remove.
1316 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1318 (zero_extendqisi2): Likewise.
1319 (zero_extendhisi2): Likewise.
1320 (extendqihi2): Likewise.
1321 (extendqisi2): Likewise.
1322 (extendhisi2): Likewise.
1326 * config/arc/constraints.md (Usd): Change it to memory constraint.
1328 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1330 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1331 as source of std instructions.
1332 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1333 allow 6-bit constants as source for store instructions.
1334 (movdi_insn): Update instruction pattern to allow 6-bit constants
1335 as source for store instructions.
1337 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1339 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1341 2018-04-30 Nathan Sidwell <nathan@acm.org>
1342 Sandra Loosemore <sandra@codesourcery.com>
1344 * dumpfile.c (dump_open): Allow '-' for stdout.
1345 * doc/invoke.texi (Developer Options): Document dump filename
1346 determination early. Document stdin/stdout selection.
1348 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1350 Microblaze Target: PIC data text relative
1352 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1353 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1355 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1356 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1357 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1358 New addressing mode for data-text relative position indepenedent code.
1359 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1360 'ADDRESS_SYMBOLIC_TXT_REL'.
1361 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1362 (microblaze_legitimate_pic_operand): Exclude function calls from
1363 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1364 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1366 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1367 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1368 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1369 for 'address + offset'.
1370 (microblaze_expand_prologue): Add new function prologue call for
1372 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1373 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1374 table in case of TARGET_PIC_DATA_TEXT_REL.
1375 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1376 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1377 Add new macros 'UNSPEC_TEXT',
1378 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1379 + exclude function calls from 'UNSPEC_PLT' in case of data text
1381 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1382 new target hook for generating address diff vector tables in case of
1384 * doc/tm.texi : Regenerate.
1385 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1386 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1387 of addr diff vector generation.
1388 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1389 target hook definition.
1390 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1391 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1392 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1393 option for data text relative.
1395 2018-04-30 Richard Biener <rguenther@suse.de>
1397 * tree-chrec.h (evolution_function_is_constant_p): Remove
1399 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1401 2018-04-30 Richard Biener <rguenther@suse.de>
1404 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1406 2018-04-30 Richard Biener <rguenther@suse.de>
1408 PR tree-optimization/28364
1409 PR tree-optimization/85275
1410 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1411 copying first exit test.
1413 2018-04-28 Mark Wielaard <mark@klomp.org>
1415 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1417 (dwarf_AT): Handle DW_AT_addr_base.
1418 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1420 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1423 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1424 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1425 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1426 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1428 2018-04-28 Richard Biener <rguenther@suse.de>
1430 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1431 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1432 to reflect use. Only add interesting stmts.
1434 2018-04-27 Martin Jambor <mjambor@suse.cz>
1437 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1438 the jump function allows for passing through aggregate values.
1440 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1442 * input.h (in_system_header_at): Convert from macro to inline
1444 (from_macro_expansion_at): Likewise.
1445 (from_macro_definition_at): Likewise.
1447 2018-04-27 Jeff Law <law@redhat.com>
1449 * config.gcc: Mark tile* targets as deprecated/obsolete.
1451 2018-04-27 Richard Biener <rguenther@suse.de>
1453 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1456 2018-04-27 Richard Biener <rguenther@suse.de>
1458 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1460 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1462 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1463 with Yd constraint. Set "preferred_for_speed" attribute from
1464 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1466 (*movdi_internal): Ditto.
1467 (movti_interunit splitters): Remove
1468 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1469 (movdi_interunit splitters): Ditto.
1470 * config/i386/constraints.md (Ye): Remove.
1471 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1473 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1476 * config/aarch64/constraints.md (Usg): Limit to 31.
1479 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1481 PR tree-optimization/85529
1482 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1483 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1484 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1485 zero extension or masking of the MSB bit.
1486 (optimize_range_tests): Add FIRST_BB argument, pass it through
1487 to optimize_range_tests_var_bound.
1488 (maybe_optimize_range_tests, reassociate_bb): Adjust
1489 optimize_range_tests callers.
1491 2018-04-26 Richard Biener <rguenther@suse.de>
1492 Jakub Jelinek <jakub@redhat.com>
1494 * cgraph.h (symbol_table): Just declare debug method here.
1495 * symtab.c (symbol_table::debug): Define.
1497 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1499 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1501 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1503 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1504 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1505 (*movdi_internal): Substitute Yi and Yj constraint with x
1506 and Ym and Yn constraint with y constraint. Update "isa"
1507 attribute and set "preferred_for_speed" attribute from
1508 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1509 (*movsi_internal): Ditto.
1510 (*movdf_internal): Ditto.
1511 (*movsf_internal): Ditto.
1512 (*zero_extendsidi2): Ditto.
1513 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1514 (sse2_loadld): Ditto.
1515 (*vec_extract<ssevecmodelower>_0): Ditto.
1516 (*vec_extractv4si_0_zext_sse4): Ditto.
1517 (vec_concatv2di): Ditto.
1518 (*vec_dup<mode>): Ditto.
1519 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1520 * config/i386/constraints.md (Yi): Remove.
1525 2018-04-26 Nathan Sidwell <nathan@acm.org>
1527 * dumpfile.c (dump_open): New.
1528 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1529 (dump_finish): Detect stdio/stderr by value not name.
1531 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1533 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1535 2018-04-26 Tom de Vries <tom@codesourcery.com>
1538 * config/nvptx/nvptx.c (verify_neutering_jumps)
1539 (verify_neutering_labels): New function
1540 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1542 2018-04-26 Tom de Vries <tom@codesourcery.com>
1545 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1546 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1549 2018-04-26 Richard Biener <rguenther@suse.de>
1550 Tom de Vries <tom@codesourcery.com>
1553 * lto-streamer-out.c (output_function): Fixup loops if required to match
1554 discovery done in the reader.
1556 2018-04-26 Richard Biener <rguenther@suse.de>
1558 PR tree-optimization/85116
1559 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1560 have a loop exit from the single latch predecessor. Remove
1561 case of header with just condition.
1562 (ch_base::copy_headers): Exclude infinite loops from any
1564 (pass_ch::execute): Record exits.
1566 2018-04-26 Richard Biener <rguenther@suse.de>
1568 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1569 prologue cost vector and pass it to vect_get_load_cost.
1570 (vect_get_peeling_costs_all_drs): Likewise.
1571 (vect_peeling_hash_get_lowest_cost): Likewise.
1572 (vect_enhance_data_refs_alignment): Likewise.
1574 2018-04-26 Richard Biener <rguenther@suse.de>
1577 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1578 checking of integer<->pointer conversions.
1579 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1580 sign-/zero-extending pointer types.
1581 (expand_omp_for_static_chunk): Likewise.
1583 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1584 Jean Lee <xiaoyur347@gmail.com>
1586 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1587 (TARGET_ASAN_SHADOW_OFFSET): Define.
1588 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1589 true for -fsanitize=address.
1591 2018-04-25 Mark Wielaard <mark@klomp.org>
1593 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1596 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1598 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1599 than "alu", remove explicit "memory" and "imm_disp" attributes.
1600 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1603 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1604 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1605 gen_lowpart_no_emit.
1607 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1610 * config/i386/i386.c (ix86_expand_builtin): Change memory
1611 operand to XI, extend p0 to Pmode.
1612 * config/i386/i386.md: Change unspec volatile and operand
1613 1 mode to XI, change operand 0 mode to P.
1615 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1617 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1618 GET_MODE_MASK before any checking.
1619 (nds32_can_use_bset_p): Likewise.
1620 (nds32_can_use_btgl_p): Likewise.
1622 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1624 * config/nds32/nds32-doubleword.md: New define_split pattern for
1625 illegal register number.
1627 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1629 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1631 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1633 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1635 2018-04-25 Richard Biener <rguenther@suse.de>
1637 * lto-streamer.h (LTO_major_version): Bump to 8.
1639 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1641 * BASE-VER: Set to 9.0.0.
1643 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1645 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1646 in __abskf2 and __powikf2.
1648 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1651 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1652 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1653 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1654 Use the above on operand 2. Reindent.
1655 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1657 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1660 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1662 * config/i386/i386.opt (mcet): Removed.
1663 * doc/install.texi: Remove -mcet documentation.
1664 * doc/invoke.texi: Likewise.
1666 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1669 * doc/install.texi: Remove -mcet from bootstrap-cet.
1671 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1674 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1675 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1679 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1680 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1681 containing a CONST_VECTOR.
1683 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1685 * doc/install.texi: Update newlib dependency for nvptx.
1687 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1690 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1691 instead of INTVAL when shifting x left.
1693 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1695 PR tree-optimization/85478
1696 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1697 vect_grouped_store_supported for single element vectors.
1699 2018-04-24 Richard Biener <rguenther@suse.de>
1702 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1703 load cost increase to the case of non-constant step.
1705 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1708 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1709 destination if any_malformed_asm.
1711 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1714 * expr.c (store_field): In the bitfield case, if the value comes from
1715 a function call and is returned in registers by means of a PARALLEL,
1716 do not change the mode of the temporary unless BLKmode and VOIDmode.
1718 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1720 PR rtl-optimization/85423
1721 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1722 dependencies to debug insns when the previous insn is non-debug.
1724 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1726 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1727 enums into a single definition.
1728 (fls): Fix predicates and printing.
1731 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1733 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1734 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1735 and short u6 immediate.
1736 (check_if_valid_sleep_operand): Remove.
1737 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1739 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1741 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1742 flag_always_save_lp condition.
1743 * config/nds32/nds32.opt (malways-save-lp): New option.
1745 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1747 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1748 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1749 * config/nds32/nds32.h
1750 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1751 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1753 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1755 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1756 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1758 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1759 Chung-Ju Wu <jasonwucj@gmail.com>
1761 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1763 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1764 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1766 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1768 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1770 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1772 * config/nds32/nds32-protos.h (nds32_data_alignment,
1773 nds32_local_alignment): Declare.
1774 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1775 nds32_local_alignment): New functions.
1776 (TARGET_CONSTANT_ALIGNMENT): Define.
1777 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1779 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1781 * config/nds32/nds32.c
1782 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1783 (TARGET_MODES_TIEABLE_P): Likewise.
1785 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1787 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1790 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1791 Chung-Ju Wu <jasonwucj@gmail.com>
1793 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1794 for unaligned access.
1795 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1797 * config/nds32/nds32-intrinsic.md: Likewise.
1798 * config/nds32/nds32_intrinsic.h: Likewise.
1799 * config/nds32/nds32.h (nds32_builtins): Likewise.
1800 * config/nds32/nds32.opt (munaligned-access): New option.
1801 * config/nds32/nds32.c (nds32_asm_file_start): Display
1802 flag_unaligned_access status.
1804 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1806 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1807 -mno-relax is present.
1808 * config/riscv/linux.h (LINK_SPEC): Ditto.
1810 2018-04-20 Martin Sebor <msebor@redhat.com>
1813 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1815 (gimple_fold_builtin_stxcpy_chk): Same.
1816 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1818 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1821 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1822 __powikf2 when long double is IEEE 128-bit.
1824 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1826 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1827 step to make sure stack always aligned.
1829 2018-04-20 Carl Love <cel@us.ibm.com>
1832 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1833 size check for arg0.
1835 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1836 Tom de Vries <tom@codesourcery.com>
1839 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1840 Emit insns for calls too.
1841 (nvptx_find_par): Always look for worker-level predecessor insn.
1842 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1844 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1845 (nvptx_process_pars): Propagate frames for calls.
1847 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1850 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1852 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1853 (ix86_handle_option): Don't handle OPT_mibt.
1854 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1856 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1858 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1859 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1860 (ix86_target_macros): Define __CET__ with flag_cf_protection
1861 for -fcf-protection.
1862 * config/i386/i386.c (isa2_opts): Remove -mibt.
1863 * config/i386/i386.h (TARGET_IBT): Removed.
1864 (TARGET_IBT_P): Likewise.
1865 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1866 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1867 * config/i386/i386.opt (mcet): Update help message.
1870 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1871 -mcet as an alias for -mshstk.
1873 2018-04-20 Richard Biener <rguenther@suse.de>
1876 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1877 complexity by forcing a single use of the multiply operand.
1879 2018-04-20 Martin Jambor <mjambor@suse.cz>
1882 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1883 recursion dependency to only apply to non-clones.
1885 2018-04-20 Martin Jambor <mjambor@suse.cz>
1888 * ipa-cp.c (create_specialized_node): Check that clones of
1889 self-recursive edges exist during IPA-CP.
1891 2018-04-19 Toon Moene <toon@moene.org>
1893 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1896 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1898 PR tree-optimization/85467
1899 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1900 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1901 VECTOR_CST element to type.
1903 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1906 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1907 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1908 (builtin_longjmp): Likewise.
1909 (save_stack_nonlocal): New pattern.
1910 (restore_stack_nonlocal): Likewise.
1912 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1915 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1916 Replace ASM_OUTPUT_LABEL with fprintf.
1918 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1921 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1922 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1923 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1924 define __IBT__ and __SHSTK__ for -fcf-protection.
1925 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1927 (ix86_trampoline_init): Likewise.
1928 (x86_output_mi_thunk): Likewise.
1929 (ix86_notrack_prefixed_insn_p): Likewise.
1930 (ix86_option_override_internal): Don't disallow -fcf-protection.
1931 * config/i386/i386.md (rdssp<mode>): Also enable for
1933 (incssp<mode>): Likewise.
1934 (nop_endbr): Likewise.
1935 * config/i386/i386.opt (mcet): Change help message to built-in
1939 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1940 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1941 enable CET built-in functions.
1943 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1945 * common/config/i386/i386-common.c
1946 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1947 OPTION_MASK_ISA_MOVDIRI_UNSET,
1948 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1949 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1950 * config.gcc (movdirintrin.h): New header.
1951 * config/i386/cpuid.h (bit_MOVDIRI,
1952 bit_MOVDIR64B): New bits.
1953 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1955 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1956 (VOID, PVOID, PCVOID)): New function types.
1957 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1958 __builtin_ia32_directstoreu_u64,
1959 __builtin_ia32_movdir64b): New builtins.
1960 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1961 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1963 (ix86_valid_target_attribute_inner_p): Ditto.
1964 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1965 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1966 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1967 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1968 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1969 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1970 (movdiri<mode>, movdir64b_<mode>): New.
1971 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1972 * config/i386/immintrin.h: Include movdirintrin.h.
1973 * config/i386/movdirintrin.h: New file.
1974 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1976 2018-04-19 Richard Biener <rguenther@suse.de>
1979 * cfg.c (clear_bb_flags): When loop state says we have
1980 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1982 2018-04-19 Richard Biener <rguenther@suse.de>
1984 PR tree-optimization/84737
1985 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1986 copying restrict info.
1987 (vect_setup_realignment): Use it.
1988 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1989 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1990 the first DR to all generated stores.
1991 (vectorizable_load): Likewise for loads.
1993 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1995 PR tree-optimization/85446
1996 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1997 the integral and pointer types to have the same precision.
1999 * doc/install.texi: Document --disable-cet being the default and
2002 2018-04-18 Martin Liska <mliska@suse.cz>
2004 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2007 2018-04-18 Martin Liska <mliska@suse.cz>
2010 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2013 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2014 arguments if they are comparable.
2016 2018-04-18 Martin Liska <mliska@suse.cz>
2019 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2022 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2025 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2028 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2029 ENDBR after calling __morestack.
2031 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2034 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2035 by using gcc_base_ver to generate a gcc_driver_version, and use
2036 it when generating GCC_DRIVER_NAME.
2037 * configure: Regenerate.
2039 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2042 * config.gcc: Obsolete powerpc*-*-*spe*.
2044 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2047 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2048 (stabstr_D): Change type of unum from unsigned int to
2049 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2052 2018-04-17 Jim Wilson <jimw@sifive.com>
2055 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2056 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2057 Set arg_pointer_offset after using pretend_args_size.
2059 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2061 PR rtl-optimization/85431
2062 * dse.c (record_store): Ignore zero width stores.
2065 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2066 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2067 __builtin_stack_restore rather than after it.
2068 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2069 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2070 argument instead of virtual_dynamic_stack_rtx.
2072 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2074 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2076 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2077 Add note to error message to explain internal mapping of overloaded
2078 built-in function name to non-overloaded built-in function name.
2079 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2082 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2085 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2086 where the inputs overlap with the output.
2088 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2090 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2091 (=v, v) alternative and explicit "memory" attribute.
2092 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2093 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2095 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2096 "sselog1" type instead of "sselog".
2097 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2098 "sselog". Remove explicit "memory" attribute.
2099 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2100 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2102 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2103 "sselog1" type instead of "sselog". Remove explicit "memory"
2105 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2106 use "sselog1" type instead of "sselog". Remove explicit "memory"
2108 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2109 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2111 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2112 "sselog1" type instead of "sselog". Remove explicit "memory"
2114 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2115 use "sselog1" type instead of "sselog". Remove explicit "memory"
2119 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2122 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2125 2018-04-17 Martin Jambor <mjambor@suse.cz>
2128 * ipa-cp.c (create_specialized_node): Call
2129 expand_all_artificial_thunks if necessary.
2131 2018-04-17 Martin Liska <mliska@suse.cz>
2134 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2135 in message, remote space in between '_G' and '('.
2137 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2140 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2141 avx512f_vmcmp<mode>3<round_saeonly_name>,
2142 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2143 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2144 avx512f_rndscale<mode><round_saeonly_name>,
2145 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2146 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2147 Use %<iptr>2 instead of %2 for -masm=intel.
2148 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2149 avx512f_vcvttss2usi<round_saeonly_name>,
2150 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2152 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2153 avx512f_vcvttsd2usi<round_saeonly_name>,
2154 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2155 Use %q1 instead of %1 for -masm=intel.
2156 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2157 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2158 of %3 for -masm=intel.
2159 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2161 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2163 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2165 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2166 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2168 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2170 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2171 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2172 %g1 and one with %0 and %1.
2173 (avx512er_vmrcp28<mode><round_saeonly_name>,
2174 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2176 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2177 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2178 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2179 of %0 and %{%4%} for -masm=intel.
2180 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2181 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2182 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2183 order of %0 and %{%5%}%{z%} for -masm=intel.
2185 2018-04-17 Jan Hubicka <jh@suse.cz>
2188 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2190 2018-04-17 Martin Liska <mliska@suse.cz>
2193 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2194 for target_clone error message. Make default implementation
2195 clone to be a local declaration.
2196 (separate_attrs): Add new argument and check for an empty
2198 (expand_target_clones): Handle it.
2199 (ipa_target_clone): Make redirection just for target_clones
2202 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2203 Tom de Vries <tom@codesourcery.com>
2206 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2207 tiled basic blocks without omp continue statements.
2209 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2212 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2213 vec_extract expression as having side effects to make sure it gets
2216 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2219 * config/i386/i386.c (get_builtin_code_for_version): Check
2222 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2225 * gcc/config.gcc: Support "skylake".
2226 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2228 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2229 (processor_target_table): Add "skylake".
2230 (ix86_option_override_internal): Add "skylake".
2231 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2232 PROCESSOR_CANNONLAKE.
2233 (get_builtin_code_for_version): Fix priority for
2234 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2235 PROCESSOR_SKYLAKE-AVX512.
2236 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2237 (processor_type): Add PROCESSOR_SKYLAKE.
2239 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2240 Jason Merrill <jason@redhat.com>
2243 * convert.c (convert_to_integer_1): Use direct recursion for
2244 enumeral types and types with a precision less than the number
2245 of bits in their mode.
2247 2018-04-16 Julia Koval <julia.koval@intel.com>
2250 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2251 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2253 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2256 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2257 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2258 and -mno-direct-move.
2260 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2263 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2264 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2265 Ensure negative shifts result in {0}.
2267 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2269 PR rtl-optimization/79916
2270 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2271 regs (if any) to define how to gnerate SD moves when LRA is in
2274 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2276 PR rtl-optimization/85393
2277 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2278 * except.c (expand_dw2_landing_pad_for_region): Make static.
2279 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2280 a label and unconditional jump to old_bb, rather than
2281 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2284 PR rtl-optimization/85376
2285 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2286 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2287 instead of a specific value.
2289 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2290 Bin Cheng <bin.cheng@arm.com>
2292 PR tree-optimization/82965
2293 PR tree-optimization/83991
2294 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2295 by_profile_only parameter.
2296 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2297 information if the loop was predicted to iterate too many times.
2298 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2300 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2303 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2306 2018-04-13 Martin Liska <mliska@suse.cz>
2307 Jakub Jelinek <jakub@redhat.com>
2310 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2311 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2312 * builtins.c (expand_builtin_memory_copy_args): Use
2313 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2314 handle dest_addr == pc_rtx.
2316 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2319 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2320 asked to not generate direct moves.
2321 (fix_trunc<mode>si2_stfiwx): Similar.
2322 (fix_trunc<mode>si2_internal): Similar.
2324 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2327 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2328 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2329 lookup if dest in some wider mode is known to be const0_rtx and
2330 if so, record permanent equivalence for it to be ZERO_EXTEND of
2331 the narrower mode destination.
2333 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2335 * lto-streamer-out.c (output_function): Revert 259346.
2336 * omp-expand.c (expand_oacc_for): Likewise.
2338 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2340 PR rtl-optimization/85354
2341 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2342 * sel-sched.c (sel_global_init): ... here.
2344 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2347 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2348 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2349 mode for PE-COFF targets.
2350 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2351 (i386_pe_asm_lto_end): Likewise.
2352 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2353 (TARGET_ASM_LTO_END): Likewise.
2354 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2355 (i386_pe_asm_lto_start): New function.
2356 (i386_pe_asm_lto_end): Likewise.
2358 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2359 Richard Biener <rguenther@suse.de>
2362 * lto-streamer-out.c (output_function): Fix CFG loop state before
2364 * omp-expand.c (expand_oacc_for): Handle calls to internal
2365 functions like regular functions.
2367 2018-04-12 Richard Biener <rguenther@suse.de>
2370 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2371 for the early LTO debug to properly generate references to it
2372 during DIE emission. Do not re-use that for the skeleton for
2374 (dwarf2out_early_finish): Likewise.
2376 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2379 * config/i386/sse.md
2380 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2381 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2382 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2383 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2384 and output is a reg, avoid creating invalid lowpart subreg, but
2385 instead split into a 512-bit move. Don't split if not AVX512VL,
2386 input is xmm16+ reg and output is a mem.
2387 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2388 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2389 xmm16+ reg and output is a mem.
2391 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2393 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2394 also for flag_dwarf2_cfi_asm.
2396 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2398 PR rtl-optimization/85342
2399 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2400 a bool scalar var inside of the loop instead. Don't try to update
2401 recog_data.operand after failed apply_change_group.
2403 2018-04-12 Tom de Vries <tom@codesourcery.com>
2406 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2407 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2408 array with flexible array member as array without given dimension.
2409 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2410 argument for undefined param to true.
2412 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2415 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2416 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2417 from PowerPC section.
2418 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2419 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2420 help text that is too long.
2421 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2422 help text that is too long.
2423 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2424 help text that is too long.
2426 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2428 * config/alpha/alpha.md (stack_probe_internal): Rename
2429 from "probe_stack". Update all callers.
2431 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2433 PR rtl-optimization/84566
2434 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2435 sched_macro_fuse_insns.
2437 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2440 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2441 (compute_block_dependences): ... from here.
2443 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2445 PR tree-optimization/85331
2446 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2447 from int to HOST_WIDE_INT.
2449 2018-04-11 Martin Jambor <mjambor@suse.cz>
2452 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2453 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2454 not the same as the source val.
2455 (cgraph_edge_brings_value_p): New parameter.
2456 (gather_edges_for_value): Pass destination value to
2457 cgraph_edge_brings_value_p.
2458 (perhaps_add_new_callers): Likewise.
2459 (get_info_about_necessary_edges): Likewise and exclude values brought
2460 only by self-recursive edges.
2461 (create_specialized_node): Redirect only clones of self-calling edges.
2462 (+self_recursive_pass_through_p): New function.
2463 (find_more_scalar_values_for_callers_subset): Use it.
2464 (find_aggregate_values_for_callers_subset): Likewise.
2465 (known_aggs_to_agg_replacement_list): Removed.
2466 (decide_whether_version_node): Re-calculate known constants for all
2467 remaining context clones.
2469 2018-04-11 Richard Biener <rguenther@suse.de>
2472 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2473 from early DWARF output.
2474 (dwarf2out_early_finish): Output line info unconditionally into
2475 early DWARF and add reference to it.
2477 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2480 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2481 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2482 other than V2DFmode using iptr mode attribute.
2483 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2485 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2487 PR rtl-optimization/84659
2488 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2490 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2493 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2495 (output_loc_list): Pass address of a dummy size variable even in the
2496 locview handling loop.
2497 (index_location_lists): Add comment on why skip_loc_list_entry can't
2500 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2503 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2506 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2509 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2510 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2511 and -mstring-compare-inline-limit.
2513 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2516 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2517 for stack clash protection in a register whenever we need it to be in
2520 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2522 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2523 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2525 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2528 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2530 (mlong-double-): Ditto.
2531 * config/rs6000/sysv4.opt (msdata=): Ditto.
2532 (mtls-size=): Ditto.
2534 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2536 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2537 erroneous entries for
2538 "vector int vec_ldl (int, long int *)", and
2539 "vector unsigned int vec_ldl (int, unsigned long int *)".
2540 Add comments and entries for
2541 "vector bool char vec_ldl (int, bool char *)",
2542 "vector bool short vec_ldl (int, bool short *)",
2543 "vector bool int vec_ldl (int, bool int *)",
2544 "vector bool long long vec_ldl (int, bool long long *)",
2545 "vector pixel vec_ldl (int, pixel *)",
2546 "vector long long vec_ldl (int, long long *)",
2547 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2548 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2549 type tree bool_long_long_type_node and correct definition of
2550 bool_V2DI_type_node to make reference to this new type tree.
2551 (rs6000_mangle_type): Replace erroneous reference to
2552 bool_long_type_node with bool_long_long_type_node.
2553 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2554 comments to emphasize sign distinctions for char and int types and
2555 replace RS6000_BTI_bool_long constant with
2556 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2557 use of RS6000_BTI_pixel.
2558 (bool_long_type_node): Remove this macro definition.
2559 (bool_long_long_type_node): New macro definition
2561 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2563 PR rtl-optimization/85300
2564 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2565 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2566 simplify_unary_operation fails.
2568 2018-04-10 Martin Liska <mliska@suse.cz>
2570 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2571 cgraph_edge and ipa_ref.
2573 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2577 * config/i386/sse.md
2578 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2579 computation of the VEC_MERGE selector from mask.
2580 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2581 Fix decoding of the VEC_MERGE selector into mask.
2583 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2585 PR tree-optimization/85286
2586 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2588 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2590 * final.c (final_1): Set insn_last_address as well as
2591 insn_current_address.
2593 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2596 * explow.c (emit_stack_probe): Call validize_mem on memory location
2597 before passing it to gen_probe_stack. Create address operand and
2598 legitimize it for the probe_stack_address case.
2600 2018-04-09 Jan Hubicka <jh@suse.cz>
2603 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2604 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2605 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2607 (free_lang_data): Rebuild type inheritance graph.
2609 2018-04-09 Martin Sebor <msebor@redhat.com>
2611 * invoke.texi (-finline-small-functions): Mention other optimization
2613 (-findirect-inlining, -fpartial-inlining): Same.
2614 (-finline-functions-called-once): Same.
2615 (-freorder-blocks-and-partition): Same.
2617 2018-04-09 Jan Hubicka <jh@suse.cz>
2620 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2621 jumps; choose last target that matches the criteria (i.e.
2622 no partition changes for non-crossing jumps).
2623 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2624 support for redirecting crossing jumps to non-crossing.
2626 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2628 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2629 also for naked functions.
2631 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2633 * config/arc/arc.md (add_shift): New pattern.
2634 (add_shift2): Likewise.
2635 (sub_shift): Likewise.
2636 (sub_shift_cmp0_noout): Likewise.
2637 (compare_si_ashiftsi): Likewise.
2638 (xbfu_cmp0_noout): New combine pattern.
2639 (xbfu_cmp0"): Likewise.
2640 (movsi_set_cc_insn): Place the predicable variant first.
2641 (commutative_binary_cmp0_noout): Remove clobber.
2642 (commutative_binary_cmp0): New pattern.
2643 (noncommutative_binary_cmp0): Likewise.
2644 (noncommutative_binary_cmp0_noout): Likewise.
2645 (noncommutative_binary_comparison_result_used): Removed.
2646 (rsub_cmp0): New pattern.
2647 (rsub_cmp0_noout): Likewise.
2648 (extzvsi): Changed, keep only meaningful variants.
2649 (SQH, SEZ): New iterators.
2650 (SQH_postfix): New mode attribute.
2651 (SEZ_prefix): New code attribute.
2652 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2653 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2654 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2656 (noncommutative_operator): Check the availability of barrel
2659 2018-04-09 Richard Biener <rguenther@suse.de>
2661 PR tree-optimization/85284
2662 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2663 Only use the niter constraining form of simple_iv when the exit
2666 2018-04-09 Tom de Vries <tom@codesourcery.com>
2669 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2670 (define_expand "*memory_barrier"): New define_expand.
2671 (define_insn "memory_barrier"): New insn.
2673 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2675 PR rtl-optimization/80463
2676 PR rtl-optimization/83972
2677 PR rtl-optimization/83480
2679 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2680 correct producer for the insn.
2681 (tidy_control_flow): Fixup seqnos in case of debug insns.
2683 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2685 PR rtl-optimization/83913
2687 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2688 different sched-times when merging exprs.
2690 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2692 PR rtl-optimization/83962
2694 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2695 tidy_fallthru_edge and tidy_control_flow.
2697 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2699 PR rtl-optimization/83530
2701 * sel-sched.c (force_next_insn): New global variable.
2702 (remove_insn_for_debug): When force_next_insn is true, also leave only
2703 next insn in the ready list.
2704 (sel_sched_region): When the region wasn't scheduled, make another pass
2705 over it with force_next_insn set to 1.
2707 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2709 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2711 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2712 for interrupt control.
2713 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2714 functions for interrupt control.
2715 * config/nds32/nds32-intrinsic.md: Likewise.
2716 * config/nds32/nds32_intrinsic.h: Likewise.
2717 * config/nds32/nds32.h (nds32_builtins): Likewise.
2719 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2721 * config/nds32/nds32.c (nds32_init_machine_status,
2722 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2723 strict_aligned_p field.
2724 (nds32_expand_to_rtl_hook): New function.
2725 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2726 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2728 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2729 Chung-Ju Wu <jasonwucj@gmail.com>
2731 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2732 * config/nds32/nds32-n7.md: New file.
2733 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2734 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2736 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2737 * config/nds32/nds32.md (pipeline_model): Add n7.
2738 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2739 * config/nds32/pipelines.md: Include n7 settings.
2741 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2742 Chung-Ju Wu <jasonwucj@gmail.com>
2744 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2745 * config/nds32/nds32-e8.md: New file.
2746 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2747 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2749 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2750 * config/nds32/nds32.md (pipeline_model): Add e8.
2751 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2752 * config/nds32/pipelines.md: Include e8 settings.
2754 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2755 Chung-Ju Wu <jasonwucj@gmail.com>
2757 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2758 * config/nds32/nds32-n8.md: New file.
2759 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2760 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2762 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2763 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2764 * config/nds32/nds32.md (pipeline_model): Add n8.
2765 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2766 * config/nds32/pipelines.md: Include n8 settings.
2768 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2769 Chung-Ju Wu <jasonwucj@gmail.com>
2771 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2772 * config/nds32/nds32-n9-2r1w.md: New file.
2773 * config/nds32/nds32-n9-3r2w.md: New file.
2774 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2775 nds32_register_ports): New or modify for cpu n9.
2776 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2778 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2779 * config/nds32/nds32-utils.c: New file.
2780 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2781 TARGET_MUL_SLOW): Define.
2782 * config/nds32/nds32.md (pipeline_model): New attribute.
2783 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2784 New options that support cpu n9.
2785 * config/nds32/pipelines.md: Include n9 settings.
2786 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2788 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2790 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2791 information if necessary.
2792 (output_cond_branch_compare_zero): Likewise.
2793 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2794 (nds32_target_alignment): Refine for alignment.
2795 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2796 (FUNCTION_BOUNDARY): Modify.
2797 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2799 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2801 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2803 * config/nds32/constants.md (unspec_volatile_element): Add values for
2804 TLB operation and data prefetch.
2805 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2806 functions for TLB operation and data prefetch.
2807 * config/nds32/nds32-intrinsic.md: Likewise.
2808 * config/nds32/nds32_intrinsic.h: Likewise.
2809 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2810 (nds32_print_operand): Likewise.
2811 * config/nds32/nds32.h (nds32_builtins): Likewise.
2813 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2814 Andrew Pinski <pinsika@gcc.gnu.org>
2817 * match.pd: Use constant_boolean_node of correct type instead of
2818 boolean_true_node or boolean_false_node for simplifying
2819 pointer comparisons to zero.
2821 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2823 PR tree-optimization/80021
2824 * tree.c (verify_type_variant): Make error call in verify_variant_match
2825 translatable and remove final full stop.
2827 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2829 * config/nds32/constants.md (unspec_volatile_element): Add
2830 UNSPEC_VOLATILE_EH_RETURN.
2831 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2832 nds32_output_stack_pop): Support dwarf exception handling process.
2833 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2834 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2835 exception handling process.
2836 (nds32_compute_stack_frame): Likewise.
2837 (nds32_return_addr_rtx): Likewise.
2838 (nds32_initial_elimination_offset): Likewise.
2839 (nds32_expand_prologue): Likewise.
2840 (nds32_expand_epilogue): Likewise.
2841 (nds32_dynamic_chain_address): New function.
2842 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2844 (DYNAMIC_CHAIN_ADDRESS): Define.
2845 (EH_RETURN_DATA_REGNO): Define.
2846 (EH_RETURN_STACKADJ_RTX): Define.
2847 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2848 patterns for dwarf exception handling.
2850 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2852 * config/nds32/nds32.h: Clean up obsolete macros.
2854 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2856 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2857 Add enum values for particular instructions.
2858 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2859 particular intrinsic functions.
2860 * config/nds32/nds32-intrinsic.md: Likewise.
2861 * config/nds32/nds32_intrinsic.h: Likewise.
2862 * config/nds32/nds32.h (nds32_builtins): Likewise.
2863 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2864 (btst, ave): New patterns for particular instructions.
2866 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2868 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2869 Add enum values for atomic load/store and memory sync.
2870 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2872 * config/nds32/nds32-intrinsic.md: Likewise.
2873 * config/nds32/nds32_intrinsic.h: Likewise.
2874 * config/nds32/nds32.h (nds32_builtins): Likewise.
2876 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2878 PR tree-optimization/85257
2879 * fold-const.c (native_encode_vector): If not all elts could fit
2880 and off is -1, return 0 rather than offset.
2881 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2882 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2883 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2884 adjust buffer in native_interpret_expr call.
2886 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2888 * config/nds32/constants.md (unspec_volatile_element): Add cache
2889 control enum values.
2890 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2891 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2892 * config/nds32/nds32.c (nds32_cctl_names): New.
2893 (nds32_print_operand): Handle cache control register names.
2894 * config/nds32/nds32.h (nds32_builtins): New enum values.
2895 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2897 * config/nds32/nds32.md (type): Add mmu.
2898 * config/nds32/pipelines.md (simple_insn): Add mmu.
2900 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2902 * config/nds32/nds32.md (type): Remove call.
2903 * config/nds32/pipelines.md (simple_insn): Likewise.
2905 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2907 * config/nds32/constants.md (unspec_volatile_element): Add
2908 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2909 UNSPEC_VOLATILE_FMFCFG.
2910 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2911 description for fmfcfg and fmfcsr.
2912 (bdesc_1arg): Add fmtcsr.
2913 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2914 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2915 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2916 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2917 unspec_fmfcfg): New patterns.
2918 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2919 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2920 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2921 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2922 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2923 __nds32__fmfcfg): Define.
2925 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2927 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2928 intrinsic register names.
2929 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2930 intrinsic register enum values and macros.
2932 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2934 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2935 for load/store addressing form.
2936 (nds32_print_operand_address): Likewise.
2938 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2941 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2942 based on LABEL_REF. Remove useless assertion.
2943 (pic_address_needs_scratch): Fix formatting.
2944 (sparc_legitimize_pic_address): Minor tweaks.
2945 (sparc_delegitimize_address): Adjust assertion accordingly.
2946 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2947 into symbolic_operand.
2948 (movsi_high_pic_label_ref): Likewise.
2949 (movsi_lo_sum_pic_label_ref): Likewise.
2950 (movdi_pic_label_ref): Likewise.
2951 (movdi_high_pic_label_ref): Likewise.
2952 (movdi_lo_sum_pic_label_ref): Likewise.
2954 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2956 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2957 custom LIB_SPEC setup.
2959 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2960 Kito Cheng <kito.cheng@gmail.com>
2962 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2963 * config/riscv/freebsd.h: New.
2965 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2967 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2968 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2971 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2972 Kito Cheng <kito.cheng@gmail.com>
2974 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2975 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2976 * config/nds32/nds32-protos.h (nds32_output_call,
2977 nds32_output_return): Declare.
2978 * config/nds32/nds32.md: Refine all the call and return patterns.
2980 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2983 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2984 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2986 PR rtl-optimization/84872
2987 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2988 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2991 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2993 * expr.c (copy_blkmode_to_reg): Revert 254862.
2994 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2996 2018-04-06 Richard Biener <rguenther@suse.de>
2999 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3000 after seeing a component reference with an adjacent field. Treat
3001 refs to arrays at struct end of external decls similar to
3002 refs to unconstrained commons.
3004 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3007 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3008 look through SAVE_EXPRs with non-side-effects argument. Adjust
3010 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3013 2018-04-06 Richard Biener <rguenther@suse.de>
3016 * alias.c (find_base_term): New wrapper around find_base_term
3017 unwinding CSELIB_VAL_PTR changes.
3018 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3021 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3023 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3025 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3026 constant definitions.
3027 ("nop"): lr 0,0 -> nopr r0
3028 ("nop_lr0", "nop_lr1"): New insn definitions.
3030 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3032 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3033 NDS32_V3PUSH_AVAILABLE_P macro.
3035 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3036 Chung-Ju Wu <jasonwucj@gmail.com>
3038 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3039 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3040 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3041 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3042 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3043 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3044 * config/nds32/constraints.md: New constraints and checking for hard
3045 float configuration.
3046 * config/nds32/iterators.md: New mode iterator and attribute for hard
3047 float configuration.
3048 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3050 * config/nds32/nds32-fpu.md: New file.
3051 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3052 deal with hard float code generation.
3053 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3055 (abi_type, float_reg_number): New enum type.
3056 * config/nds32/nds32-predicates.c: New predicates for hard float.
3057 * config/nds32/nds32-protos.h: Declare functions for hard float.
3058 * config/nds32/nds32.c: Implementation for hard float configuration.
3059 * config/nds32/nds32.h: Definitions for hard float configuration.
3060 * config/nds32/nds32.md: Include hard float machine description and
3061 modify patterns for hard float configuration.
3062 * config/nds32/nds32.opt: New options for hard float configuration.
3063 * config/nds32/predicates.md: New predicates for hard float
3066 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3068 * common/config/nds32/nds32-common.c
3069 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3071 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3074 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3075 CONSTRUCTOR_ELT (ctor, ...)->value.
3077 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3080 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3082 2018-04-05 Tom de Vries <tom@codesourcery.com>
3085 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3088 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3089 Kito Cheng <kito.cheng@gmail.com>
3091 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3092 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3093 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3094 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3096 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3097 Kito Cheng <kito.cheng@gmail.com>
3099 * config/nds32/constraints.md (Ufe): New memory constraint.
3100 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3101 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3102 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3104 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3105 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3107 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3109 * config/nds32/nds32.md: Use optimize_size in the condition for
3110 alu-shift instructions.
3112 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3114 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3116 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3118 * config/nds32/nds32.md (negsi2): Refine pattern.
3120 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3121 Chung-Ju Wu <jasonwucj@gmail.com>
3123 * config/nds32/iterators.md (shift_rotate): New code iterator.
3124 (shift): New code attribute.
3125 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3126 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3127 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3128 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3129 bit-wise operations.
3130 (andsi3, *andsi3): Ditto.
3131 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3132 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3133 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3134 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3135 nds32_ior_operand, nds32_xor_operand): New predicates.
3137 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3139 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3140 (addsi3, subsi3): ... this.
3142 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3144 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3146 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3148 * config/nds32/nds32.md: Adjust indention.
3150 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3152 * config/nds32/nds32.md (feature): New attribute.
3154 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3156 * config/nds32/nds32.md (subtype): New attribute.
3158 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3161 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3162 expansion to perform a bitwise AND of the argument followed by a
3163 boolean negation of the result.
3165 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3167 PR rtl-optimization/84878
3168 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3169 the basic block. Assert the use reference is not artificial and that
3170 it has an associated insn.
3172 2018-04-04 Michael Matz <matz@suse.de>
3174 * builtins.c (compute_objsize): Pass correct operand
3175 to array_at_struct_end_p.
3177 2018-04-04 Richard Biener <rguenther@suse.de>
3180 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3181 from contexts for DINFO_LEVEL_TERSE and below.
3183 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3185 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3186 resiter_operand condition.
3187 * config/nds32/nds32.md (*move<mode>): Ditto.
3189 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3190 Monk Chiang <sh.chiang04@gmail.com>
3192 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3194 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3196 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3198 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3199 Kito Cheng <kito.cheng@gmail.com>
3201 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3202 nds32_cond_code_str, output_cond_branch,
3203 output_cond_branch_compare_zero, nds32_expand_cbranch,
3204 nds32_expand_cstore, nds32_expand_movcc,
3205 nds32_output_cbranchsi4_equality_zero,
3206 nds32_output_cbranchsi4_equality_reg,
3207 nds32_output_cbranchsi4_equality_reg_or_const_int,
3208 nds32_output_cbranchsi4_greater_less_zero: New functions.
3209 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3210 nds32_expand_cstore, nds32_expand_movcc,
3211 nds32_output_cbranchsi4_equality_zero,
3212 nds32_output_cbranchsi4_equality_reg,
3213 nds32_output_cbranchsi4_equality_reg_or_const_int,
3214 nds32_output_cbranchsi4_greater_less_zero): Declare.
3215 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3216 nds32_rimm11s_operand): New predicates.
3217 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3218 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3221 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3223 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3224 * config/nds32/nds32.md: Ditto.
3225 * config/nds32/pipelines.md: Ditto.
3227 2018-04-04 Richard Biener <rguenther@suse.de>
3229 PR tree-optimization/85168
3230 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3231 propagating abnormals.
3233 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3235 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3237 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3238 Kito Cheng <kito.cheng@gmail.com>
3240 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3241 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3242 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3243 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3244 * config/nds32/nds32.md (sibcall_internal): New.
3245 (sibcall_register): Remove.
3246 (sibcall_immediate): Remove.
3247 (sibcall_value_internal): New.
3248 (sibcall_value_register): Remove.
3249 (sibcall_value_immediate): Remove.
3250 * config/nds32/predicates.md (nds32_general_register_operand): New.
3251 (nds32_call_address_operand): New.
3253 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3255 PR rtl-optimization/85167
3256 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3257 bb_defs if *split_p, instead preinitialize it to NULL.
3259 PR tree-optimization/85156
3260 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3261 evaluating the argument multiple times.
3263 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3265 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3267 (_mm_cvtpd_ps): Likewise.
3268 (_mm_cvttpd_epi32): Likewise.
3269 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3270 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3271 vector, pixel, and bool following altivec.h include.
3273 2018-04-03 Martin Sebor <msebor@redhat.com>
3275 * doc/extend.texi (Common Function Attributes): Clarify.
3276 (const attribute): Likewise.
3277 (pure attribute): Likewise.
3279 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3282 * config/i386/i386.c (ix86_expand_vector_set): Use
3283 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3285 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3287 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3288 instructions when changing rounding bits to preserve precision bits
3289 in the x87 control word.
3291 2018-04-03 Martin Liska <mliska@suse.cz>
3293 PR tree-optimization/82491
3294 * rtl.h (strip_offset_and_add): Replace += suboffset with
3295 poly_uint64 () + suboffset.
3297 2018-03-29 Martin Liska <mliska@suse.cz>
3298 Martin Jambor <mjambor@suse.cz>
3301 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3302 param_type is not an integral or pointer type.
3304 2018-04-03 Richard Biener <rguenther@suse.de>
3306 * sese.h (recompute_all_dominators): Remove.
3308 2018-04-02 Martin Sebor <msebor@redhat.com>
3310 * doc/invoke.texi (-Wrestrict): Fix typos.
3312 2018-04-02 Jim Wilson <jimw@sifive.com>
3314 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3315 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3316 (<optab>di3, <optab>si3_extend): Likewise.
3317 (<optab>si3_mask, <optab>si3_mask_1): New.
3318 (<optab>di3_mask, <optab>di3_mask_1): New.
3319 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3320 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3321 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3323 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3325 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3328 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3330 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3331 (nds32_canonicalize_comparison): New function.
3333 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3334 Kito Cheng <kito.cheng@gmail.com>
3335 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3337 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3338 * config/nds32/constants.md (unspec_volatile_element): Add
3339 UNSPEC_VOLATILE_RELAX_GROUP.
3340 * config/nds32/nds32-relax-opt.c: New file.
3341 * config/nds32/nds32-predicates.c
3342 (nds32_symbol_load_store_p): New function.
3343 * config/nds32/nds32-protos.h
3344 (nds32_symbol_load_store_p): Declare function.
3345 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3346 * config/nds32/nds32.c
3347 (nds32_register_pass): New function to register pass.
3348 (nds32_register_passes): New function to register passes.
3349 * config/nds32/nds32.md (relax_group): New pattern.
3350 * config/nds32/nds32.opt (mrelax-hint): New option.
3351 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3353 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3355 * config/nds32/t-nds32: Modify files dependency.
3357 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3359 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3360 (PROFILE_HOOK): Define its implementation.
3362 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3364 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3365 type and 32-bit size.
3367 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3370 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3371 (V_128_256): New mode iterator.
3372 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3373 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3374 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3376 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3377 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3379 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3382 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3383 NaN inputs correctly.
3385 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3388 * config/rs6000/vsx.md (??r): New mode attribute.
3389 (*vsx_mov<mode>_64bit): Use it.
3390 (*vsx_mov<mode>_32bit): Likewise.
3392 2018-03-30 Martin Sebor <msebor@redhat.com>
3394 PR tree-optimization/84818
3395 * builtins.c (check_access): Use warning_n.
3397 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3400 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3402 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3405 2018-03-30 Julia Koval <julia.koval@intel.com>
3408 * x86-tune.def (movx, partial_reg_dependency): Enable for
3411 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3414 * lra-constraints.c (process_alt_operands): Move setting
3415 this_alternative_matches below.
3417 2018-03-29 Martin Liska <mliska@suse.cz>
3420 * doc/invoke.texi: Document how LTO works with debug info.
3421 Describe auto-load support of binutils. Mention 'x86-64'
3422 as valid option value of -march option.
3424 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3426 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3429 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3430 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3431 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3434 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3437 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3438 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3439 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3440 for RS6000_BTM_POWERPC64.
3441 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3442 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3443 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3448 2018-03-28 Carl Love <cel@us.ibm.com>
3451 2017-09-27 Carl Love <cel@us.ibm.com>
3453 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3454 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3455 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3458 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3460 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3461 instead of __vector bool.
3462 (_mm_max_pu8): Likewise.
3463 (_mm_min_pi16): Likewise.
3465 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3468 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3469 (DIVWEUO): Likewise.
3471 (DIVDEUO): Likewise.
3472 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3473 DIVWEUO and DIVDEUO.
3474 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3475 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3476 (div_extend): Likewise.
3477 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3479 (__builtin_divweuo): Likewise.
3480 (__builtin_divdeo): Likewise.
3481 (__builtin_divdeuo): Likewise.
3483 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3486 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3487 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3489 PR tree-optimization/82004
3490 * gimple-match-head.c (optimize_pow_to_exp): New function.
3491 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3492 Don't fold to exp if optimize_pow_to_exp is false.
3494 2018-03-28 Martin Liska <mliska@suse.cz>
3497 * calls.c (initialize_argument_information): Fix trailing space.
3498 * common.opt: Fix typo and provide better explanation for
3499 -fsanitize-coverage option.
3500 * config/i386/i386.opt: Fix typo.
3502 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3503 Martin Liska <mliska@suse.cz>
3506 * gimplify.c (asan_poison_variable): Don't do the check for
3507 gimplify_omp_ctxp here.
3508 (gimplify_decl_expr): Do it here.
3509 (gimplify_target_expr): Likewise.
3511 2018-03-28 Martin Liska <mliska@suse.cz>
3514 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3515 chkp_type_bounds_count if MPX is not enabled.
3517 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3519 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3521 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3524 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3525 function to create the function decl for complex long double
3526 multiply and divide for -mabi=ieeelongdouble.
3527 (init_float128_ieee): Call it.
3529 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3532 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3533 -fcf-protection=branch -mibt.
3534 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3536 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3539 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3541 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3544 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3545 extern array declarations.
3547 2018-03-27 Richard Biener <rguenther@suse.de>
3550 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3551 explicit single_use checks.
3553 2018-03-27 Richard Biener <rguenther@suse.de>
3555 PR tree-optimization/85082
3556 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3559 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3561 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3562 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3563 Turn on fasynchronous-unwind-tables and funwind-tables.
3565 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3568 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3569 (*bmi_blsr_<mode>_ccz): Ditto.
3571 2018-03-26 Tom de Vries <tom@codesourcery.com>
3573 PR tree-optimization/85063
3574 * omp-general.c (offloading_function_p): New function. Factor out
3576 * omp-offload.c (pass_omp_target_link::gate): ... here.
3577 * omp-general.h (offloading_function_p): Declare.
3578 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3579 with attribute omp declare target for offloading functions.
3581 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3583 PR tree-optimization/84005
3584 * tree-data-ref.h (get_base_for_alignment): Declare.
3585 * tree-data-ref.c (get_base_for_alignment_1): New function.
3586 (get_base_for_alignment): Likewise.
3587 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3588 get_base_for_alignment to find a suitable base object, instead
3589 of always using drb->base_address.
3591 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3594 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3595 known size by default.
3597 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3600 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3601 and non BLKmode operands.
3603 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3606 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3607 Clean up attributes.
3609 2018-03-23 Richard Biener <rguenther@suse.de>
3612 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3613 we are going to emit early debug for LTO.
3615 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3618 * function.c (match_asm_constraints_1): Don't optimize if input
3619 doesn't satisfy general_operand predicate for output's mode.
3622 * alias.c (write_dependence_p): Don't require for x_canonicalized
3623 non-VOIDmode if x has VOIDmode.
3626 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3627 just don't try to optimize it rather than assert it never happens.
3629 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3631 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3632 macro expansions for definition of ST_INTERNAL_<mode> and
3633 LD_INTERNAL_<mode> builtins.
3634 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3636 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3638 (altivec_expand_st_builtin): Likewise.
3639 (altivec_expand_builtin): Remove calls to deleted functions.
3640 (rs6000_address_for_altivec): Delete this function.
3641 * config/rs6000/vector.md: Remove expands for
3642 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3644 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3647 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3648 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3649 re-computing once computed.
3650 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3651 (arm_init_machine_status): Initialize
3652 machine->static_chain_stack_bytes.
3654 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3657 * doc/extend.texi: Add four new prototypes for vec_ld.
3658 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3659 definitions for more logical presentation.
3660 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3661 entries for V1TI variants of __builtin_altivec_ld builtin.
3662 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3663 handling of V1TI variant of LVX icode pattern.
3664 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3665 (rs6000_gimple_fold_builtin): Likewise.
3666 (altivec_init_builtins): Add code to define
3667 __builtin_altivec_lvx_v1ti function.
3669 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3672 * function.c (match_asm_constraints_1): Don't do the optimization
3673 if input isn't a REG, SUBREG, MEM or constant.
3675 2018-03-22 Tom de Vries <tom@codesourcery.com>
3677 PR tree-optimization/84956
3678 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3679 bb_has_abnormal_pred.
3681 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3684 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3685 DECL_INITIAL (decl) to decl at the end.
3686 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3689 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3691 * doc/extend.texi (__builtin_tgmath): Document when complex
3692 integer types are treated as _Complex _Float64.
3694 2018-03-21 Tom de Vries <tom@codesourcery.com>
3696 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3698 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3700 PR tree-optimization/84960
3701 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3702 if it is ENTRY block, move them into single succ of ENTRY in that case.
3704 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3706 PR tree-optimization/84811
3707 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3708 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3709 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3710 (known_subrange_p): Update accordingly. Cast each value involved
3711 in the size comparison, rather than casting the result of the
3714 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3716 PR tree-optimization/84982
3717 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3718 by flipping the least significant bit rather than all bits from
3719 bitpos to bitpos + bitsize - 1.
3721 2018-03-21 Nathan Sidwell <nathan@acm.org>
3723 * doc/extend.texi (Deprecated Features): Remove mention of
3724 long-deleted deprecations.
3726 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3729 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3730 * configure: Regenerate.
3732 2018-03-21 Tom de Vries <tom@codesourcery.com>
3734 PR tree-optimization/83126
3735 * tree-parloops.c (num_phis): New function.
3736 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3738 2018-03-21 Nathan Sidwell <nathan@acm.org>
3740 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3741 mention anon-struct/union members and trailing attributes.
3743 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3745 PR tree-optimization/84969
3746 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3747 builtin memset partitions if they set different rhs values.
3749 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3751 PR rtl-optimization/84989
3752 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3753 VEC_DUPLICATE with scalar result mode.
3755 2018-03-21 Martin Liska <mliska@suse.cz>
3758 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3759 not intended return statement.
3761 2018-03-21 Martin Liska <mliska@suse.cz>
3764 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3765 (chkp_find_bound_slots_1): Limit number of iterations.
3767 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3770 * Minor grammar fixes for x86 options.
3772 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3775 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3776 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3778 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3781 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3782 (altivec_lvx_<mode>_1op): Likewise.
3783 (altivec_stvx_<mode>_2op): Likewise.
3784 (altivec_stvx_<mode>_1op): Likewise.
3785 (altivec_lvx_<VM2:mode>): New define_expand.
3786 (altivec_stvx_<VM2:mode>): Likewise.
3787 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3788 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3789 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3790 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3791 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3792 (rs6000_gen_lvx): Likewise.
3793 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3794 (altivec_expand_stv_builtin): Likewise.
3795 (altivec_expand_builtin): Likewise.
3796 * config/rs6000/vector.md: Likewise.
3798 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3801 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3804 2018-03-20 Richard Biener <rguenther@suse.de>
3807 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3808 sign-conversions as zero, fall back to standard scalar_stmt
3811 2018-03-20 Martin Liska <mliska@suse.cz>
3814 * predict.c (rebuild_frequencies): Handle case when we have
3815 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3817 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3820 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3821 flag_section_anchors.
3822 * varasm.c (use_blocks_for_decl_p): Remove hack for
3823 dw2_force_const_mem.
3826 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3828 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3829 be created, use lowpart_subreg of operands[0] rather than operands[0]
3831 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3832 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3833 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3834 and n constraint instead of aarch64_shift_imm_di and Usd.
3835 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3836 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3838 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3841 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3842 to favor GPR over NEON registers.
3843 (<shift>di3_neon): Likewise.
3845 2018-03-20 Tom de Vries <tom@codesourcery.com>
3848 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3849 (nvptx_process_pars): Emit bar.sync asap and alap.
3851 2018-03-20 Tom de Vries <tom@codesourcery.com>
3854 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3855 seen_label if seen_label is already set.
3857 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3860 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3861 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3862 Use 1U instead of 1. Formatting fixes.
3865 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3866 instead of TREE_TYPE (s1) for the return value.
3868 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3870 PR tree-optimization/84946
3871 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3872 bitsize + bitsize in poly_uint64 rather than poly_int64.
3875 * dwarf2asm.c: Include fold-const.c.
3876 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3877 of decl rather than decl itself.
3879 PR rtl-optimization/84643
3880 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3882 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3885 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3886 calling assemble_variable.
3888 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3891 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3892 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3894 2018-03-19 Jim Wilson <jimw@sifive.com>
3897 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3898 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3899 (riscv_first_stack_step): Likewise.
3900 (riscv_option_override): Use STACK_BOUNDARY instead of
3902 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3904 (BIGGEST_ALIGNMENT): Set to 128.
3905 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3906 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3909 2018-03-19 Richard Biener <rguenther@suse.de>
3911 PR tree-optimization/84933
3912 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3913 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3915 2018-03-19 Richard Biener <rguenther@suse.de>
3917 PR tree-optimization/84859
3918 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3919 (cond_if_else_store_replacement): Perform sinking operation on
3920 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3921 Generalize what a BB with a single eligible store is.
3923 2018-03-19 Richard Biener <rguenther@suse.de>
3925 PR tree-optimization/84929
3926 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3927 chrec_is_positive against non-chrec arg.
3929 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3932 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3934 2018-03-18 Martin Liska <mliska@suse.cz>
3936 PR rtl-optimization/84635
3937 * regrename.c (build_def_use): Use matches_mode only when
3940 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3942 PR tree-optimization/84913
3943 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3944 vectorize chains of COND_EXPRs.
3946 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3948 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3950 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3952 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3954 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3956 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3958 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3959 Kito Cheng <kito.cheng@gmail.com>
3961 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3962 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3963 (nds32_adjust_reg_alloc_order): New function.
3964 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3966 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3968 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3969 nds32_print_operand, nds32_print_operand_address): Use
3970 HOST_WIDE_INT_PRINT_DEC instead.
3972 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3974 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3976 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3979 * config/i386/i386.c (initial_ix86_tune_features,
3980 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3982 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3983 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3984 rather than 1u << ix86_tune. Formatting fix.
3985 (ix86_option_override_internal): Change ix86_arch_mask from
3986 unsigned int to unsigned HOST_WIDE_INT, initialize to
3987 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3988 (ix86_function_specific_restore): Likewise.
3990 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3993 * postreload.c (reload_combine_recognize_pattern): Perform
3994 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3995 truncate_int_for_mode the result for the destination's mode.
3998 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3999 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4001 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4004 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4007 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4008 regno_allocno_class_array and sorted_pseudos.
4009 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4010 insns where regno is used.
4012 2018-03-16 Martin Liska <mliska@suse.cz>
4015 * multiple_target.c (create_dispatcher_calls): Redirect
4016 reference in the symbol table.
4018 2018-03-16 Martin Liska <mliska@suse.cz>
4021 * multiple_target.c (create_dispatcher_calls): Redirect also
4024 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4028 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4029 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4030 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4032 2018-03-16 Julia Koval <julia.koval@intel.com>
4034 * doc/invoke.texi (Skylake Server): Add CLWB.
4035 Cannonlake): Remove CLWB.
4037 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4039 PR tree-optimization/84841
4040 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4042 (FLOAT_ONE_CONST_TYPE): Define.
4043 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4044 (sort_by_operand_rank): Put entries with higher constant_type last
4045 rather than first to match comments.
4047 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4049 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4052 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4055 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4056 check and use address of target always.
4058 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4061 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4062 (indirect_thunk_bnd_needed): Likewise.
4063 (indirect_thunks_used): Likewise.
4064 (indirect_thunks_bnd_used): Likewise.
4065 (indirect_return_needed): New.
4066 (indirect_return_bnd_needed): Likewise.
4067 (output_indirect_thunk_function): Add a bool argument for
4069 (output_indirect_thunk_function): Don't generate alias for
4070 function return thunk.
4071 (ix86_code_end): Call output_indirect_thunk_function to generate
4072 function return thunks.
4073 (ix86_output_function_return): Set indirect_return_bnd_needed
4074 and indirect_return_needed instead of indirect_thunk_bnd_needed
4075 and indirect_thunk_needed.
4077 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4079 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4080 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4081 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4083 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4084 Paul Hua <paul.hua.gm@gmail.com>
4087 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4089 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4091 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4092 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4095 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4098 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4099 instead of GET_MODE_SIZE when comparing Units.
4101 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4104 * varasm.c (hash_section): Return an unchangeble hash value
4105 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4106 Return !aarch64_can_use_per_function_literal_pools_p ().
4108 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4111 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4112 rather than address of cmode as last argument to prepare_cmp_insn.
4114 2018-03-15 Julia Koval <julia.koval@intel.com>
4116 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4117 F_AVX512VNNI, F_AVX512BITALG): New.
4119 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4122 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4123 insn for floating-point loads and stores.
4125 2018-03-14 Carl Love <cel@us.ibm.com>
4127 * config/rs6000/rs6000-c.c: Add macro definitions for
4128 ALTIVEC_BUILTIN_VEC_PERMXOR.
4129 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4130 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4131 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4132 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4134 * config/doc/extend.texi: Add prototypes for vec_permxor.
4136 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4139 * diagnostic-show-locus.c (class layout_point): Convert m_line
4140 from int to linenum_type.
4141 (line_span::comparator): Use linenum "compare" function when
4142 comparing line numbers.
4143 (test_line_span): New function.
4144 (layout_range::contains_point): Convert param "row" from int to
4146 (layout_range::intersects_line_p): Likewise.
4147 (layout::will_show_line_p): Likewise.
4148 (layout::print_source_line): Likewise.
4149 (layout::should_print_annotation_line_p): Likewise.
4150 (layout::print_annotation_line): Likewise.
4151 (layout::print_leading_fixits): Likewise.
4152 (layout::annotation_line_showed_range_p): Likewise.
4153 (struct line_corrections): Likewise for field m_row.
4154 (line_corrections::line_corrections): Likewise for param "row".
4155 (layout::print_trailing_fixits): Likewise.
4156 (layout::get_state_at_point): Likewise.
4157 (layout::get_x_bound_for_row): Likewise.
4158 (layout::print_line): Likewise.
4159 (diagnostic_show_locus): Likewise for locals "last_line" and
4161 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4162 * input.c (selftest::test_linenum_comparisons): New function.
4163 (selftest::input_c_tests): Call it.
4164 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4165 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4166 * selftest.h (ASSERT_GT): New macro.
4167 (ASSERT_GT_AT): New macro.
4168 (ASSERT_LT): New macro.
4169 (ASSERT_LT_AT): New macro.
4171 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4173 PR rtl-optimization/84780
4174 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4176 2018-03-14 Martin Liska <mliska@suse.cz>
4178 * tree.c (record_node_allocation_statistics): Use
4179 get_stats_node_kind.
4180 (get_stats_node_kind): New function extracted from
4181 record_node_allocation_statistics.
4182 (free_node): Use get_stats_node_kind.
4184 2018-03-14 Richard Biener <rguenther@suse.de>
4186 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4187 that the value-set of ANTIC_IN doesn't grow.
4190 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4192 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4193 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4195 2018-03-14 Julia Koval <julia.koval@intel.com>
4197 * config.gcc (icelake-client, icelake-server): New.
4199 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4200 (initial_ix86_arch_features): Ditto.
4201 (PTA_SKYLAKE): Add SGX.
4202 (PTA_ICELAKE): Remove.
4203 (PTA_ICELAKE_CLIENT): New.
4204 (PTA_ICELAKE_SERVER): New.
4205 (ix86_option_override_internal): Split up icelake on icelake client and
4207 (get_builtin_code_for_version): Ditto.
4208 (fold_builtin_cpu): Ditto.
4209 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4210 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4211 * config/i386/i386.h (processor_type): Ditto.
4212 * doc/invoke.texi: Ditto.
4214 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4217 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4218 INTEGER_CST offset, add it together with bitpos / 8 and
4219 sign extend based on POINTER_SIZE.
4223 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4226 * config/i386/constraints.md (Yc): New register constraint.
4227 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4228 Use Yc constraint for alternative 2 of operand 0. Remove
4229 preferred_for_speed attribute.
4231 2018-03-14 Richard Biener <rguenther@suse.de>
4233 PR tree-optimization/84830
4234 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4235 with the old one to avoid oscillations.
4237 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4240 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4242 (assign_by_spills): Return a flag of reload assignment failure.
4243 Do not process the reload assignment failures. Do not spill other
4244 reload pseudos if they has the same reg class. Update n if
4246 (lra_assign): Add a return arg. Set up from the result of
4247 assign_by_spills call.
4248 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4249 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4250 usage_insns if it is not NULL.
4251 (spill_hard_reg_in_range): New function.
4252 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4253 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4254 function prototypes.
4255 (lra_assign): Change prototype.
4256 * lra.c (lra): Add code to deal with fails by splitting hard reg
4259 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4261 * config/riscv/riscv.opt (mrelax): New option.
4262 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4263 "norelax" when riscv_mrelax is disabled.
4264 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4266 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4269 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4270 reassociation for int modes.
4272 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4274 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4275 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4277 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4278 * config/aarch64/aarch64-sve.md
4279 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4280 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4281 (*extend<mode><Vwide>2): Rename to...
4282 (aarch64_sve_extend<mode><Vwide>2): ...this.
4283 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4284 renaming the old pattern to...
4285 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4287 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4288 define_expand, renaming the old pattern to...
4289 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4290 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4291 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4292 account when deciding which SVE instruction the optab should use.
4293 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4295 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4297 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4298 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4299 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4300 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4301 (tlsdesc_small_<mode>): Turn a define_expand and use
4302 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4303 (tlsdesc_small_advsimd_<mode>): ...this.
4304 (tlsdesc_small_sve_<mode>): New pattern.
4306 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4308 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4309 (UNSPEC_UMUL_HIGHPART): New constants.
4310 (MUL_HIGHPART): New int iteraor.
4311 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4312 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4314 (*<su>mul<mode>3_highpart): New define_insn.
4316 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4319 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4322 2018-03-13 Martin Liska <mliska@suse.cz>
4325 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4327 (sem_item_optimizer::~sem_item_optimizer): Release it.
4328 (sem_item_optimizer::merge_classes): Register variable aliases.
4329 (sem_item_optimizer::fixup_pt_set): New function.
4330 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4331 * ipa-icf.h: Declare new variables and functions.
4333 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4336 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4337 integer_pow2p@2 and test integer_pow2p in condition.
4338 (A < 0 ? C : 0): Similarly for @1.
4341 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4342 characters starting at p contain '\0' character, don't look beyond
4346 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4347 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4350 * reg-stack.c (change_stack): Change update_end var from int to
4351 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4352 also call set_block_for_insn on the newly added insns and rescan.
4355 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4356 on the last operand.
4359 * tree.c (stabilize_reference_1): Return save_expr (e) for
4360 STATEMENT_LIST even if it doesn't have side-effects.
4362 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4364 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4366 2018-03-12 Renlin Li <renlin.li@arm.com>
4368 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4369 aarch64_output_scalar_simd_mov_immediate.
4371 2018-03-12 Martin Sebor <msebor@redhat.com>
4373 PR tree-optimization/83456
4374 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4375 for perfectly overlapping calls to memcpy.
4376 (gimple_fold_builtin_memory_chk): Same.
4377 (gimple_fold_builtin_strcpy): Handle no-warning.
4378 (gimple_fold_builtin_stxcpy_chk): Same.
4379 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4381 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4383 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4384 parameter. Use it for SFmode.
4385 (rs6000_function_arg_advance_1): Adjust.
4386 (rs6000_function_arg): Adjust.
4387 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4389 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4391 PR rtl-optimization/84169
4392 PR rtl-optimization/84780
4393 * combine.c (can_combine_p): Check for a 2-insn combination whether
4394 the destination register is used between the two insns, too.
4396 2018-03-12 Richard Biener <rguenther@suse.de>
4398 PR tree-optimization/84803
4399 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4400 for refs DR analysis didn't process.
4402 2018-03-12 Richard Biener <rguenther@suse.de>
4404 PR tree-optimization/84777
4405 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4406 force-vectorize loops ignore whether we are optimizing for size.
4408 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4410 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4411 (TARGET_MD_ASM_ADJUST): Define.
4413 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4414 Kito Cheng <kito.cheng@gmail.com>
4415 Chung-Ju Wu <jasonwucj@gmail.com>
4417 * config/nds32/nds32.c (nds32_compute_stack_frame,
4418 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4419 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4420 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4421 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4422 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4423 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4424 * config/nds32/nds32.md (prologue, epilogue): Use macro
4425 NDS32_V3PUSH_AVAILABLE_P to do checking.
4427 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4430 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4431 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4432 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4433 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4434 addition of most attributes on !orig_type_die or the attribute not
4435 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4437 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4438 Chung-Ju Wu <jasonwucj@gmail.com>
4440 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4442 * config/nds32/nds32.opt (mvh): New option.
4444 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4445 Chung-Ju Wu <jasonwucj@gmail.com>
4447 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4449 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4450 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4453 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4454 Chung-Ju Wu <jasonwucj@gmail.com>
4456 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4458 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4459 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4461 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4462 Kito Cheng <kito.cheng@gmail.com>
4463 Chung-Ju Wu <jasonwucj@gmail.com>
4465 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4466 UNSPEC_FFMISM and UNSPEC_FLMISM.
4467 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4468 for ffb, ffmism and flmism.
4469 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4470 (unspec_ffmism): Ditto.
4471 (unspec_flmism): Ditto.
4472 (nds32_expand_builtin_impl): Check if string extension is available.
4473 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4474 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4476 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4479 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4482 * lra-assigns.c (assign_by_spills): Return a flag of reload
4483 assignment failure. Do not process the reload assignment
4484 failures. Do not spill other reload pseudos if they has the same
4486 (lra_assign): Add a return arg. Set up from the result of
4487 assign_by_spills call.
4488 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4489 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4490 usage_insns if it is not NULL.
4491 (spill_hard_reg_in_range): New function.
4492 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4493 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4494 function prototypes.
4495 (lra_assign): Change prototype.
4496 * lra.c (lra): Add code to deal with fails by splitting hard reg
4499 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4502 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4504 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4507 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4508 (dw_val_node): Add val_symbolic_view.
4509 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4510 (symview_upper_bound): New.
4511 (new_line_info_table): Initialize symviews_since_reset.
4512 (dwarf2out_source_line): Count symviews_since_reset and set
4513 symview_upper_bound.
4514 (dw_val_equal_p): Handle symview.
4515 (add_AT_symview): New.
4516 (print_dw_val): Handle symview.
4517 (attr_checksum, attr_checksum_ordered): Likewise.
4518 (same_dw_val_p, size_of_die): Likewise.
4519 (value_format, output_die): Likewise.
4520 (add_high_low_attributes): Use add_AT_symview for entry_view.
4521 (dwarf2out_finish): Reset symview_upper_bound, clear
4524 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4527 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4528 Add strict argument and use it.
4529 (rs6000_split_multireg_move): Update for new strict argument.
4530 (mem_operand_gpr): Disallow all non-offsettable addresses.
4531 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4533 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4536 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4537 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4538 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4541 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4542 decl, use remap_type if we want to use the type.
4544 2018-03-09 Martin Sebor <msebor@redhat.com>
4546 PR tree-optimization/84526
4547 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4549 (builtin_access::generic_overlap): Be prepared to handle non-array
4552 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4554 PR rtl-optimization/84682
4555 * lra-constraints.c (process_address_1): Check is_address flag
4556 for address constraints.
4557 (process_alt_operands): Likewise.
4558 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4559 preprocess_constraints.
4560 * recog.h (preprocess_constraints): Add oploc parameter.
4562 * recog.c (preprocess_constraints): Test address_operand for
4563 CT_ADDRESS constraints.
4565 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4568 * lra-assigns.c (assign_by_spills): Return a flag of reload
4569 assignment failure. Do not process the reload assignment
4570 failures. Do not spill other reload pseudos if they has the same
4572 (lra_assign): Add a return arg. Set up from the result of
4573 assign_by_spills call.
4574 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4575 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4576 usage_insns if it is not NULL.
4577 (spill_hard_reg_in_range): New function.
4578 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4579 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4580 function prototypes.
4581 (lra_assign): Change prototype.
4582 * lra.c (lra): Add code to deal with fails by splitting hard reg
4585 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4588 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4589 Accept complain bool parameter. Only emit errors if it is true.
4590 (arm_parse_cpu_option_name): Likewise.
4591 (arm_target_thumb_only): Adjust callers of the above.
4592 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4593 prototype to take a default true bool parameter.
4594 (arm_parse_arch_option_name): Likewise.
4596 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4597 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4601 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4602 * configure: Regenerate.
4603 * configure.ac ("linker --version-script option"): New.
4604 ("linker soname option"): New.
4606 2018-03-09 Richard Biener <rguenther@suse.de>
4608 PR tree-optimization/84775
4609 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4610 immediate uses of predicate stmts and mark them modified.
4613 PR tree-optimization/84178
4614 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4616 (version_loop_for_if_conversion): Delay update_ssa call.
4617 (tree_if_conversion): Delay update_ssa until after predicate
4620 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4623 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4624 when the function accesses prior frames.
4626 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4629 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4630 gen_llsym, otherwise call maybe_gen_llsym.
4633 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4634 has ',' character inside of it.
4636 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4639 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4640 as clobbering CC_REGNUM.
4642 2018-03-08 Richard Biener <rguenther@suse.de>
4645 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4646 (follow_copies_to_constant): Do not follow SSA names registered
4649 2018-03-08 Richard Biener <rguenther@suse.de>
4651 PR tree-optimization/84178
4652 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4654 (version_loop_for_if_conversion): Delay update_ssa call.
4655 (tree_if_conversion): Delay update_ssa until after predicate
4658 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4660 PR tree-optimization/84178
4661 * tree-if-conv.c (release_bb_predicate): Remove the
4662 the assertion that the stmts have NULL use_ops.
4663 Discard the statements, asserting that they haven't
4664 yet been added to a BB.
4666 2018-03-08 Richard Biener <rguenther@suse.de>
4668 PR tree-optimization/84746
4669 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4670 (phi_translate): Pass in destination ANTIC_OUT set.
4671 (phi_translate_1): Likewise. For a simplified result lookup
4672 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4673 (phi_translate_set): Adjust.
4674 (do_pre_regular_insertion): Likewise.
4675 (do_pre_partial_partial_insertion): Likewise.
4677 2018-03-08 Martin Liska <mliska@suse.cz>
4679 PR gcov-profile/84735
4680 * doc/gcov.texi: Document usage of profile files.
4681 * gcov-io.h: Document changes in the format.
4683 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4687 * dwarf2out.c (struct dw_line_info_table): Update comments for
4689 (FORCE_RESET_NEXT_VIEW): New.
4690 (FORCE_RESETTING_VIEW_P): New.
4691 (RESETTING_VIEW_P): Check for -1 too.
4692 (ZERO_VIEW_P): Likewise.
4693 (new_line_info_table): Force-reset next view.
4694 (dwarf2out_begin_function): Likewise.
4695 (dwarf2out_source_line): Simplify zero_view_p initialization.
4696 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4697 view directly. Omit view when omitting .loc at line 0.
4699 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4701 PR tree-optimization/84740
4702 * tree-switch-conversion.c (process_switch): Call build_constructors
4703 only if info.phi_count is non-zero.
4705 PR tree-optimization/84739
4706 * tree-tailcall.c (find_tail_calls): Check call arguments against
4707 DECL_ARGUMENTS (current_function_decl) rather than
4708 DECL_ARGUMENTS (func) when checking for tail recursion.
4710 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4712 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4713 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4714 Volker Reichelt's entry and add entries for people that perform
4715 GCC fuzzy testing and report numerous bugs.
4717 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4720 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4721 readonly data in sdata, if that is disabled.
4722 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4723 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4724 -mreadonly-in-sdata option.
4726 2018-03-07 Martin Sebor <msebor@redhat.com>
4728 PR tree-optimization/84468
4729 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4730 basic block when looking for nul assignment.
4732 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4735 * except.h (output_function_exception_table): Adjust prototype.
4736 * except.c (output_function_exception_table): Remove FNNAME parameter
4737 and add SECTION parameter. Ouput one part of the table at a time.
4738 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4739 the first part of the exception table and emit unwind directives.
4740 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4741 (i386_pe_seh_cold_init): Likewise.
4742 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4743 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4744 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4745 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4746 * config/i386/winnt.c: Include except.h.
4747 (struct seh_frame_state): Add reg_offset, after_prologue and
4748 in_cold_section fields.
4749 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4750 (i386_pe_seh_cold_init): New function.
4751 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4752 to seh->in_cold_section.
4753 (seh_emit_push): Record the offset of the push.
4754 (seh_emit_save): Record the offet of the save.
4755 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4756 Test seh->after_prologue to disregard the epilogue.
4757 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4758 (i386_pe_end_cold_function): New function.
4760 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4763 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4764 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4767 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4769 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4770 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4773 * multiple_target.c: Include tree-inline.h and intl.h.
4774 (expand_target_clones): Diagnose and fail if node->definition and
4775 !tree_versionable_function_p (node->decl).
4777 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4779 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4781 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4782 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4783 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4785 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4788 * combine.c (try_combine): Use reg_or_subregno instead of handling
4789 just paradoxical SUBREGs and REGs.
4791 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4793 * config/arc/arc.c (arc_finalize_pic): Remove function.
4794 (arc_must_save_register): We use single base PIC register, remove
4795 checks to save/restore the PIC register.
4796 (arc_expand_prologue): Likewise.
4797 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4799 (arc_verify_short): Likewise.
4800 (arc_attr_type): Likewise.
4801 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4802 (walk_stores): Likewise.
4803 (arc_address_cost): Make it static.
4804 (arc_verify_short): Likewise.
4805 (branch_dest): Likewise.
4806 (arc_attr_type): Likewise.
4807 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4808 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4809 (arc_final_prescan_insn): Remove inserting the nops due to
4810 hardware hazards. It is done in reorg step.
4811 (insn_length_variant_t): Remove.
4812 (insn_length_parameters_t): Likewise.
4813 (arc_insn_length_parameters): Likewise.
4814 (arc_get_insn_variants): Likewise.
4815 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4817 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4820 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4823 PR tree-optimization/84687
4824 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4826 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4828 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4830 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4831 Rename to ppc_speculation_barrier.
4832 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4833 __builtin_ppc_speculation_barrier.
4835 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4838 * combine.c (combine_simplify_rtx): Don't try to simplify if
4839 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4842 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4844 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4845 to 32 bytes when compiling for POWER9.
4847 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4850 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4851 regparm >= 3 with no arg reg available also for calls with
4852 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4855 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4857 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4859 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4862 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4864 2018-03-05 Richard Biener <rguenther@suse.de>
4866 PR tree-optimization/84486
4867 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4868 When inserting a __builtin_assume_aligned call set the LHS
4869 SSA name alignment info accordingly.
4871 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4873 PR tree-optimization/84114
4874 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4875 Avoid reassociation of FLOAT_MODE addition.
4877 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4879 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4880 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4881 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4882 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4883 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4884 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4885 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4887 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4888 __builtin_ia32_wbinvd): New builtins.
4889 (SPECIAL_ARGS2): New.
4890 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4891 (SPECIAL_ARGS2): New.
4892 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4893 (ix86_valid_target_attribute_inner_p): Ditto.
4894 (ix86_init_mmx_sse_builtins): Add special_args2.
4895 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4896 TARGET_WBNOINVD_P): New.
4897 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4898 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4899 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4900 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4901 * config/i386/pconfigintrin.h: New file.
4902 * config/i386/wbnoinvdintrin.h: Ditto.
4903 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4904 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4906 2018-03-05 Richard Biener <rguenther@suse.de>
4908 PR tree-optimization/84670
4909 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4911 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4912 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4913 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4914 doesn't grow if all successors (recursively) were visited at least
4917 2018-03-05 Richard Biener <rguenther@suse.de>
4919 PR tree-optimization/84650
4920 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4921 if executed in the loop pipeline.
4923 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4925 * doc/configfiles.texi (Configuration Files): Move info about
4926 conditionalizing $target-protos.h to...
4927 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4928 differs from $target-protos.h.
4930 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4931 Chung-Ju Wu <jasonwucj@gmail.com>
4933 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4934 * config/nds32/nds32-multiple.md (setmemsi): Define.
4935 * config/nds32/nds32-memory-manipulation.c
4936 (nds32_gen_dup_4_byte_to_word_value): New.
4937 (emit_setmem_word_loop): New.
4938 (emit_setmem_byte_loop): New.
4939 (nds32_expand_setmem_loop): New.
4940 (nds32_expand_setmem_loop_v3m): New.
4941 (nds32_expand_setmem_unroll): New.
4942 (nds32_expand_setmem): New.
4944 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4945 Chung-Ju Wu <jasonwucj@gmail.com>
4947 * config/nds32/nds32-memory-manipulation.c
4948 (nds32_emit_load_store): New.
4949 (nds32_emit_post_inc_load_store): New.
4950 (nds32_emit_mem_move): New.
4951 (nds32_emit_mem_move_block): New.
4952 (nds32_expand_movmemsi_loop_unknown_size): New.
4953 (nds32_expand_movmemsi_loop_known_size): New.
4954 (nds32_expand_movmemsi_loop): New.
4955 (nds32_expand_movmemsi_unroll): New.
4956 (nds32_expand_movmemqi): Rename ...
4957 (nds32_expand_movmemsi): ... to this.
4958 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4959 (movmemsi): ... to this.
4960 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4961 (nds32_expand_movmemsi): ... to this.
4963 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4964 Monk Chiang <sh.chiang04@gmail.com>
4965 Chung-Ju Wu <jasonwucj@gmail.com>
4967 * config/nds32/nds32-protos.h
4968 (nds32_expand_load_multiple): New arguments.
4969 (nds32_expand_store_multiple): Ditto.
4970 (nds32_valid_multiple_load_store): Rename ...
4971 (nds32_valid_multiple_load_store_p): ... to this.
4972 * config/nds32/nds32-memory-manipulation.c
4973 (nds32_expand_load_multiple): Refine implementation.
4974 (nds32_expand_store_multiple): Ditto.
4975 * config/nds32/nds32-multiple.md
4976 (load_multiple): Update nds32_expand_load_multiple interface.
4977 (store_multiple): Update nds32_expand_store_multiple interface.
4978 * config/nds32/nds32-predicates.c
4979 (nds32_valid_multiple_load_store): Rename ...
4980 (nds32_valid_multiple_load_store_p): ... to this and refine
4982 * config/nds32/predicates.md
4983 (nds32_load_multiple_and_update_address_operation): New predicate.
4984 (nds32_store_multiple_and_update_address_operation): New predicate.
4986 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4987 Chung-Ju Wu <jasonwucj@gmail.com>
4989 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4990 (combo): New attribute.
4991 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4993 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4995 * config/nds32/nds32.opt: Change -mcmodel= default value.
4997 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4998 Monk Chiang <sh.chiang04@gmail.com>
4999 Chung-Ju Wu <jasonwucj@gmail.com>
5001 * config/nds32/constants.md (unspec_element): New enum.
5002 * config/nds32/constraints.md (Umw): New constraint.
5003 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5004 * config/nds32/nds32-intrinsic.md: Likewise.
5005 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5006 (nds32_valid_smw_lwm_base_p): New.
5007 (nds32_output_smw_single_word): New.
5008 (nds32_output_lmw_single_word): New.
5009 (nds32_expand_unaligned_load): New.
5010 (nds32_expand_unaligned_store): New.
5011 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5012 (nds32_output_smw_single_word): Declare.
5013 (nds32_output_lmw_single_word): Declare.
5014 (nds32_expand_unaligned_load): Declare.
5015 (nds32_expand_unaligned_store): Declare.
5016 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5017 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5018 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5019 NDS32_BUILTIN_UASTORE_DW.
5020 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5023 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5024 Kito Cheng <kito.cheng@gmail.com>
5025 Chung-Ju Wu <jasonwucj@gmail.com>
5027 * config/nds32/nds32-intrinsic.c
5028 (nds32_expand_builtin_null_ftype_reg): Delete.
5029 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5030 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5031 (nds32_read_argument): New.
5032 (nds32_legitimize_target): Ditto.
5033 (nds32_legitimize_argument): Ditto.
5034 (nds32_check_constant_argument): Ditto.
5035 (nds32_expand_unop_builtin): Ditto.
5036 (nds32_expand_unopimm_builtin): Ditto.
5037 (nds32_expand_binop_builtin): Ditto.
5038 (nds32_builtin_decl_impl): Ditto.
5039 (builtin_description): Ditto.
5040 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5041 (nds32_init_builtins_impl): Ditto.
5042 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5043 (nds32_builtin_decl): New.
5044 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5045 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5047 2018-03-02 Jeff Law <law@redhat.com>
5049 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5050 (redundant_insn, fill_simple_delay_slots): Likewise.
5051 (fill_slots_from_thread): Likewise.
5052 * resource.c (mark_referenced_resources): Likewise.
5053 (mark_set_resources, find_dead_or_set_registers): Likewise.
5055 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5057 * substring-locations.h (format_warning_va): Formatting fix for
5059 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5061 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5062 * substring-locations.c: Include intl.h.
5063 (format_warning_va): Turned into small wrapper around
5064 format_warning_n_va, renamed to ...
5065 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5066 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5068 (format_warning_at_substring_n): New function.
5069 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5070 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5071 format_warning_at_substring with just a shorter name instead of
5072 const function pointer.
5073 (fmtwarn_n): New function.
5074 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5075 appropriate, get rid of all the fmtstr temporaries, move conditionals
5076 with G_() wrapped string literals directly into fmtwarn arguments,
5077 cast dir.len to (int), formatting fixes.
5079 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5081 * doc/invoke.texi: Remove "Cilk Plus" references.
5083 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5084 Richard Biener <rguenther@suse.de>
5087 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5088 for error or warning attributes if CALL_FROM_THUNK_P is set.
5091 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5094 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5095 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5098 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5099 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5101 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5102 instead of -1U in last predictors element's probability member.
5104 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5107 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5108 arguments if they are comparable.
5110 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5112 PR tree-optimization/84634
5113 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5114 masks and masked_loop_p with a single loop_masks, making sure it's
5115 null for bb vectorization.
5117 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5119 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5120 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5121 loop->force_vectorize to check whether there is no alias.
5123 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5126 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5128 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5130 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5131 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5132 instead of a loop around prev_real_insn.
5133 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5137 * config/i386/i386.c (ix86_print_operand): Use conditional
5138 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5141 2018-03-02 Richard Biener <rguenther@suse.de>
5143 PR tree-optimization/84427
5144 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5145 (bitmap_set_subtract_values): Rewrite to handle multiple
5148 (prune_clobbered_mems): Likewise.
5149 (phi_translate): Take edge instead of pred/phiblock.
5150 (phi_translate_1): Likewise.
5151 (phi_translate_set): Likewise. Insert all translated
5152 exprs for a value into the set, keeping possibly multiple
5153 expressions per value.
5154 (compute_antic_aux): Adjust for phi_translate changes.
5155 When intersecting union the expressions and prune those
5156 not in the final value set, keeping possibly multiple
5157 expressions per value. Do not use value-insertion
5158 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5159 all expressions. Add verification that the value-sets
5160 only shrink during iteration.
5161 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5162 (do_pre_regular_insertion): Likewise.
5163 (do_pre_partial_partial_insertion): Likewise.
5165 2018-03-02 Richard Biener <rguenther@suse.de>
5168 * config/darwin.c (saved_debug_info_level): New static global.
5169 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5170 (darwin_asm_lto_end): Restore debug info generation settings.
5172 2018-03-01 Martin Liska <mliska@suse.cz>
5175 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5178 2018-03-01 Richard Biener <rguenther@suse.de>
5181 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5182 type annotation with LTO when debug was disabled at compile-time.
5184 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5186 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5188 (mips_final_postscan_insn): Likewise.
5190 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5192 PR rtl-optimization/84528
5193 * alias.c (init_alias_target): Add commentary.
5194 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5195 a unique base value if the frame pointer is not eliminated
5196 to the stack pointer.
5198 2018-03-01 Tom de Vries <tom@codesourcery.com>
5200 PR rtl-optimization/83327
5201 * lra-int.h (hard_regs_spilled_into): Declare.
5202 * lra.c (hard_regs_spilled_into): Define.
5203 (init_reg_info): Init hard_regs_spilled_into.
5204 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5205 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5206 (process_bb_lives): Handle hard_regs_spilled_into.
5207 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5208 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5210 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5212 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5213 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5214 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5215 * config/rs6000/aix72.h: New file.
5217 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5219 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5220 instead of warning_at with conditional singular and plural messages
5224 * stor-layout.c (update_alignment_for_field): For
5225 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5226 && !DECL_PACKED (field), do the alignment update, just use
5227 only desired_align instead of MAX (type_align, desired_align)
5229 (place_field): Don't do known_align < desired_align handling
5230 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5231 is non-NULL, instead do it after rli->prev_field handling and
5232 only if not within a bitfield word. For DECL_PACKED (field)
5233 use type_align of BITS_PER_UNIT.
5235 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5237 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5238 superfluous parentheses and trailing spaces.
5240 2018-02-28 Richard Biener <rguenther@suse.de>
5242 PR tree-optimization/84584
5243 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5244 SCoPs with fake exit edge.
5246 2018-02-28 Martin Liska <mliska@suse.cz>
5249 * timevar.c (timer::print): Fix format to properly print 100%
5252 2018-02-28 Richard Biener <rguenther@suse.de>
5255 * genmatch.c (capture_info::walk_match): Do not mark
5256 captured expressions without operands as expr_p given
5257 they act more like predicates and should be subject to
5258 "lost tail" side-effect preserving.
5260 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5262 PR rtl-optimization/81611
5263 * auto-inc-dec.c (attempt_change): Move dead note from
5264 mem_insn if it's the next use of regno
5265 (find_address): Take address use of reg holding
5266 non-incremented value. Add parm to limit search to the named
5268 (merge_in_block): Attempt to use a mem insn that is the next
5269 use of the original regno.
5271 2018-02-27 Martin Sebor <msebor@redhat.com>
5274 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5275 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5277 2018-02-27 Martin Sebor <msebor@redhat.com>
5279 PR translation/84207
5280 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5281 n argument to unsigned HOST_WIDE_INT.
5282 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5283 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5284 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5285 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5287 2018-02-27 Richard Biener <rguenther@suse.de>
5289 PR tree-optimization/84512
5290 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5291 Do not use the estimate returned from record_stmt_cost for
5292 the scalar iteration cost but sum properly using add_stmt_cost.
5294 2018-02-27 Richard Biener <rguenther@suse.de>
5296 PR tree-optimization/84466
5297 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5298 Adjust last change to less strictly validate use operands.
5300 2018-02-27 Martin Liska <mliska@suse.cz>
5302 PR gcov-profile/84548
5303 * gcov.c (process_file): Allow partial overlap and consider it
5304 also as group functions.
5305 (output_lines): Properly calculate range of lines for a group.
5307 2018-02-27 Martin Liska <mliska@suse.cz>
5309 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5310 'ggc' suffixes. Change first column width.
5311 (timer::print): Fix formatting of the column.
5313 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5315 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5316 preserve inline entry blocks for the sake of debug inline
5317 entry point markers alone.
5318 (remove_unused_locals): Suggest in comments a better place to
5319 force the preservation of inline entry blocks that are
5320 otherwise unused, but do not preserve them.
5322 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5324 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5326 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5329 * config/i386/constraints.md (Bs): Replace
5330 ix86_indirect_branch_register with
5331 TARGET_INDIRECT_BRANCH_REGISTER.
5333 * config/i386/i386.md (indirect_jump): Likewise.
5334 (tablejump): Likewise.
5335 (*sibcall_memory): Likewise.
5336 (*sibcall_value_memory): Likewise.
5337 Peepholes of indirect call and jump via memory: Likewise.
5338 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5339 (*sibcall_value_GOT_32): Likewise.
5340 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5341 (GOT_memory_operand): Likewise.
5342 (call_insn_operand): Likewise.
5343 (sibcall_insn_operand): Likewise.
5344 (GOT32_symbol_operand): Likewise.
5345 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5347 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5349 PR rtl-optimization/83496
5350 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5351 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5352 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5353 redundant insn, if any.
5354 (relax_delay_slots): Likewise.
5355 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5357 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5359 PR tree-optimization/83965
5360 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5361 that grouped statements are part of a reduction chain. Return
5362 true if the statement is not marked as a reduction itself but
5364 (vect_recog_dot_prod_pattern): Don't check whether the statement
5365 is part of a group here.
5366 (vect_recog_sad_pattern): Likewise.
5367 (vect_recog_widen_sum_pattern): Likewise.
5369 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5372 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5374 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5377 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5379 (ix86_output_indirect_function_return): New prototype.
5380 (ix86_split_simple_return_pop_internal): Likewise.
5381 * config/i386/i386.c (indirect_return_via_cx): New.
5382 (indirect_return_via_cx_bnd): Likewise.
5383 (indirect_thunk_name): Handle return va CX_REG.
5384 (output_indirect_thunk_function): Create alias for
5385 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5386 (ix86_output_indirect_jmp): Remove the bool argument.
5387 (ix86_output_indirect_function_return): New function.
5388 (ix86_split_simple_return_pop_internal): Likewise.
5389 * config/i386/i386.md (*indirect_jump): Don't pass false
5390 to ix86_output_indirect_jmp.
5391 (*tablejump_1): Likewise.
5392 (simple_return_pop_internal): Change it to define_insn_and_split.
5393 Call ix86_split_simple_return_pop_internal to split it for
5395 (simple_return_indirect_internal): Call
5396 ix86_output_indirect_function_return instead of
5397 ix86_output_indirect_jmp.
5399 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5402 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5403 memset and value initialization afterwards.
5405 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5407 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5409 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5412 * common/config/aarch64/aarch64-common.c
5413 (aarch_option_optimization_table[]): Switch
5414 off fomit-frame-pointer
5416 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5417 Chung-Ju Wu <jasonwucj@gmail.com>
5419 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5421 (store_multiple): Ditto.
5423 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5425 * config.gcc: Add --with-cpu support for nds32 target.
5426 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5427 * config/nds32/nds32.opt: Add -mcpu= option.
5429 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5431 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5432 isel=yes): Warn for these deprecated options.
5434 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5436 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5437 ISA_2_5_MASKS_EMBEDDED.
5439 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5441 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5442 p->max as pointers rather than using iterative_hash_expr.
5444 2018-02-23 Carl Love <cel@us.ibm.com>
5446 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5447 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5449 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5450 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5451 P8V_BUILTIN_VEC_VUNSIGNED2.
5453 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5456 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5457 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5459 * lra-constraints.c (curr_insn_transform): Set up
5460 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5462 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5463 * lra-eliminations.c (spill_pseudos): Ditto.
5464 (process_insn_for_elimination): Ditto.
5465 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5466 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5467 LRA_NON_CLOBBERED_ALT.
5469 2018-02-22 Martin Sebor <msebor@redhat.com>
5471 PR tree-optimization/84480
5472 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5473 to maybe_diag_stxncpy_trunc. Call it.
5474 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5475 from gimple_fold_builtin_strcpy. Print inlining stack.
5476 (handle_builtin_stxncpy): Print inlining stack.
5477 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5479 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5482 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5483 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5484 and -fcheck-pointer-bounds are used together.
5485 (indirect_thunk_prefix): New enum.
5486 (indirect_thunk_need_prefix): New function.
5487 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5488 "_nt" instead of "_bnd" for NOTRACK prefix.
5489 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5490 (output_indirect_thunk_function): Likewise.
5492 (ix86_code_end): Update output_indirect_thunk_function calls.
5493 (ix86_output_indirect_branch_via_reg): Replace
5494 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5495 (ix86_output_indirect_branch_via_push): Likewise.
5496 (ix86_output_function_return): Likewise.
5497 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5498 incompatible with -fcf-protection=branch and
5499 -fcheck-pointer-bounds.
5501 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5504 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5505 Change gcc_assert call to output_operand_lossage.
5507 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5509 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5511 2018-02-22 DJ Delorie <dj@redhat.com>
5512 Sebastian Perta <sebastian.perta@renesas.com>
5513 Oleg Endo <olegendo@gcc.gnu.org>
5515 * config/rx/rx.c (rx_rtx_costs): New function.
5516 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5518 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5520 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5522 2018-02-22 Martin Liska <mliska@suse.cz>
5525 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5526 Add "native" as a possible value.
5528 2018-02-22 Martin Liska <mliska@suse.cz>
5531 * config/i386/i386.c (ix86_option_override_internal):
5532 Add "native" as a possible value for -march and -mtune.
5534 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5537 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5538 to all type variants.
5540 PR tree-optimization/84503
5541 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5542 width as info->bitpos + info->bitsize - start.
5543 (merged_store_group::merge_overlapping): Simplify width computation.
5544 (check_no_overlap): New function.
5545 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5546 start + width and last_order of the group, fail if check_no_overlap
5548 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5549 to group if check_no_overlap fails.
5551 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5553 * config/rs6000/altivec.md: Delete contraint arguments to
5554 define_expand, define_split, and define_peephole2, and in
5555 define_insn_and_split if always unused.
5556 * config/rs6000/darwin.md: Ditto.
5557 * config/rs6000/dfp.md: Ditto.
5558 * config/rs6000/rs6000.md: Ditto.
5559 * config/rs6000/sync.md: Ditto.
5560 * config/rs6000/vector.md: Ditto.
5561 * config/rs6000/vsx.md: Ditto.
5563 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5565 * config/rs6000/altivec.md: Write output control strings as braced
5566 blocks instead of double-quoted strings.
5567 * config/rs6000/darwin.md: Ditto.
5568 * config/rs6000/rs6000.md: Ditto.
5569 * config/rs6000/vector.md: Ditto.
5570 * config/rs6000/vsx.md: Ditto.
5572 2018-02-21 Jason Merrill <jason@redhat.com>
5574 PR c++/84314 - ICE with templates and fastcall attribute.
5575 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5577 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5579 * ipa-cp.c (determine_versionability): Fix comment typos.
5581 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5584 * ipa-cp.c (determine_versionability): Do not version functions caling
5587 2018-02-21 Martin Liska <mliska@suse.cz>
5590 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5591 Add "native" as a possible value.
5592 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5593 the macro when native cpu detection is available.
5595 2018-02-21 Martin Liska <mliska@suse.cz>
5598 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5599 Add "native" as a possible value.
5600 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5601 when native cpu detection is available.
5603 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5604 Martin Sebor <msebor@redhat.com>
5606 PR tree-optimization/84478
5607 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5609 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5610 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5611 support which is conservatively correct, for 2 only stay conservative
5612 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5613 argument to the 2 argument get_range_strlen, adjust 6 arg
5614 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5616 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5617 (gimple_fold_builtin_strlen): Pass true as last argument to
5620 2018-02-20 Martin Sebor <msebor@redhat.com>
5623 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5624 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5625 (builtin_memref::builtin_memref): Factor out parts into
5626 set_base_and_offset and call it.
5628 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5631 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5632 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5633 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5634 search at the associated MODE_INT.
5636 2018-02-20 Jeff Law <law@redhat.com>
5639 PR tree-optimization/81592
5641 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5642 for range data rather than using global data.
5643 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5644 range data rather than using global data.
5645 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5646 pass it to children as needed.
5647 (struct directive::fmtresult): Similarly.
5648 (struct directive::set_width): Similarly.
5649 (struct directive::set_precision): Similarly.
5650 (format_integer, format_directive, parse_directive): Similarly.
5651 (format_none): Accept unnamed vr_values parameter.
5652 (format_percent, format_floating, format_character): Similarly.
5653 (format_string, format_plain): Similarly.
5654 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5655 the EVRP range analyzer for range data rather than using global data.
5656 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5657 gimple-ssa-evrp-analyze.h
5658 (class sprintf_dom_walker): Add after_dom_children member function.
5659 Add evrp_range_analyzer member.
5660 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5661 range analyzer as needed.
5662 (sprintf_dom_walker::after_dom_children): New member function.
5663 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5665 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5666 (evrp_range_analyzer::pop_to_marker): Likewise.
5668 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5670 PR tree-optimization/84419
5671 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5672 with the required type if its current type is compatible but
5675 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5678 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5679 after vectorization.
5681 2018-02-20 Martin Liska <mliska@suse.cz>
5684 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5685 possible values if we don't have a hint.
5687 2018-02-20 Martin Liska <mliska@suse.cz>
5691 * final.c (shorten_branches): Build align_tab array with one
5693 * opts.c (finish_options): Add alignment option limit check.
5694 (MAX_CODE_ALIGN): Likewise.
5695 (MAX_CODE_ALIGN_VALUE): Likewise.
5696 * doc/invoke.texi: Document maximum allowed option value for
5697 all -falign-* options.
5699 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5702 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5703 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5704 * var-tracking.c (emit_note_insn_var_location): Remove all references
5705 to NOTE_INSN_CALL_ARG_LOCATION.
5706 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5707 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5708 Use copy_rtx_if_shared.
5709 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5710 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5711 (dwarf2out_var_location): Remove handling of
5712 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5714 * final.c (final_scan_insn): Remove all references to
5715 NOTE_INSN_CALL_ARG_LOCATION.
5716 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5717 before dumping final insns.
5718 * except.c (emit_note_eh_region_end): Remove all references to
5719 NOTE_INSN_CALL_ARG_LOCATION.
5720 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5721 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5722 * config/arc/arc.c (hwloop_optimize): Likewise.
5723 * config/arm/arm.c (create_fix_barrier): Likewise.
5724 * config/s390/s390.c (s390_chunkify_start): Likewise.
5725 * config/sh/sh.c (find_barrier): Likewise.
5726 * config/i386/i386.c (rest_of_insert_endbranch,
5727 ix86_seh_fixup_eh_fallthru): Likewise.
5728 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5729 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5730 * config/frv/frv.c (frv_function_prologue): Likewise.
5731 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5733 (note_outside_basic_block_p): Remove all references to
5734 NOTE_INSN_CALL_ARG_LOCATION.
5735 * gengtype.c (adjust_field_rtx_def): Likewise.
5736 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5738 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5739 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5742 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5745 PR tree-optimization/84452
5746 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5747 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5750 2018-02-19 Martin Liska <mliska@suse.cz>
5753 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5755 2018-02-19 Martin Liska <mliska@suse.cz>
5756 Richard Sandiford <richard.sandiford@linaro.org>
5758 PR tree-optimization/82491
5759 * gimple-fold.c (get_base_constructor): Make earlier bail out
5762 2018-02-19 Carl Love <cel@us.ibm.com>
5764 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5765 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5767 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5768 P8V_BUILTIN_VEC_NEG.
5770 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5772 * config/rl78/rl78.md (movdf): New define expand.
5774 2018-02-19 Martin Liska <mliska@suse.cz>
5777 * doc/invoke.texi: Fix typo.
5778 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5780 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5782 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5783 handle rs6000_single_float and rs6000_double_float specially for
5786 2018-02-16 Jeff Law <law@redhat.com>
5788 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5791 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5794 * ipa-inline.c (inline_small_functions): Fix a typo.
5796 2018-02-16 Nathan Sidwell <nathan@acm.org>
5798 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5800 2018-02-16 Carl Love <cel@us.ibm.com>
5802 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5803 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5804 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5805 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5806 expansion to P8V_BUILTIN_VEC_FLOAT2.
5808 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5810 PR rtl-optimization/70023
5811 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5812 src_regno into account.
5814 2018-02-16 Carl Love <cel@us.ibm.com>
5816 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5817 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5818 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5819 * config/rs6000/rs6000.c: Remove case statements for
5820 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5821 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5822 and P9V_BUILTIN_VEC_VINSERT4B.
5823 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5824 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5825 * config/rs6000/vsx.md:
5826 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5829 2018-02-16 Carl Love <cel@us.ibm.com>
5831 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5833 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5835 * config/rs6000/rs6000-c.c: Add the definitions for
5836 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5837 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5838 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5839 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5840 definition for insert4b and define insn *insert3b_internal.
5841 * doc/extend.texi: Add documentation for vec_extract4b.
5843 2018-02-16 Nathan Sidwell <nathan@acm.org>
5845 * doc/extend.texi (Backwards Compatibility): Mention friend
5846 injection. Note for-scope is deprecated.
5847 * doc/invoke.texi (-ffriend-injection): Deprecate.
5849 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5851 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5852 that moved to I2, also allow destinations that are a paradoxical
5853 subreg (instead of a normal reg).
5855 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5858 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5861 2018-02-16 Richard Biener <rguenther@suse.de>
5863 PR tree-optimization/84037
5864 PR tree-optimization/84016
5866 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5867 Adjust vec_construct for the fact we need additional higher latency
5868 128bit inserts for AVX256 and AVX512 vector builds.
5869 (ix86_add_stmt_cost): Scale vector construction cost for
5872 2018-02-16 Richard Biener <rguenther@suse.de>
5874 PR tree-optimization/84417
5875 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5876 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5877 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5879 2018-02-16 Martin Liska <mliska@suse.cz>
5882 * internal-fn.def (ASAN_CHECK): Set proper flags.
5883 (ASAN_MARK): Likewise.
5885 2018-02-16 Julia Koval <julia.koval@intel.com>
5887 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5888 from PTA_CANNONLAKE.
5890 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5893 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5894 Use ++iter rather than iter++ for std::list iterators.
5895 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5896 defer deleting them until all nodes in the forest are processed. Do
5897 free even leaf nodes. Change to_process into auto_vec.
5900 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5901 * vec.h (vec_default_construct): Use memset instead of placement new
5902 if BROKEN_VALUE_INITIALIZATION is defined.
5903 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5904 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5907 PR rtl-optimization/83723
5908 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5909 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5910 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5912 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5914 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5916 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5918 PR rtl-optimization/81443
5919 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5920 from inner REGs to paradoxical SUBREGs.
5922 2018-02-16 Richard Biener <rguenther@suse.de>
5924 PR tree-optimization/84399
5925 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5926 For operands we can analyze at their definition make sure we can
5927 analyze them at each use as well.
5929 2018-02-16 Richard Biener <rguenther@suse.de>
5931 PR tree-optimization/84190
5932 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5933 volatile accesses if the decl isn't volatile.
5935 2018-02-15 Jason Merrill <jason@redhat.com>
5937 PR c++/84314 - ICE with templates and fastcall attribute.
5938 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5939 TYPE_CANONICAL on an existing type.
5941 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5943 PR tree-optimization/84383
5944 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5945 dstoff nor call operand_equal_p if dstbase is NULL.
5947 PR tree-optimization/84334
5948 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5949 also a CONSTANT_CLASS_P, punt.
5951 2018-02-14 Jim Wilson <jimw@sifive.com>
5953 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5954 first SMALL_OPERAND check. New local min_second_step. Move assert
5955 to where locals are set. Add TARGET_RVC support.
5956 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5958 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5960 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5962 2018-02-14 Martin Sebor <msebor@redhat.com>
5964 PR tree-optimization/83698
5965 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5966 arrays constrain the offset range to their bounds.
5967 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5968 (builtin_access::overlap): Avoid setting the size of overlap if it's
5970 (maybe_diag_overlap): Also consider arrays when deciding what values
5971 of offsets to include in diagnostics.
5973 2018-02-14 Martin Sebor <msebor@redhat.com>
5976 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5977 that correspond to the kind of a declaration.
5979 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5982 * config/pa/pa.md: Load address of PIC label using the linkage table
5983 if the label is nonlocal.
5985 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5987 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5988 warning message if user requests -maltivec=be.
5989 * doc/invoke.texi: Document deprecation of -maltivec=be.
5991 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5994 * config/rs6000/rs6000-c.c: Update definitions for
5995 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5996 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5998 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6001 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6002 add _get_ssp intrinsics. Remove argument from
6003 __builtin_ia32_rdssp[d|q].
6004 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6005 * config/i386/i386-builtin.def: Remove argument from
6006 __builtin_ia32_rdssp[d|q].
6007 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6008 ix86_expand_special_args_builtin for _rdssp[d|q].
6009 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6010 Clear register before usage.
6011 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6012 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6014 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6016 PR tree-optimization/84357
6017 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6018 operand 1 of an ARRAY_REF too.
6020 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6023 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6024 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6026 (set_of_reg): New struct.
6027 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6028 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6029 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6031 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6032 Split into bitclr, bitset, bitinvert patterns if appropriate.
6033 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6034 use rx_fuse_in_memory_bitop.
6035 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6036 to named insn, correct maximum insn length.
6038 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6041 * machmode.def: Define a complex mode for PARTIAL_INT.
6042 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6044 * doc/rtl.texi: Document CSPImode.
6045 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6047 (msp430_hard_regno_nregs_with_padding): Likewise.
6049 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6052 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6054 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6056 PR rtl-optimization/84169
6057 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6058 we generated a parallel as new i3 and we split that to new i2 and i3
6059 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6060 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6061 those to i2, not i1. Partially rewrite this scan code.
6063 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6066 * stor-layout.c (place_field): For variable length fields, adjust
6067 offset_align afterwards not just based on the field's alignment,
6068 but also on the size.
6071 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6072 of exps and logs in the use_exp2 case.
6074 2018-02-13 Jeff Law <law@redhat.com>
6076 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6079 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6082 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6086 * common.opt (gas-loc-support, gas-locview-support): New.
6087 (ginline-points, ginternal-reset-location-views): New.
6088 * doc/invoke.texi: Document them. Use @itemx where intended.
6089 (gvariable-location-views): Adjust.
6090 * target.def (reset_location_view): New.
6091 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6092 (TARGET_RESET_LOCATION_VIEW): New.
6093 * doc/tm.texi: Rebuilt.
6094 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6095 (dwarf2out_default_as_locview_support): New.
6096 (output_asm_line_debug_info): Use option variables.
6097 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6098 (output_loc_list): Likewise.
6099 (add_high_low_attributes): Check option variables.
6100 Don't output entry view attribute in strict mode.
6101 (gen_inlined_subroutine_die): Check option variables.
6102 (dwarf2out_inline_entry): Likewise.
6103 (init_sections_and_labels): Likewise.
6104 (dwarf2out_early_finish): Likewise.
6105 (maybe_reset_location_view): New, from...
6106 (dwarf2out_var_location): ... here. Call it.
6107 * debug.h (dwarf2out_default_as_loc_support): Declare.
6108 (dwarf2out_default_as_locview_support): Declare.
6109 * hooks.c (hook_int_rtx_insn_0): New.
6110 * hooks.h (hook_int_rtx_insn_0): Declare.
6111 * toplev.c (process_options): Take -gas-loc-support and
6112 -gas-locview-support from dwarf2out. Enable
6113 -gvariable-location-views by default only with locview
6114 assembler support. Enable -ginternal-reset-location-views by
6115 default only if the target defines the corresponding hook.
6116 Enable -ginline-points by default if location views are
6117 enabled; force it disabled if statement frontiers are
6119 * tree-inline.c (expand_call_inline): Check option variables.
6120 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6122 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6124 PR tree-optimization/84321
6125 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6126 handling. Also check whether the anti-range contains any values
6127 that satisfy the mask; switch to a VR_RANGE if not.
6129 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6132 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6134 2018-02-13 Martin Jambor <mjambor@suse.cz>
6137 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6138 of call statements, also set location of a load to a temporary.
6140 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6142 * config/rl78/rl78.c (add_vector_labels): New function.
6143 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6144 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6145 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6146 which checks that no arguments are passed.
6147 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6148 * doc/extend.texi: Documentation for the new attribute.
6150 2018-02-13 Andreas Schwab <schwab@suse.de>
6152 * config/riscv/linux.h (CPP_SPEC): Define.
6154 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6157 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6158 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6159 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6160 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6161 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6162 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6163 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6164 temporarily for AES and PCLMUL builtins.
6166 PR tree-optimization/84339
6167 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6168 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6172 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6173 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6174 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6176 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6178 * omp-simd-clone.h: New file.
6179 * omp-simd-clone.c: Include omp-simd-clone.h.
6180 (expand_simd_clones): No longer static.
6181 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6182 cgraph.h and omp-simd-clone.h.
6183 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6184 (vect_recog_widen_shift_pattern): Formatting fix.
6185 (vect_pattern_recog_1): Don't check optab for calls.
6188 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6189 operands[2] into a REG before using gen_lowpart on it.
6191 2018-02-12 Jeff Law <law@redhat.com>
6194 * config/sh/sh.c (find_barrier): Consider a sibling call
6197 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6198 successfully back substituting a reg.
6200 2018-02-12 Richard Biener <rguenther@suse.de>
6202 PR tree-optimization/84037
6203 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6204 parameter, move visited init to caller.
6205 (vect_slp_analyze_operations): Separate cost from validity
6206 check, initialize visited once for all instances.
6207 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6209 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6211 (vect_model_promotion_demotion_cost): Likewise.
6212 (vectorizable_bswap): Guard cost modeling with !slp_node
6213 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6215 (vectorizable_call): Likewise.
6216 (vectorizable_conversion): Likewise.
6217 (vectorizable_assignment): Likewise.
6218 (vectorizable_shift): Likewise.
6219 (vectorizable_operation): Likewise.
6220 (vectorizable_store): Likewise.
6221 (vectorizable_load): Likewise.
6222 (vectorizable_condition): Likewise.
6223 (vectorizable_comparison): Likewise.
6225 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6228 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6229 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6230 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6232 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6235 * params.def (inline-min-speedup): Increase from 8 to 15.
6236 (max-inline-insns-auto): Decrease from 40 to 30.
6237 * ipa-split.c (consider_split): Add some buffer for function to
6238 be considered inlining candidate.
6239 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6242 2018-02-12 Richard Biener <rguenther@suse.de>
6244 PR tree-optimization/84037
6245 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6246 matched stmts if we cannot swap the non-matched ones.
6248 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6250 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6251 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6252 _mm_maskz_scalef_round_ss): New intrinsics.
6253 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6254 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6255 __builtin_ia32_scalefss_round): Remove.
6256 (__builtin_ia32_scalefsd_mask_round,
6257 __builtin_ia32_scalefss_mask_round): New intrinsics.
6258 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6259 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6260 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6261 "<round_constraint>")): Changed to ...
6262 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6263 "<round_scalar_constraint>")): ... this.
6264 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6265 %0, %1, %2<round_op3>}"): Changed to ...
6266 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6267 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6268 %2<round_scalar_mask_op3>}"): ... this.
6269 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6271 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6273 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6274 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6275 (_mm_maskz_sqrt_round_ss): New intrinsics.
6276 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6277 (__builtin_ia32_sqrtsd_mask_round)
6278 (__builtin_ia32_sqrtss_mask_round): New builtins.
6279 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6280 (__builtin_ia32_sqrtss_round): Remove.
6281 (__builtin_ia32_sqrtsd_mask_round)
6282 (__builtin_ia32_sqrtss_mask_round): New builtins.
6283 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6284 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6285 ((match_operand:VF_128 1 "vector_operand"
6286 "xBm,<round_constraint>")): Changed to ...
6287 ((match_operand:VF_128 1 "vector_operand"
6288 "xBm,<round_scalar_constraint>")): ... this.
6289 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6290 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6291 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6292 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6293 %<iptr>1<round_scalar_mask_op3>}): ... this.
6294 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6295 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6297 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6300 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6301 Cast vec_cmpeq result to correct type.
6302 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6303 Cast vec_cmpgt result to correct type.
6305 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6307 * final.c (final_scan_insn_1): Renamed from...
6308 (final_scan_insn): ... this. New wrapper, to recover
6309 seen from the outermost call in recursive ones.
6310 * config/sparc/sparc.c (output_return): Drop seen from call.
6311 (output_sibcall): Likewise.
6312 * config/visium/visium.c (output_branch): Likewise.
6314 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6316 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6319 2018-02-10 Alan Modra <amodra@gmail.com>
6322 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6323 Specify LR as an input.
6325 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6328 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6329 remove_member_access_dummy_vars): New functions.
6330 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6331 lower_omp_1, execute_lower_omp): Use them.
6333 PR rtl-optimization/84308
6334 * shrink-wrap.c (spread_components): Release todo vector.
6336 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6338 PR rtl-optimization/57193
6339 * ira-color.c (struct allocno_color_data): Add member
6340 conflict_allocno_hard_prefs.
6341 (update_conflict_allocno_hard_prefs): New.
6342 (bucket_allocno_compare_func): Add a preference based on
6343 conflict_allocno_hard_prefs.
6344 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6345 (color_allocnos): Remove a dead code. Initiate
6346 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6348 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6351 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6352 constraint from =wa to wa. Avoid a subreg on the output operand,
6353 instead use a pseudo and subreg it in a move.
6354 (p9_xxbrd_<mode>): Changed to ...
6355 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6356 (p9_xxbrd_v2df): New expander.
6357 (p9_xxbrw_<mode>): Changed to ...
6358 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6359 (p9_xxbrw_v4sf): New expander.
6361 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6363 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6365 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6368 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6369 multiply in 32-bit mode.
6370 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6371 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6374 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6376 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6377 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6378 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6379 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6381 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6384 * dwarf2out.c (is_trivial_indirect_ref): New function.
6385 (dwarf2out_late_global_decl): Do not generate a location
6386 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6387 and that are not defined in the current unit.
6389 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6391 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6392 instead of a libcall for UNORDERED.
6394 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6397 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6398 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6400 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6403 * config/s390/s390.c (s390_set_current_function): Invoke
6404 s390_indirect_branch_settings also if fndecl didn't change.
6406 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6408 * config/rs6000/rs6000.md (blockage): Set length to zero.
6410 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6412 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6414 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6417 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6418 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6422 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6423 PARALLEL incoming that failed vt_get_decl_and_offset check.
6426 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6427 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6429 (get_variable_section): For decls in named .bss* sections pass true as
6430 second argument to bss_initializer_p.
6432 2018-02-09 Marek Polacek <polacek@redhat.com>
6433 Jakub Jelinek <jakub@redhat.com>
6436 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6437 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6438 Sync some changes from cxx_fold_indirect_ref.
6440 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6442 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6444 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6445 (BLOCK_INLINE_ENTRY_LABEL): New.
6446 (dwarf2out_var_location): Disregard inline entry markers.
6447 (inline_entry_data): New struct.
6448 (inline_entry_data_hasher): New hashtable type.
6449 (inline_entry_data_hasher::hash): New.
6450 (inline_entry_data_hasher::equal): New.
6451 (inline_entry_data_table): New variable.
6452 (add_high_low_attributes): Add DW_AT_entry_pc and
6453 DW_AT_GNU_entry_view attributes if a pending entry is found
6454 in inline_entry_data_table. Add old entry_pc attribute only
6455 if debug nonbinding markers are disabled.
6456 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6457 markers are enabled.
6458 (block_within_block_p, dwarf2out_inline_entry): New.
6459 (dwarf2out_finish): Check that no entries remained in
6460 inline_entry_data_table.
6461 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6462 (final_scan_insn, notice_source_line): Likewise.
6463 (rest_of_clean_state): Skip inline entry markers.
6464 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6466 * gimple.c (gimple_build_debug_inline_entry): New.
6467 * gimple.h (enum gimple_debug_subcode): Add
6468 GIMPLE_DEBUG_INLINE_ENTRY.
6469 (gimple_build_debug_inline_entry): Declare.
6470 (gimple_debug_inline_entry_p): New.
6471 (gimple_debug_nonbind_marker_p): Adjust.
6472 * insn-notes.def (INLINE_ENTRY): New.
6473 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6474 inline entry marker notes.
6475 (print_insn): Likewise.
6476 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6477 (INSN_DEBUG_MARKER_KIND): Likewise.
6478 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6479 * tree-inline.c (expand_call_inline): Build and insert
6480 debug_inline_entry stmt.
6481 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6482 inline entry blocks early, if nonbind markers are enabled.
6483 (dump_scope_block): Dump fragment info.
6484 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6485 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6486 (gimple_build_debug_inline_entry): New.
6487 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6488 Enable/disable inline entry points too.
6489 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6490 (DEBUG_INSN): Describe inline entry markers.
6492 * common.opt (gvariable-location-views): New.
6493 (gvariable-location-views=incompat5): New.
6494 * config.in: Rebuilt.
6495 * configure: Rebuilt.
6496 * configure.ac: Test assembler for view support.
6497 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6498 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6499 * dwarf2out.c (var_loc_view): New typedef.
6500 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6501 (dwarf2out_locviews_in_attribute): New.
6502 (dwarf2out_locviews_in_loclist): New.
6503 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6504 (enum dw_line_info_opcode): Add LI_adv_address.
6505 (struct dw_line_info_table): Add view.
6506 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6507 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6508 (zero_view_p): New variable.
6509 (ZERO_VIEW_P): New macro.
6510 (output_asm_line_debug_info): New.
6511 (struct var_loc_node): Add view.
6512 (add_AT_view_list, AT_loc_list): New.
6513 (add_var_loc_to_decl): Add view param. Test it against last.
6514 (new_loc_list): Add view params. Record them.
6515 (AT_loc_list_ptr): Handle loc and view lists.
6516 (view_list_to_loc_list_val_node): New.
6517 (print_dw_val): Handle dw_val_class_view_list.
6518 (size_of_die): Likewise.
6519 (value_format): Likewise.
6520 (loc_list_has_views): New.
6521 (gen_llsym): Set vl_symbol too.
6522 (maybe_gen_llsym, skip_loc_list_entry): New.
6523 (dwarf2out_maybe_output_loclist_view_pair): New.
6524 (output_loc_list): Output view list or entries too.
6525 (output_view_list_offset): New.
6526 (output_die): Handle dw_val_class_view_list.
6527 (output_dwarf_version): New.
6528 (output_compilation_unit_header): Use it.
6529 (output_skeleton_debug_sections): Likewise.
6530 (output_rnglists, output_line_info): Likewise.
6531 (output_pubnames, output_aranges): Update version comments.
6532 (output_one_line_info_table): Output view numbers in asm comments.
6533 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6534 Call maybe_gen_llsym.
6535 (loc_list_from_tree_1): Adjust.
6536 (add_AT_location_description): Create view list attribute if
6537 needed, check it's absent otherwise.
6538 (convert_cfa_to_fb_loc_list): Adjust.
6539 (maybe_emit_file): Call output_asm_line_debug_info for test.
6540 (dwarf2out_var_location): Reset views as needed. Precompute
6541 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6542 attribute. Set view.
6543 (new_line_info_table): Reset next view.
6544 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6545 (dwarf2out_source_line): Likewise. Output view resets and labels to
6546 the assembler, or select appropriate line info opcodes.
6547 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6548 (optimize_string_length): Catch it. Adjust.
6549 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6550 dw_val_class_view_list, and remove it if no longer needed.
6551 (hash_loc_list): Hash view numbers.
6552 (loc_list_hasher::equal): Compare them.
6553 (optimize_location_lists): Check whether a view list symbol is
6554 needed, and whether the locview attribute is present, and
6555 whether they match. Remove the locview attribute if no longer
6557 (index_location_lists): Call skip_loc_list_entry for test.
6558 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6559 Use output_dwarf_version.
6560 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6561 (struct dw_val_node): Add val_view_list.
6562 * final.c (SEEN_NEXT_VIEW): New.
6563 (set_next_view_needed): New.
6564 (clear_next_view_needed): New.
6565 (maybe_output_next_view): New.
6566 (final_start_function): Rename to...
6567 (final_start_function_1): ... this. Take pointer to FIRST,
6568 add SEEN parameter. Emit param bindings in the initial view.
6569 (final_start_function): Reintroduce SEEN-less interface.
6570 (final): Rename to...
6571 (final_1): ... this. Take SEEN parameter. Output final pending
6572 next view at the end.
6573 (final): Reintroduce seen-less interface.
6574 (final_scan_insn): Output pending next view before switching
6575 sections or ending a block. Mark the next view as needed when
6576 outputting variable locations. Notify debug backend of section
6577 changes, and of location view changes.
6578 (rest_of_handle_final): Adjust.
6579 * toplev.c (process_options): Autodetect value for debug variable
6580 location views option. Warn on incompat5 without -gdwarf-5.
6581 * doc/invoke.texi (gvariable-location-views): New.
6582 (gvariable-location-views=incompat5): New.
6583 (gno-variable-location-views): New.
6585 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6587 PR tree-optimization/84136
6588 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6589 that the result of find_edge is non-NULL.
6591 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6594 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6595 storing integer register in SImode. Fix cost of 256 and 512
6596 byte aligned SSE register store.
6598 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6600 * config/i386/i386.c (ix86_multiplication_cost): Fix
6601 multiplication cost for TARGET_AVX512DQ.
6603 2018-02-08 Marek Polacek <polacek@redhat.com>
6605 PR tree-optimization/84238
6606 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6609 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6611 PR tree-optimization/84265
6612 * tree-vect-stmts.c (vectorizable_store): Don't treat
6613 VMAT_CONTIGUOUS accesses as grouped.
6614 (vectorizable_load): Likewise.
6616 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6618 PR tree-optimization/81635
6619 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6620 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6621 (test_round_for_mask): New functions.
6622 (wide_int_cc_tests): Call test_round_for_mask.
6623 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6624 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6625 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6626 range returned by get_range_info.
6628 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6631 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6632 * symtab.c: Include builtins.h
6633 (symtab_node::output_to_lto_symbol_table_p): Move here
6634 from lto-streamer-out.c:output_symbol_p.
6635 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6636 (output_symbol_p): Move all logic to symtab.c
6637 (produce_symtab): Update.
6639 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6641 * config/s390/s390-opts.h (enum indirect_branch): Define.
6642 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6643 (s390_indirect_branch_via_thunk)
6644 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6645 (enum s390_indirect_branch_type): Define.
6646 * config/s390/s390.c (struct s390_frame_layout, struct
6647 machine_function): Remove.
6648 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6649 (indirect_branch_table_label_no, indirect_branch_table_name):
6651 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6652 (enum s390_indirect_branch_option): Define.
6653 (s390_return_addr_from_memory): New function.
6654 (s390_handle_string_attribute): New function.
6655 (s390_attribute_table): Add new attribute handler.
6656 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6657 (s390_indirect_branch_via_thunk): New function.
6658 (s390_indirect_branch_via_inline_thunk): New function.
6659 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6660 sibling call optimization for non z10 compiles.
6661 (s390_emit_call): Force indirect branch target to be a single
6662 register. Add r1 clobber for non-z10 compiles.
6663 (s390_emit_epilogue): Emit return jump via return_use expander.
6664 (s390_reorg): Handle JUMP_INSNs as execute targets.
6665 (s390_option_override_internal): Perform validity checks for the
6666 new command line options.
6667 (s390_indirect_branch_attrvalue): New function.
6668 (s390_indirect_branch_settings): New function.
6669 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6670 (s390_output_indirect_thunk_function): New function.
6671 (s390_code_end): Implement target hook.
6672 (s390_case_values_threshold): Implement target hook.
6673 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6675 * config/s390/s390.h (struct s390_frame_layout)
6676 (struct machine_function): Move here from s390.c.
6677 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6678 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6679 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6680 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6681 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6682 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6683 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6684 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6685 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6686 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6687 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6688 (mnemonic attribute): Add values which aren't recognized
6690 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6691 pattern for branch conversion. Fix mnemonic attribute.
6692 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6693 indirect branch via thunk if requested.
6694 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6695 ("*indirect_jump"): Disable for branch conversion using out of
6697 ("indirect_jump_via_thunk<mode>_z10")
6698 ("indirect_jump_via_thunk<mode>")
6699 ("indirect_jump_via_inlinethunk<mode>_z10")
6700 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6701 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6702 ("casesi_jump_via_inlinethunk<mode>_z10")
6703 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6704 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6705 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6706 ("*indirect2_jump"): Disable for branch conversion.
6707 ("casesi_jump"): Turn into expander and expand patterns for branch
6709 ("return_use"): New expander.
6710 ("*return"): Emit return via thunk and rename it to ...
6711 ("*return<mode>"): ... this one.
6712 * config/s390/s390.opt: Add new options and and enum for the
6715 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6717 * lra-constraints.c (match_reload): Unconditionally use
6718 gen_lowpart_SUBREG, rather than selecting between that
6719 and equivalent gen_rtx_SUBREG code.
6721 2018-02-08 Richard Biener <rguenther@suse.de>
6723 PR tree-optimization/84233
6724 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6725 changed flag instead of boguously re-using phi_inserted.
6727 2018-02-08 Martin Jambor <mjambor@suse.cz>
6729 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6730 static local variables.
6732 2018-02-08 Richard Biener <rguenther@suse.de>
6734 PR tree-optimization/84278
6735 * tree-vect-stmts.c (vectorizable_store): When looking for
6736 smaller vector types to perform grouped strided loads/stores
6737 make sure the mode is supported by the target.
6738 (vectorizable_load): Likewise.
6740 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6742 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6743 Increase LDP/STP opportunities by adding adjacent callee-saves.
6745 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6747 PR rtl-optimization/84068
6748 PR rtl-optimization/83459
6749 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6751 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6753 PR tree-optimization/84224
6754 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6755 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6758 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6761 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6762 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6763 count, remove one USE.
6765 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6767 * doc/install.texi (Configuration): Document the
6768 --with-long-double-format={ibm,ieee} PowerPC configuration
6772 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6773 Convert from define_expand to be define_insn_and_split. Rework
6774 float/double/_Float128 conversions to QI/HI/SImode to work with
6775 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6776 conversions to QI/HImode types did a store and then a load to
6777 truncate the value. For conversions to VSX registers, don't split
6778 the insn, instead emit the code directly. Use the code iterator
6779 any_fix to combine signed and unsigned conversions.
6780 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6781 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6782 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6783 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6784 (fix_<mode>di2_hw): Likewise.
6785 (fixuns_<mode>di2_hw): Likewise.
6786 (fix_<mode>si2_hw): Likewise.
6787 (fixuns_<mode>si2_hw): Likewise.
6788 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6789 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6790 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6791 fix<uns>_trunc<SFDF:mode>si2_p8.
6792 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6794 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6795 (fix<uns>_<mode>_mem): Likewise.
6796 (fctiw<u>z_<mode>_mem): Likewise.
6797 (fix<uns>_<mode>_mem): Likewise.
6798 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6799 the register allocator from doing a direct move to the GPRs to do
6800 a store, and instead use the ISA 3.0 store byte/half-word from
6801 vector register instruction. For IEEE 128-bit floating point,
6802 also optimize stores of 32-bit ints.
6803 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6805 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6807 * genextract.c (push_pathstr_operand): New function to support
6809 (walk_rtx): Call push_pathstr_operand.
6810 (print_path): Support [a-zA-Z].
6812 2018-02-07 Richard Biener <rguenther@suse.de>
6814 PR tree-optimization/84037
6815 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6816 (cse_and_gimplify_to_preheader): Declare.
6817 (vect_get_place_in_interleaving_chain): Likewise.
6818 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6820 (_loop_vec_info::~_loop_vec_info): Delete it.
6821 (cse_and_gimplify_to_preheader): New function.
6822 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6823 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6824 (vectorizable_load): Likewise. For grouped stores always base
6825 the IV on the first element.
6826 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6827 condition before gimplifying.
6829 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6831 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6832 *DIV_EXPR and *MOD_EXPR.
6834 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6837 * config/i386/i386.c (ix86_option_override_internal): Mask out
6838 the CF_SET bit when checking -fcf-protection.
6840 2018-02-07 Tom de Vries <tom@codesourcery.com>
6843 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6846 2018-02-07 Richard Biener <rguenther@suse.de>
6848 PR tree-optimization/84204
6849 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6852 PR tree-optimization/84205
6853 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6854 special-case isl_ast_op_zdiv_r.
6856 PR tree-optimization/84223
6857 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6858 Only add conditions from within the region.
6859 (gather_bbs::after_dom_children): Adjust.
6861 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6864 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6865 * config/avr/avr.md: Only post-reload split REG-REG moves if
6866 either register is GENERAL_REG_P.
6868 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6870 PR tree-optimization/84235
6871 * tree-ssa-scopedtables.c
6872 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6873 if the subtraction is performed in floating point type where NaNs are
6874 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6875 build 1. Formatting fix.
6877 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6880 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6881 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6882 and skip it regardless of bb boundaries. Use CALL_P macro,
6883 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6884 unnecessarily, formatting fix.
6886 2018-02-06 Michael Collison <michael.collison@arm.com>
6888 * config/arm/thumb2.md:
6889 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6890 (*thumb_mov_notscc): Ditto.
6892 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6895 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6896 unsigned_fix, not "s".
6898 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6900 * configure.ac (gcc_fn_eh_frame_ro): New function.
6901 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6902 correct .eh_frame permissions.
6903 * configure: Regenerate.
6905 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6907 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6910 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6912 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6913 Display warning message for -mno-speculate-indirect-jumps.
6915 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6917 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6919 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6921 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6923 PR tree-optimization/84225
6924 * tree-eh.c (find_trapping_overflow): Only call
6925 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6927 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6930 * config/i386/i386.c: Reimplement the check of possible options
6931 -mibt/-mshstk conbination. Change error messages.
6932 * doc/invoke.texi: Fix a typo: remove extra '='.
6934 2018-02-06 Marek Polacek <polacek@redhat.com>
6936 PR tree-optimization/84228
6937 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6939 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6942 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6943 emitted arch directives.
6944 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6945 __ARM_FEATURE_COPROC before changing architectures.
6947 2018-02-06 Richard Biener <rguenther@suse.de>
6949 * config/i386/i386.c (print_reg): Fix typo.
6950 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6952 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6954 * configure: Regenerate.
6956 2018-02-05 Martin Sebor <msebor@redhat.com>
6958 PR tree-optimization/83369
6959 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6962 2018-02-05 Martin Liska <mliska@suse.cz>
6964 * doc/invoke.texi: Cherry-pick upstream r323995.
6966 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6968 * ira.c (ira_init_register_move_cost): Adjust comment.
6970 2018-02-05 Martin Liska <mliska@suse.cz>
6972 PR gcov-profile/84137
6973 * doc/gcov.texi: Fix typo in documentation.
6975 2018-02-05 Martin Liska <mliska@suse.cz>
6977 PR gcov-profile/83879
6978 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6979 using dlopen functionality.
6981 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6983 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6984 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6985 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6986 _mm_maskz_range_round_ss): New intrinsics.
6987 (__builtin_ia32_rangesd128_round)
6988 (__builtin_ia32_rangess128_round): Remove.
6989 (__builtin_ia32_rangesd128_mask_round,
6990 __builtin_ia32_rangess128_mask_round): New builtins.
6991 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6992 __builtin_ia32_rangess128_round): Remove.
6993 (__builtin_ia32_rangesd128_mask_round,
6994 __builtin_ia32_rangess128_mask_round): New builtins.
6995 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6996 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6997 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6998 "<round_saeonly_constraint>")): Changed to ...
6999 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7000 "<round_saeonly_scalar_constraint>")): ... this.
7001 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7002 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7003 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7004 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7005 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7007 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7009 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7011 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7012 Remove all values except native, 8540 and 8548.
7014 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7016 * config/i386/i386.c (ix86_output_function_return): Pass
7017 INVALID_REGNUM, instead of -1, as invalid register number to
7018 indirect_thunk_name and output_indirect_thunk.
7020 2018-02-02 Julia Koval <julia.koval@intel.com>
7022 * config.gcc: Add -march=icelake.
7023 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7024 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7025 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7026 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7027 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7028 (processor_target_table): Add icelake.
7029 (ix86_option_override_internal): Handle new PTAs.
7030 (get_builtin_code_for_version): Handle icelake.
7031 (M_INTEL_COREI7_ICELAKE): New.
7032 (fold_builtin_cpu): Handle icelake.
7033 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7034 * doc/invoke.texi: Add -march=icelake.
7036 2018-02-02 Julia Koval <julia.koval@intel.com>
7038 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7039 to wide_int_bitmask.
7040 * wide-int-bitmask.h: New.
7042 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7045 * config/i386/i386.md: Replace Pmode with word_mode in
7046 builtin_setjmp_setup and builtin_longjmp to support x32.
7048 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7052 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7054 (rs6000_supported_cpu_names): New static variable.
7055 (linux_cpu_translation_table): Likewise.
7056 (elf_platform) <cpu>: Define new static variable and use it.
7057 Translate kernel AT_PLATFORM name to canonical name if needed.
7058 Error if platform name is unknown.
7060 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7063 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7065 2018-02-01 Jeff Law <law@redhat.com>
7068 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7069 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7070 the scratch if RELEASE_VIA_POP is false.
7071 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7072 If we have to save a temporary register, decrement SIZE appropriately.
7073 Pass new arguments to release_scratch_register_on_entry.
7074 (ix86_adjust_stack_and_probe): Likewise.
7075 (ix86_emit_probe_stack_range): Pass new arguments to
7076 release_scratch_register_on_entry.
7078 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7080 PR rtl-optimization/84157
7081 * combine.c (change_zero_ext): Use REG_P predicate in
7082 front of HARD_REGISTER_P predicate.
7084 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7086 * config/avr/avr.c (avr_option_override): Move disabling of
7087 -fdelete-null-pointer-checks to...
7088 * common/config/avr/avr-common.c (avr_option_optimization_table):
7091 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7093 PR tree-optimization/81635
7094 * tree-data-ref.c (split_constant_offset_1): For types that
7095 wrap on overflow, try to use range info to prove that wrapping
7098 2018-02-01 Renlin Li <renlin.li@arm.com>
7101 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7103 (aarch64_register_move_cost): Likewise.
7104 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7106 (REG_CLASS_NAMES): Likewise.
7107 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7108 TAILCALL_ADDR_REGS. Remove IP registers.
7109 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7111 2018-02-01 Richard Biener <rguenther@suse.de>
7113 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7114 for specifying RPO order and allow NULL for that.
7115 * domwalk.c (dom_walker::dom_walker): Likewise.
7116 (dom_walker::walk): Handle NULL RPO order.
7117 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7119 (rewrite_update_dom_walker): Likewise.
7120 (mark_def_dom_walker): Likewise.
7122 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7124 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7125 (aarch64_maybe_expand_sve_subreg_move): Declare.
7126 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7127 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7129 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7130 that are semantically a reverse operation.
7131 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7132 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7133 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7135 (aarch64_can_change_mode_class): For big-endian, forbid changes
7136 between two SVE modes if they have different element sizes.
7138 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7140 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7141 the TImode handling for big-endian targets.
7143 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7145 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7146 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7148 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7149 Remove BSWAP handing for big-endian targets and use the form of
7150 LD1RQ appropariate for the mode.
7152 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7154 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7155 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7158 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7161 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7162 check for operands that need to go through aarch64_sve_reload_be.
7164 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7166 PR tree-optimization/81661
7167 PR tree-optimization/84117
7168 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7169 * tree-eh.c: Include gimplify.h.
7170 (find_trapping_overflow, replace_trapping_overflow,
7171 rewrite_to_non_trapping_overflow): New functions.
7172 * tree-vect-loop.c: Include tree-eh.h.
7173 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7174 * tree-data-ref.c: Include tree-eh.h.
7175 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7177 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7179 PR rtl-optimization/84123
7180 * combine.c (change_zero_ext): Check if hard register satisfies
7181 can_change_dest_mode before calling gen_lowpart_SUBREG.
7183 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7186 * ira.c (ira_init_register_move_cost): Remove assert.
7188 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7190 PR rtl-optimization/84071
7191 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7192 * doc/tm.texi: Regenerate.
7194 2018-01-31 Richard Biener <rguenther@suse.de>
7196 PR tree-optimization/84132
7197 * tree-data-ref.c (analyze_miv_subscript): Properly
7198 check whether evolution_function_is_affine_multivariate_p
7199 before calling gcd_of_steps_may_divide_p.
7201 2018-01-31 Julia Koval <julia.koval@intel.com>
7204 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7205 * config/i386/i386.md (rdpid_rex64) New.
7206 (rdpid): Make 32bit only.
7208 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7211 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7212 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7214 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7217 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7219 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7221 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7223 PR rtl-optimization/84071
7224 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7225 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7227 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7229 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7230 (arc_attribute_table): Add 'aux' attribute.
7231 (arc_in_small_data_p): Consider aux like variables.
7232 (arc_is_aux_reg_p): New function.
7233 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7234 (arc_get_aux_arg): New function.
7235 (prepare_move_operands): Handle aux-register access.
7236 (arc_handle_aux_attribute): New function.
7237 * doc/extend.texi (ARC Variable attributes): Add subsection.
7239 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7241 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7242 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7243 (arc_attribute_table): Add 'uncached' attribute.
7244 (arc_print_operand): Print '.di' flag for uncached memory
7246 (arc_in_small_data_p): Do not consider for small data the uncached
7248 (arc_is_uncached_mem_p): New function.
7249 * config/arc/predicates.md (compact_store_memory_operand): Check
7250 for uncached memory accesses.
7251 (nonvol_nonimm_operand): Likewise.
7252 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7254 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7257 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7258 falign-loops=): Add Optimization flag.
7260 2018-01-30 Jeff Law <law@redhat.com>
7263 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7264 INT_REGISTERS_SAVED. Check it prior to calling
7265 get_scratch_register_on_entry.
7266 (ix86_adjust_stack_and_probe): Similarly.
7267 (ix86_emit_probe_stack_range): Similarly.
7268 (ix86_expand_prologue): Corresponding changes.
7270 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7273 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7274 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7276 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7279 * lra-constraints.c (curr_insn_transform): Process AND in the
7282 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7284 PR rtl-optimization/83986
7285 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7286 dependence against last_pending_memory_flush in addition to
7289 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7291 PR tree-optimization/81611
7292 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7295 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7298 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7301 2018-01-30 Richard Biener <rguenther@suse.de>
7302 Jakub Jelinek <jakub@redhat.com>
7304 PR tree-optimization/84111
7305 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7306 inner loops added during recursion, as they don't have up-to-date
7309 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7312 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7313 (can_inline_edge_by_limits_p): ... here.
7314 (can_early_inline_edge_p, check_callers,
7315 update_caller_keys, update_callee_keys, recursive_inlining,
7316 add_new_edges_to_heap, speculation_useful_p,
7317 inline_small_functions,
7318 inline_small_functions, flatten_function,
7319 inline_to_all_callers_1): Update.
7321 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7323 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7326 2018-01-30 Richard Biener <rguenther@suse.de>
7328 PR tree-optimization/83008
7329 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7330 invariant and constant vector uses in stmts when they need
7333 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7336 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7337 * configure: Regenerate.
7339 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7341 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7343 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7344 Use gen_rtx_REG rather than gen_lowpart.
7346 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7348 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7349 rather than 0 when creating partial subregs.
7351 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7353 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7356 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7359 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7360 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7361 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7362 flags. This restores the settings used before the 2017-07-24.
7363 Turning off pre increment/decrement/modify allows IVOPTS to
7364 optimize DF/SF loops where the index is an int.
7366 2018-01-29 Richard Biener <rguenther@suse.de>
7367 Kelvin Nilsen <kelvin@gcc.gnu.org>
7370 * tree-vect-stmts.c (vectorizable_call): Don't call
7371 targetm.vectorize_builtin_md_vectorized_function if callee is
7374 2018-01-22 Carl Love <cel@us.ibm.com>
7376 * doc/extend.tex: Fix typo in second arg in
7377 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7379 2018-01-29 Richard Biener <rguenther@suse.de>
7381 PR tree-optimization/84086
7382 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7383 (flush_ssaname_freelist): When SSA names were released reset
7384 the SCEV hash table.
7386 2018-01-29 Richard Biener <rguenther@suse.de>
7388 PR tree-optimization/84057
7389 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7390 removed paths when removing edges.
7392 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7394 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7395 -mfunction-return=@var{choice}.
7397 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7400 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7402 (layout::print_source_line): Likewise.
7403 (test_get_line_width_without_trailing_whitespace): Add test cases.
7405 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7408 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7411 2018-01-26 Jim Wilson <jimw@sifive.com>
7413 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7415 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7418 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7420 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7421 and CMP + SUB-immediate -> SUBS.
7423 2018-01-26 Martin Sebor <msebor@redhat.com>
7425 PR tree-optimization/83896
7426 * tree-ssa-strlen.c (get_string_len): Rename...
7427 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7428 Avoid assuming length is constant.
7429 (handle_char_store): Use HOST_WIDE_INT for string length.
7431 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7434 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7435 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7437 2018-01-26 Richard Biener <rguenther@suse.de>
7439 PR rtl-optimization/84003
7440 * dse.c (record_store): Only record redundant stores when
7441 the earlier store aliases at least all accesses the later one does.
7443 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7445 PR rtl-optimization/83985
7446 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7447 REG_CFA_RESTORE insns.
7448 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7449 REG_CFA_RESTORE insns here.
7452 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7453 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7455 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7457 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7458 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7459 (arc_init): Likewise.
7460 (arc_override_options): Likewise.
7461 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7463 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7465 * config/arc/arc.h (TARGET_DBNZ): Define.
7466 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7467 properly set the tune attribute.
7468 (dbnz): Use TARGET_DBNZ guard.
7469 * config/arc/arc.opt (mtune): Add core3 option.
7471 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7473 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7474 recognize new pic like addresses.
7475 (arc_delegitimize_address): Clean up.
7477 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7479 * config/arc/arc-arches.def: Option mrf16 valid for all
7481 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7482 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7483 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7484 * config/arc/arc-tables.opt: Regenerate.
7485 * config/arc/arc.c (arc_conditional_register_usage): Handle
7486 reduced register file case.
7487 (arc_file_start): Set must have build attributes.
7488 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7490 * config/arc/arc.opt (mrf16): Add new option.
7491 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7492 * config/arc/genmultilib.awk: Handle new mrf16 option.
7493 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7494 * config/arc/t-multilib: Regenerate.
7495 * doc/invoke.texi (ARC Options): Document mrf16 option.
7497 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7499 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7500 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7501 (arc_attribute_table): Add 'secure_call' attribute.
7502 (arc_print_operand): Print secure call operand.
7503 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7505 (arc_is_secure_call_p): New function. * config/arc/arc.md
7506 (call_i): Add support for sjli instruction.
7507 (call_value_i): Likewise.
7508 * config/arc/constraints.md (Csc): New constraint.
7510 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7511 John Eric Martin <John.Martin@emmicro-us.com>
7513 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7514 * config/arc/arc.c (_arc_jli_section): New struct.
7515 (arc_jli_section): New type.
7516 (rc_jli_sections): New static variable.
7517 (arc_handle_jli_attribute): New function.
7518 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7519 (arc_file_end): New function.
7520 (TARGET_ASM_FILE_END): Define.
7521 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7522 (arc_add_jli_section): New function.
7523 (jli_call_scan): Likewise.
7524 (arc_reorg): Call jli_call_scan.
7525 (arc_output_addsi): Remove 'S' from printing asm operand.
7526 (arc_is_jli_call_p): New function.
7527 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7529 (movhi_insn): Likewise.
7530 (movsi_insn): Likewise.
7531 (movsi_set_cc_insn): Likewise.
7532 (loadqi_update): Likewise.
7533 (load_zeroextendqisi_update): Likewise.
7534 (load_signextendqisi_update): Likewise.
7535 (loadhi_update): Likewise.
7536 (load_zeroextendhisi_update): Likewise.
7537 (load_signextendhisi_update): Likewise.
7538 (loadsi_update): Likewise.
7539 (loadsf_update): Likewise.
7540 (movsicc_insn): Likewise.
7541 (bset_insn): Likewise.
7542 (bxor_insn): Likewise.
7543 (bclr_insn): Likewise.
7544 (bmsk_insn): Likewise.
7545 (bicsi3_insn): Likewise.
7546 (cmpsi_cc_c_insn): Likewise.
7547 (movsi_ne): Likewise.
7548 (movsi_cond_exec): Likewise.
7549 (clrsbsi2): Likewise.
7560 (call_i): Remove 'S' asm letter, add jli instruction.
7561 (call_value_i): Likewise.
7562 * config/arc/arc.op (mjli-always): New option.
7563 * config/arc/constraints.md (Cji): New constraint.
7564 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7566 (subsf3_fpx): Likewise.
7567 (mulsf3_fpx): Likewise.
7568 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7570 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7572 * doc/invoke.texi (ARC): Document mjli-always option.
7574 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7576 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7577 avoid addition with 0 and use incw and decw where possible.
7579 2018-01-26 Richard Biener <rguenther@suse.de>
7581 PR tree-optimization/81082
7582 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7583 association if it requires casting to unsigned.
7584 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7585 from fold_plusminus_mult_expr to catch important cases late when
7586 range info is available.
7588 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7590 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7591 * configure.ac (hidden_linkonce): New test.
7592 * configure: Regenerate.
7593 * config.in: Regenerate.
7595 2018-01-26 Julia Koval <julia.koval@intel.com>
7597 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7598 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7599 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7600 _mm_mask_bitshuffle_epi64_mask): Fix type.
7601 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7602 USI_FTYPE_V4DI_V4DI_USI): Remove.
7603 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7604 __builtin_ia32_vpshufbitqmb256_mask,
7605 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7606 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7607 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7609 2018-01-26 Alan Modra <amodra@gmail.com>
7612 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7613 UNSPEC_VBPERMQ. Sort other unspecs.
7615 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7617 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7619 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7622 * predict.c (drop_profile): Do not push/pop cfun; update also
7624 (handle_missing_profiles): Fix logic looking for zero profiles.
7626 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7629 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7630 on functions with #pragma omp declare simd or functions with simd
7632 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7633 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7634 Remove trailing \n from warning_at calls.
7636 2018-01-25 Tom de Vries <tom@codesourcery.com>
7639 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7640 for neutered workers.
7642 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7645 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7646 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7648 2018-01-24 Jeff Law <law@redhat.com>
7651 * i386.c (get_probe_interval): Move to earlier point.
7652 (ix86_compute_frame_layout): If -fstack-clash-protection and
7653 the frame is larger than the probe interval, then use pushes
7654 to save registers rather than reg->mem moves.
7655 (ix86_expand_prologue): Remove conditional for int_registers_saved
7658 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7661 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7662 min/max for never referenced object.
7664 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7667 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7669 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7670 attributes from DECL_ATTRIBUTES (decl) without affecting
7671 DECL_ATTRIBUTES (current_function_decl).
7672 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7673 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7675 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7677 PR tree-optimization/83979
7678 * fold-const.c (fold_comparison): Use constant_boolean_node
7679 instead of boolean_{true,false}_node.
7681 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7683 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7686 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7688 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7689 Simplify the clause that sets the length attribute.
7690 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7691 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7692 clause that sets the length attribute.
7693 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7695 2018-01-24 Tom de Vries <tom@codesourcery.com>
7698 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7699 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7700 Add strict parameter.
7701 (prevent_branch_around_nothing): Insert dummy insn between branch to
7702 label and label with no ptx insn inbetween.
7703 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7705 2018-01-24 Tom de Vries <tom@codesourcery.com>
7708 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7709 for neutered threads in warp.
7710 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7712 2018-01-24 Richard Biener <rguenther@suse.de>
7714 PR tree-optimization/83176
7715 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7718 2018-01-24 Richard Biener <rguenther@suse.de>
7720 PR tree-optimization/82819
7721 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7722 code generating pluses that are no-ops in the target precision.
7724 2018-01-24 Richard Biener <rguenther@suse.de>
7727 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7729 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7731 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7732 to merge probabilities.
7733 * predict.c (probably_never_executed): Also mark as cold functions
7734 with global 0 profile and guessed local profile.
7735 * profile-count.c (profile_probability::combine_with_count): New
7737 * profile-count.h (profile_probability::operator*,
7738 profile_probability::operator*=, profile_probability::operator/,
7739 profile_probability::operator/=): Reduce precision to adjusted
7740 and set value to guessed on contradictory divisions.
7741 (profile_probability::combine_with_freq): Remove.
7742 (profile_probability::combine_wiht_count): Declare.
7743 (profile_count::force_nonzero):: Set to adjusted.
7744 (profile_count::probability_in):: Set quality to adjusted.
7745 * tree-ssa-tail-merge.c (replace_block_by): Use
7748 2018-01-23 Andrew Waterman <andrew@sifive.com>
7749 Jim Wilson <jimw@sifive.com>
7751 * config/riscv/riscv.c (riscv_stack_boundary): New.
7752 (riscv_option_override): Set riscv_stack_boundary. Handle
7753 riscv_preferred_stack_boundary_arg.
7754 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7755 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7756 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7757 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7758 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7759 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7761 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7764 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7765 of struct ix86_frame.
7766 (ix86_expand_epilogue): Likewise. Add a local variable for
7767 the reg_save_offset field in struct ix86_frame.
7769 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7771 PR tree-optimization/82604
7772 * tree-loop-distribution.c (enum partition_kind): New enum item
7773 PKIND_PARTIAL_MEMSET.
7774 (partition_builtin_p): Support above new enum item.
7775 (generate_code_for_partition): Ditto.
7776 (compute_access_range): Differentiate cases that equality can be
7777 proven at all loops, the innermost loops or no loops.
7778 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7779 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7780 (finalize_partitions, distribute_loop): Don't fuse partition of
7781 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7782 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7785 2018-01-23 Martin Liska <mliska@suse.cz>
7787 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7788 order to ignore the predictor.
7789 (PRED_POLYMORPHIC_CALL): Likewise.
7790 (PRED_RECURSIVE_CALL): Likewise.
7792 2018-01-23 Martin Liska <mliska@suse.cz>
7794 * tree-profile.c (tree_profiling): Print function header to
7795 aware reader which function we are working on.
7796 * value-prof.c (gimple_find_values_to_profile): Do not print
7797 not interesting value histograms.
7799 2018-01-23 Martin Liska <mliska@suse.cz>
7801 * profile-count.h (enum profile_quality): Add
7802 profile_uninitialized as the first value. Do not number values
7803 as they are zero based.
7804 (profile_count::verify): Update sanity check.
7805 (profile_probability::verify): Likewise.
7807 2018-01-23 Nathan Sidwell <nathan@acm.org>
7809 * doc/invoke.texi (ffor-scope): Deprecate.
7811 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7813 PR tree-optimization/83510
7814 * domwalk.c (set_all_edges_as_executable): New function.
7815 (dom_walker::dom_walker): Convert bool param
7816 "skip_unreachable_blocks" to enum reachability. Move setup of
7817 edge flags to set_all_edges_as_executable and only do it when
7818 reachability is REACHABLE_BLOCKS.
7819 * domwalk.h (enum dom_walker::reachability): New enum.
7820 (dom_walker::dom_walker): Convert bool param
7821 "skip_unreachable_blocks" to enum reachability.
7822 (set_all_edges_as_executable): New decl.
7823 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7824 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7826 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7827 but converting true to REACHABLE_BLOCKS.
7828 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7830 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7831 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7832 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7834 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7835 if check_all_array_refs will be called.
7837 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7839 * tree.c (selftest::test_location_wrappers): Add more test
7842 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7844 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7845 (selftest::test_bit_in_range): Likewise.
7847 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7850 * doc/sourcebuild.texi (vect_float): Say that the selector
7851 only describes the situation when -funsafe-math-optimizations is on.
7852 (vect_float_strict): Document.
7854 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7856 PR tree-optimization/83965
7857 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7858 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7859 instead of checking only for a reduction.
7860 (vect_recog_widen_sum_pattern): Likewise.
7862 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7864 * predict.c (probably_never_executed): Only use precise profile info.
7865 (compute_function_frequency): Skip after inlining hack since we now
7866 have quality checking.
7868 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7870 * profile-count.h (profile_probability::very_unlikely,
7871 profile_probability::unlikely, profile_probability::even): Set
7872 precision to guessed.
7874 2018-01-23 Richard Biener <rguenther@suse.de>
7876 PR tree-optimization/83963
7877 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7878 Properly terminate dominator walk when crossing the exit edge not
7879 when visiting its source block.
7881 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7884 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7885 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7887 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7889 PR tree-optimization/83957
7890 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7891 semicolon after for body surrounded by braces.
7893 PR tree-optimization/83081
7894 * profile-count.h (profile_probability::split): New method.
7895 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7896 Use profile_probability::split.
7897 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7898 when splitting a single conditional jump into 2.
7900 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7902 PR tree-optimization/69452
7903 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7906 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7908 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7909 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7910 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7912 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7914 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7915 * config/rl78/rl78.md (movdi): New define_expand.
7916 * config/rl78/rl78.c (rl78_split_movdi): New function.
7918 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7921 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7923 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7924 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7925 128-bit to produce an UNSPEC move to get the double word with the
7926 signbit and then a shift directly to do signbit.
7927 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7928 implementation with a new version that just does either a direct
7929 move or a regular move. Move memory interface to separate insns.
7930 Move insns so they are next to the expander.
7931 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7932 with signbit move. Split big and little endian case.
7933 (signbit<mode>2_dm_mem_le): Likewise.
7934 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7935 (signbit<mode>2_dm2): Likewise.
7937 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7939 * config/rl78/rl78.md (anddi3): New define_expand.
7941 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7943 * config/rl78/rl78.md (umindi3): New define_expand.
7945 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7947 * config/rl78/rl78.md (smindi3): New define_expand.
7949 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7951 * config/rl78/rl78.md (smaxdi3): New define_expand.
7953 2018-01-22 Carl Love <cel@us.ibm.com>
7955 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7956 LVX_V1TI): Add macro expansion.
7957 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7958 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7959 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7960 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7961 Change check to determine if the instruction is a byte reversing
7962 entry. Fix typo in comment.
7963 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7964 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7965 Add def_builtin calls for new builtins.
7966 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7967 Add define_insn expansion.
7969 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7971 * config/rl78/rl78.md (umaxdi3): New define_expand.
7973 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7975 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7976 for non-QImode registers.
7978 2018-01-22 Richard Biener <rguenther@suse.de>
7980 PR tree-optimization/83963
7981 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7982 including the loop exit block.
7983 (scop_detection::merge_sese): Likewise.
7984 (scop_detection::add_scop): Do it here instead.
7986 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7988 * doc/sourcebuild.texi (arm_softfloat): Document.
7990 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7993 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7994 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7995 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7997 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7998 David Edelsohn <dje.gcc@gmail.com>
8001 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8002 Change "crset eq" to "crset 2".
8003 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8004 (*call_indirect_aix<mode>_nospec): Likewise.
8005 (*call_value_indirect_aix<mode>_nospec): Likewise.
8006 (*call_indirect_elfv2<mode>_nospec): Likewise.
8007 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8008 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8009 change assembly output from . to $.
8010 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8011 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8012 (*tablejump<mode>_internal1_nospec): Likewise.
8014 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8017 * config/sh/sh_optimize_sett_clrt.cc:
8018 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8020 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8022 PR tree-optimization/83940
8023 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8024 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8025 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8026 use it to pass back the definition type.
8027 (vect_check_store_rhs): Likewise rhs_dt_out.
8028 (vect_build_gather_load_calls): Add a mask_dt argument and use
8029 it instead of a call to vect_is_simple_use.
8030 (vectorizable_store): Update calls to vect_check_load_store_mask
8031 and vect_check_store_rhs. Use the dt returned by the latter instead
8032 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8033 instead of calls to vect_is_simple_use. Pass the scalar rather
8034 than the vector operand to vect_is_simple_use when handling
8035 second and subsequent copies of an rhs value.
8036 (vectorizable_load): Update calls to vect_check_load_store_mask
8037 and vect_build_gather_load_calls. Use the cached mask_dt and
8038 gs_info.offset_dt instead of calls to vect_is_simple_use.
8040 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8043 * tree-emutls.c: Include gimplify.h.
8044 (lower_emutls_2): New function.
8045 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8046 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8047 it before further processing.
8050 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8051 UINTVAL (trueop1) instead of INTVAL (op1).
8053 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8057 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8058 INCOMING_FRAME_SP_OFFSET if not defined.
8059 (scan_trace): Add ENTRY argument. If true and
8060 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8061 emit a note to adjust the CFA offset.
8062 (create_cfi_notes): Adjust scan_trace callers.
8063 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8064 INCOMING_FRAME_SP_OFFSET in the CIE.
8065 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8066 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8068 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8069 * doc/tm.texi: Regenerated.
8071 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8073 PR rtl-optimization/83147
8074 * lra-constraints.c (remove_inheritance_pseudos): Use
8075 lra_substitute_pseudo_within_insn.
8077 2018-01-19 Tom de Vries <tom@codesourcery.com>
8078 Cesar Philippidis <cesar@codesourcery.com>
8081 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8083 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8086 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8087 spaces for function labels.
8089 2018-01-19 Martin Liska <mliska@suse.cz>
8091 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8092 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8093 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8094 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8095 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8096 (PRED_CONST_RETURN): Change from 69 to 65.
8097 (PRED_NULL_RETURN): Change from 91 to 71.
8098 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8099 (PRED_LOOP_GUARD): Change from 66 to 73.
8101 2018-01-19 Martin Liska <mliska@suse.cz>
8103 * predict.c (predict_insn_def): Add new assert.
8104 (struct branch_predictor): Change type to signed integer.
8105 (test_prediction_value_range): Amend test to cover
8107 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8108 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8109 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8110 (PRED_LOOP_IV_COMPARE): Likewise.
8111 * predict.h (PROB_UNINITIALIZED): Define new constant.
8113 2018-01-19 Martin Liska <mliska@suse.cz>
8115 * predict.c (dump_prediction): Add new format for
8116 analyze_brprob.py script which is enabled with -details
8118 * profile-count.h (precise_p): New function.
8120 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8122 PR tree-optimization/83922
8123 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8124 there are no statements that need masking.
8125 (vect_active_double_reduction_p): New function.
8126 (vect_analyze_loop_operations): Use it when handling phis that
8127 are not in the loop header.
8129 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8131 PR tree-optimization/83914
8132 * tree-vect-loop.c (vectorizable_induction): Don't convert
8133 init_expr or apply the peeling adjustment for inductions
8134 that are nested within the vectorized loop.
8136 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8138 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8141 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8145 * function.h (gimplify_parameters): Add gimple_seq * argument.
8146 * function.c: Include gimple.h and options.h.
8147 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8148 for the added local temporaries if needed.
8149 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8150 if there are any parameter cleanups, wrap whole body into a
8151 try/finally with the cleanups.
8153 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8156 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8157 Use GET_MODE_CLASS for scalar floating point.
8159 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8163 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8164 Fix call of call_cgraph_insertion_hooks.
8166 2018-01-18 Martin Sebor <msebor@redhat.com>
8168 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8170 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8173 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8176 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8179 * common.opt: (-ffile-prefix-map): New option.
8180 * opts.c (common_handle_option): Defer it.
8181 * opts-global.c (handle_common_deferred_options): Handle it.
8182 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8183 * file-prefix-map.h: New file.
8184 (remap_debug_filename, add_debug_prefix_map): ...here.
8185 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8186 * final.c (debug_prefix_map, add_debug_prefix_map
8187 remap_debug_filename): Move to...
8188 * file-prefix-map.c: New file.
8189 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8190 generalize, get rid of alloca(), use strrchr() instead of strchr().
8191 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8192 Implement in terms of add_prefix_map().
8193 (remap_macro_filename, remap_debug_filename): Implement in term of
8195 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8196 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8197 * dbxout.c: Include file-prefix-map.h.
8198 * varasm.c: Likewise.
8199 * vmsdbgout.c: Likewise.
8200 * xcoffout.c: Likewise.
8201 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8202 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8203 * doc/invoke.texi (-ffile-prefix-map): Document.
8204 (-fdebug-prefix-map): Update description.
8206 2018-01-18 Martin Liska <mliska@suse.cz>
8208 * config/i386/i386.c (indirect_thunk_name): Document that also
8210 (output_indirect_thunk): Document why both instructions
8211 (pause and lfence) are generated.
8213 2018-01-18 Richard Biener <rguenther@suse.de>
8215 PR tree-optimization/83887
8216 * graphite-scop-detection.c
8217 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8218 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8219 (scop_detection::merge_sese): Re-implement with a flood-fill
8220 algorithm that properly finds a SESE region if it exists.
8222 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8225 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8226 pointer_diff optimizations use view_convert instead of convert.
8228 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8230 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8231 Generate different code for -mno-speculate-indirect-jumps.
8232 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8233 (*call_indirect_aix<mode>): Disable for
8234 -mno-speculate-indirect-jumps.
8235 (*call_indirect_aix<mode>_nospec): New define_insn.
8236 (*call_value_indirect_aix<mode>): Disable for
8237 -mno-speculate-indirect-jumps.
8238 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8239 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8240 -mno-speculate-indirect-jumps.
8241 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8243 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8245 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8246 long double type, set the flags for noting the default long double
8247 type, even if we don't pass or return a long double type.
8249 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8252 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8255 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8257 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8258 support for merge[hl].
8259 (fold_mergehl_helper): New helper function.
8260 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8261 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8262 (altivec_vmrglw_direct): Add xxmrglw insn.
8264 2018-01-17 Andrew Waterman <andrew@sifive.com>
8266 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8267 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8269 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8272 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8273 call the lto_location_cache before reading the
8274 DECL_SOURCE_LOCATION of the types.
8276 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8277 Richard Sandiford <richard.sandiford@linaro.org>
8279 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8280 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8281 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8282 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8283 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8285 * config/aarch64/constraints.md (aarch64_movti_operand):
8287 * config/aarch64/predicates.md (Uti): Add new constraint.
8289 2018-01-17 Carl Love <cel@us.ibm.com>
8291 * config/rs6000/vsx.md (define_expand xl_len_r,
8292 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8293 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8295 (define_expand, define_insn): Move the shift left from the
8296 define_insn to the define_expand for lxvl and stxvl instructions.
8297 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8298 and XL_LEN_R definitions to PURE.
8300 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8302 * config/i386/i386.c (indirect_thunk_name): Declare regno
8303 as unsigned int. Compare regno with INVALID_REGNUM.
8304 (output_indirect_thunk): Ditto.
8305 (output_indirect_thunk_function): Ditto.
8306 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8307 in the call to output_indirect_thunk_function.
8309 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8312 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8313 rather than the size of inner_type to determine the stack slot size
8314 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8316 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8319 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8322 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8324 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8325 endian Linux systems to optionally enable multilibs for selecting
8326 the long double type if the user configured an explicit type.
8327 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8328 have no long double multilibs if not defined.
8329 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8330 warn if the user used -mabi={ieee,ibm}longdouble and we built
8331 multilibs for long double.
8332 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8333 appropriate multilib option.
8334 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8336 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8337 for building long double multilibs.
8338 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8340 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8342 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8345 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8347 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8350 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8353 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8356 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8358 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8359 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8361 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8363 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8364 different rtl trees depending on TARGET_64BIT.
8365 (rs6000_gen_lvx): Likewise.
8367 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8369 * config/visium/visium.md (nop): Tweak comment.
8370 (hazard_nop): Likewise.
8372 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8374 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8375 -mspeculate-indirect-jumps.
8376 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8377 for -mno-speculate-indirect-jumps.
8378 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8379 (*call_value_indirect_elfv2<mode>): Disable for
8380 -mno-speculate-indirect-jumps.
8381 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8382 (indirect_jump): Emit different RTL for
8383 -mno-speculate-indirect-jumps.
8384 (*indirect_jump<mode>): Disable for
8385 -mno-speculate-indirect-jumps.
8386 (*indirect_jump<mode>_nospec): New define_insn.
8387 (tablejump): Emit different RTL for
8388 -mno-speculate-indirect-jumps.
8389 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8390 (tablejumpsi_nospec): New define_expand.
8391 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8392 (tablejumpdi_nospec): New define_expand.
8393 (*tablejump<mode>_internal1): Disable for
8394 -mno-speculate-indirect-jumps.
8395 (*tablejump<mode>_internal1_nospec): New define_insn.
8396 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8399 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8401 * caller-save.c (insert_save): Drop unnecessary parameter. All
8404 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8405 Richard Biener <rguenth@suse.de>
8408 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8409 return early, inline manually is_gimple_sizepos. Make sure if we
8410 call gimplify_expr we don't end up with a gimple constant.
8411 * tree.c (variably_modified_type_p): Don't return true for
8412 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8413 * gimplify.h (is_gimple_sizepos): Remove.
8415 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8417 PR tree-optimization/83857
8418 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8419 vectorizable_live_operation for pure SLP statements.
8420 (vectorizable_live_operation): Handle PHIs.
8422 2018-01-16 Richard Biener <rguenther@suse.de>
8424 PR tree-optimization/83867
8425 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8426 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8428 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8431 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8432 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8433 If off is not INTEGER_CST, issue a may not be aligned warning
8434 rather than isn't aligned. Use isn%'t rather than isn't.
8435 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8437 <case MULT_EXPR>: Improve the case when bottom and one of the
8438 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8439 operand, in that case check if the other operand is multiple of
8440 bottom divided by the INTEGER_CST operand.
8442 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8445 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8446 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8447 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8448 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8449 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8450 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8451 (pa_function_arg_size): New function.
8453 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8455 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8456 in a separate statement.
8458 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8460 PR tree-optimization/83847
8461 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8462 group gathers and scatters.
8464 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8466 PR rtl-optimization/86620
8467 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8469 PR rtl-optimization/83213
8470 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8471 to last if both are JUMP_INSNs.
8473 PR tree-optimization/83843
8474 * gimple-ssa-store-merging.c
8475 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8476 store_immediate_info for bswap/nop orig_stores.
8478 2018-01-15 Andrew Waterman <andrew@sifive.com>
8480 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8482 <UDIV>: Increase cost if !TARGET_DIV.
8484 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8486 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8487 (define_attr "cr_logical_3op"): New.
8488 (cceq_ior_compare): Adjust.
8489 (cceq_ior_compare_complement): Adjust.
8490 (*cceq_rev_compare): Adjust.
8491 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8492 (is_cracked_insn): Adjust.
8493 (insn_must_be_first_in_group): Adjust.
8494 * config/rs6000/40x.md: Adjust.
8495 * config/rs6000/440.md: Adjust.
8496 * config/rs6000/476.md: Adjust.
8497 * config/rs6000/601.md: Adjust.
8498 * config/rs6000/603.md: Adjust.
8499 * config/rs6000/6xx.md: Adjust.
8500 * config/rs6000/7450.md: Adjust.
8501 * config/rs6000/7xx.md: Adjust.
8502 * config/rs6000/8540.md: Adjust.
8503 * config/rs6000/cell.md: Adjust.
8504 * config/rs6000/e300c2c3.md: Adjust.
8505 * config/rs6000/e500mc.md: Adjust.
8506 * config/rs6000/e500mc64.md: Adjust.
8507 * config/rs6000/e5500.md: Adjust.
8508 * config/rs6000/e6500.md: Adjust.
8509 * config/rs6000/mpc.md: Adjust.
8510 * config/rs6000/power4.md: Adjust.
8511 * config/rs6000/power5.md: Adjust.
8512 * config/rs6000/power6.md: Adjust.
8513 * config/rs6000/power7.md: Adjust.
8514 * config/rs6000/power8.md: Adjust.
8515 * config/rs6000/power9.md: Adjust.
8516 * config/rs6000/rs64.md: Adjust.
8517 * config/rs6000/titan.md: Adjust.
8519 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8521 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8522 ix86_indirect_branch_register logic.
8524 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8526 * config/i386/constraints.md (Bs): Update
8527 ix86_indirect_branch_register check. Don't check
8528 ix86_indirect_branch_register with GOT_memory_operand.
8530 * config/i386/predicates.md (GOT_memory_operand): Don't check
8531 ix86_indirect_branch_register here.
8532 (GOT32_symbol_operand): Likewise.
8534 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8536 * config/i386/predicates.md (constant_call_address_operand):
8537 Rewrite ix86_indirect_branch_register logic.
8538 (sibcall_insn_operand): Likewise.
8540 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8542 * config/i386/constraints.md (Bs): Replace
8543 ix86_indirect_branch_thunk_register with
8544 ix86_indirect_branch_register.
8546 * config/i386/i386.md (indirect_jump): Likewise.
8547 (tablejump): Likewise.
8548 (*sibcall_memory): Likewise.
8549 (*sibcall_value_memory): Likewise.
8550 Peepholes of indirect call and jump via memory: Likewise.
8551 * config/i386/i386.opt: Likewise.
8552 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8553 (GOT_memory_operand): Likewise.
8554 (call_insn_operand): Likewise.
8555 (sibcall_insn_operand): Likewise.
8556 (GOT32_symbol_operand): Likewise.
8558 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8561 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8562 type rather than type addr's type points to.
8563 (expand_omp_atomic_mutex): Likewise.
8564 (expand_omp_atomic): Likewise.
8566 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8569 * config/i386/i386.c (output_indirect_thunk_function): Use
8570 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8571 for __x86_return_thunk.
8573 2018-01-15 Richard Biener <rguenther@suse.de>
8576 * expmed.c (extract_bit_field_1): Fix typo.
8578 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8581 * config/arm/iterators.md (VF): New mode iterator.
8582 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8583 Remove integer-related logic from pattern.
8584 (neon_vabd<mode>_3): Likewise.
8586 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8589 * common.opt (fstrict-overflow): No longer an alias.
8590 (fwrapv-pointer): New option.
8591 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8592 also for pointer types based on flag_wrapv_pointer.
8593 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8594 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8595 opts->x_flag_wrapv got set.
8596 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8597 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8598 POINTER_TYPE_OVERFLOW_UNDEFINED.
8599 * match.pd: Likewise in address comparison pattern.
8600 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8602 2018-01-15 Richard Biener <rguenther@suse.de>
8605 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8606 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8607 Reset type names to their identifier if their TYPE_DECL doesn't
8608 have linkage (and thus is used for ODR and devirt).
8609 (save_debug_info_for_decl): Remove.
8610 (save_debug_info_for_type): Likewise.
8611 (add_tree_to_fld_list): Adjust.
8612 * tree-pretty-print.c (dump_generic_node): Make dumping of
8613 type names more robust.
8615 2018-01-15 Richard Biener <rguenther@suse.de>
8617 * BASE-VER: Bump to 8.0.1.
8619 2018-01-14 Martin Sebor <msebor@redhat.com>
8622 * builtins.c (check_access): Avoid warning when the no-warning bit
8625 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8627 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8628 * ira-color (allocno_hard_regs_compare): Likewise.
8630 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8633 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8634 Use .pushsection/.popsection.
8636 2018-01-14 Martin Sebor <msebor@redhat.com>
8639 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8641 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8643 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8644 entry from extra_headers.
8645 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8646 extra_headers, make the list bitwise identical to the i?86-*-* one.
8648 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8650 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8651 -mcmodel=large with -mindirect-branch=thunk,
8652 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8653 -mfunction-return=thunk-extern.
8654 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8655 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8656 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8658 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8660 * config/i386/i386.c (print_reg): Print the name of the full
8661 integer register without '%'.
8662 (ix86_print_operand): Handle 'V'.
8663 * doc/extend.texi: Document 'V' modifier.
8665 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8667 * config/i386/constraints.md (Bs): Disallow memory operand for
8668 -mindirect-branch-register.
8670 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8671 (GOT_memory_operand): Likewise.
8672 (call_insn_operand): Likewise.
8673 (sibcall_insn_operand): Likewise.
8674 (GOT32_symbol_operand): Likewise.
8675 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8676 for -mindirect-branch-register.
8677 (tablejump): Likewise.
8678 (*sibcall_memory): Likewise.
8679 (*sibcall_value_memory): Likewise.
8680 Disallow peepholes of indirect call and jump via memory for
8681 -mindirect-branch-register.
8682 (*call_pop): Replace m with Bw.
8683 (*call_value_pop): Likewise.
8684 (*sibcall_pop_memory): Replace m with Bs.
8685 * config/i386/i386.opt (mindirect-branch-register): New option.
8686 * doc/invoke.texi: Document -mindirect-branch-register option.
8688 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8690 * config/i386/i386-protos.h (ix86_output_function_return): New.
8691 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8692 set function_return_type.
8693 (indirect_thunk_name): Add ret_p to indicate thunk for function
8695 (output_indirect_thunk_function): Pass false to
8696 indirect_thunk_name.
8697 (ix86_output_indirect_branch_via_reg): Likewise.
8698 (ix86_output_indirect_branch_via_push): Likewise.
8699 (output_indirect_thunk_function): Create alias for function
8700 return thunk if regno < 0.
8701 (ix86_output_function_return): New function.
8702 (ix86_handle_fndecl_attribute): Handle function_return.
8703 (ix86_attribute_table): Add function_return.
8704 * config/i386/i386.h (machine_function): Add
8705 function_return_type.
8706 * config/i386/i386.md (simple_return_internal): Use
8707 ix86_output_function_return.
8708 (simple_return_internal_long): Likewise.
8709 * config/i386/i386.opt (mfunction-return=): New option.
8710 (indirect_branch): Mention -mfunction-return=.
8711 * doc/extend.texi: Document function_return function attribute.
8712 * doc/invoke.texi: Document -mfunction-return= option.
8714 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8716 * config/i386/i386-opts.h (indirect_branch): New.
8717 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8718 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8719 with local indirect jump when converting indirect call and jump.
8720 (ix86_set_indirect_branch_type): New.
8721 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8722 (indirectlabelno): New.
8723 (indirect_thunk_needed): Likewise.
8724 (indirect_thunk_bnd_needed): Likewise.
8725 (indirect_thunks_used): Likewise.
8726 (indirect_thunks_bnd_used): Likewise.
8727 (INDIRECT_LABEL): Likewise.
8728 (indirect_thunk_name): Likewise.
8729 (output_indirect_thunk): Likewise.
8730 (output_indirect_thunk_function): Likewise.
8731 (ix86_output_indirect_branch_via_reg): Likewise.
8732 (ix86_output_indirect_branch_via_push): Likewise.
8733 (ix86_output_indirect_branch): Likewise.
8734 (ix86_output_indirect_jmp): Likewise.
8735 (ix86_code_end): Call output_indirect_thunk_function if needed.
8736 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8738 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8739 (ix86_attribute_table): Add indirect_branch.
8740 * config/i386/i386.h (machine_function): Add indirect_branch_type
8741 and has_local_indirect_jump.
8742 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8744 (tablejump): Likewise.
8745 (*indirect_jump): Use ix86_output_indirect_jmp.
8746 (*tablejump_1): Likewise.
8747 (simple_return_indirect_internal): Likewise.
8748 * config/i386/i386.opt (mindirect-branch=): New option.
8749 (indirect_branch): New.
8752 (thunk-inline): Likewise.
8753 (thunk-extern): Likewise.
8754 * doc/extend.texi: Document indirect_branch function attribute.
8755 * doc/invoke.texi: Document -mindirect-branch= option.
8757 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8760 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8762 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8764 * ipa-inline.c (want_inline_small_function_p): Return false if
8765 inlining has already failed with CIF_FINAL_ERROR.
8766 (update_caller_keys): Call want_inline_small_function_p before
8768 (update_callee_keys): Likewise.
8770 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8772 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8774 (rs6000_quadword_masked_address_p): Likewise.
8775 (quad_aligned_load_p): Likewise.
8776 (quad_aligned_store_p): Likewise.
8777 (const_load_sequence_p): Add comment to describe the outer-most loop.
8778 (mimic_memory_attributes_and_flags): New function.
8779 (rs6000_gen_stvx): Likewise.
8780 (replace_swapped_aligned_store): Likewise.
8781 (rs6000_gen_lvx): Likewise.
8782 (replace_swapped_aligned_load): Likewise.
8783 (replace_swapped_load_constant): Capitalize argument name in
8784 comment describing this function.
8785 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8786 and stores that access quad-word aligned addresses and replace
8787 with stvx or lvx instructions when appropriate.
8788 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8789 New function prototype.
8790 (rs6000_quadword_masked_address_p): Likewise.
8791 (rs6000_gen_lvx): Likewise.
8792 (rs6000_gen_stvx): Likewise.
8793 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8794 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8795 when memory address is aligned.
8796 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8797 this split to select lvx instruction when memory address is aligned.
8798 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8799 instruction when memory address is aligned.
8800 (*vsx_le_perm_load_v16qi): Likewise.
8801 (four unnamed splitters): Modify to select the stvx instruction
8802 when memory is aligned.
8804 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8806 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8807 which appears in the queue multiple times.
8809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8810 Alan Hayward <alan.hayward@arm.com>
8811 David Sherwood <david.sherwood@arm.com>
8813 * tree-vectorizer.h (vec_lower_bound): New structure.
8814 (_loop_vec_info): Add check_nonzero and lower_bounds.
8815 (LOOP_VINFO_CHECK_NONZERO): New macro.
8816 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8817 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8818 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8819 fields. Make seg_len the distance travelled, not including the
8821 (dr_direction_indicator): Declare.
8822 (dr_zero_step_indicator): Likewise.
8823 (dr_known_forward_stride_p): Likewise.
8824 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8826 (runtime_alias_check_p): Allow runtime alias checks with
8828 (operator ==): Compare access_size and align.
8829 (prune_runtime_alias_test_list): Rework for new distinction between
8830 the access_size and seg_len.
8831 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8833 (get_segment_min_max): New function.
8834 (create_intersect_range_checks): Use it.
8835 (dr_step_indicator): New function.
8836 (dr_direction_indicator): Likewise.
8837 (dr_zero_step_indicator): Likewise.
8838 (dr_known_forward_stride_p): Likewise.
8839 * tree-loop-distribution.c (data_ref_segment_size): Return
8840 DR_STEP * (niters - 1).
8841 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8843 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8844 (vect_preserves_scalar_order_p): New function, split out from...
8845 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8846 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8847 (vect_vfa_access_size): New function.
8848 (vect_vfa_align): Likewise.
8849 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8850 (dump_lower_bound): New function.
8851 (vect_check_lower_bound): Likewise.
8852 (vect_small_gap_p): Likewise.
8853 (vectorizable_with_step_bound_p): Likewise.
8854 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8855 depencies if the vectorization factor is 1. Convert the checks
8856 for nonzero steps into checks on the bounds of DR_STEP. Try using
8857 a bunds check for variable steps if the minimum required step is
8858 relatively small. Update calls to the dr_with_seg_len
8859 constructor and to vect_compile_time_alias.
8860 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8862 (vect_loop_versioning): Call it.
8863 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8865 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8868 Alan Hayward <alan.hayward@arm.com>
8869 David Sherwood <david.sherwood@arm.com>
8871 * doc/sourcebuild.texi (vect_scatter_store): Document.
8872 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8874 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8876 * genopinit.c (main): Add supports_vec_scatter_store and
8877 supports_vec_scatter_store_cached to target_optabs.
8878 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8879 IFN_MASK_SCATTER_STORE.
8880 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8882 * internal-fn.h (internal_store_fn_p): Declare.
8883 (internal_fn_stored_value_index): Likewise.
8884 * internal-fn.c (scatter_store_direct): New macro.
8885 (expand_scatter_store_optab_fn): New function.
8886 (direct_scatter_store_optab_supported_p): New macro.
8887 (internal_store_fn_p): New function.
8888 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8889 IFN_MASK_SCATTER_STORE.
8890 (internal_fn_mask_index): Likewise.
8891 (internal_fn_stored_value_index): New function.
8892 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8894 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8895 * optabs-query.c (supports_vec_scatter_store_p): New function.
8896 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8897 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8898 true for scatter stores.
8899 (vect_gather_scatter_fn_p): Handle scatter stores too.
8900 (vect_check_gather_scatter): Consider using scatter stores if
8901 supports_vec_scatter_store_p.
8902 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8904 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8905 internal_fn_stored_value_index.
8906 (check_load_store_masking): Handle scatter stores too.
8907 (vect_get_store_rhs): Make public.
8908 (vectorizable_call): Use internal_store_fn_p.
8909 (vectorizable_store): Handle scatter store internal functions.
8910 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8911 when deciding whether the end of the group has been reached.
8912 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8913 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8914 (mask_scatter_store<mode>): New insns.
8916 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8917 Alan Hayward <alan.hayward@arm.com>
8918 David Sherwood <david.sherwood@arm.com>
8920 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8921 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8922 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8924 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8925 Use vect_truncate_gather_scatter_offset if we can't treat the
8926 operation as a normal gather load or scatter store.
8927 (get_group_load_store_type): Take the gather_scatter_info
8928 as argument. Try using a gather load or scatter store for
8929 single-element groups.
8930 (get_load_store_type): Update calls to get_group_load_store_type
8931 and vect_use_strided_gather_scatters_p.
8933 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8934 Alan Hayward <alan.hayward@arm.com>
8935 David Sherwood <david.sherwood@arm.com>
8937 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8938 optional tree argument.
8939 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8941 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8942 but continue to use the current value as a fallback.
8943 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8944 to compare the updates.
8945 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8946 (get_load_store_type): Use it when handling a strided access.
8947 (vect_get_strided_load_store_ops): New function.
8948 (vect_get_data_ptr_increment): Likewise.
8949 (vectorizable_load): Handle strided gather loads. Always pass
8950 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8953 Alan Hayward <alan.hayward@arm.com>
8954 David Sherwood <david.sherwood@arm.com>
8956 * doc/md.texi (gather_load@var{m}): Document.
8957 (mask_gather_load@var{m}): Likewise.
8958 * genopinit.c (main): Add supports_vec_gather_load and
8959 supports_vec_gather_load_cached to target_optabs.
8960 * optabs-tree.c (init_tree_optimization_optabs): Use
8961 ggc_cleared_alloc to allocate target_optabs.
8962 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8963 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8965 * internal-fn.h (internal_load_fn_p): Declare.
8966 (internal_gather_scatter_fn_p): Likewise.
8967 (internal_fn_mask_index): Likewise.
8968 (internal_gather_scatter_fn_supported_p): Likewise.
8969 * internal-fn.c (gather_load_direct): New macro.
8970 (expand_gather_load_optab_fn): New function.
8971 (direct_gather_load_optab_supported_p): New macro.
8972 (direct_internal_fn_optab): New function.
8973 (internal_load_fn_p): Likewise.
8974 (internal_gather_scatter_fn_p): Likewise.
8975 (internal_fn_mask_index): Likewise.
8976 (internal_gather_scatter_fn_supported_p): Likewise.
8977 * optabs-query.c (supports_at_least_one_mode_p): New function.
8978 (supports_vec_gather_load_p): Likewise.
8979 * optabs-query.h (supports_vec_gather_load_p): Declare.
8980 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8981 and memory_type field.
8982 (NUM_PATTERNS): Bump to 15.
8983 * tree-vect-data-refs.c: Include internal-fn.h.
8984 (vect_gather_scatter_fn_p): New function.
8985 (vect_describe_gather_scatter_call): Likewise.
8986 (vect_check_gather_scatter): Try using internal functions for
8987 gather loads. Recognize existing calls to a gather load function.
8988 (vect_analyze_data_refs): Consider using gather loads if
8989 supports_vec_gather_load_p.
8990 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8991 (vect_get_gather_scatter_offset_type): Likewise.
8992 (vect_convert_mask_for_vectype): Likewise.
8993 (vect_add_conversion_to_patterm): Likewise.
8994 (vect_try_gather_scatter_pattern): Likewise.
8995 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8996 (vect_vect_recog_func_ptrs): Add it.
8997 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8998 internal_fn_mask_index and internal_gather_scatter_fn_p.
8999 (check_load_store_masking): Take the gather_scatter_info as an
9000 argument and handle gather loads.
9001 (vect_get_gather_scatter_ops): New function.
9002 (vectorizable_call): Check internal_load_fn_p.
9003 (vectorizable_load): Likewise. Handle gather load internal
9005 (vectorizable_store): Update call to check_load_store_masking.
9006 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9007 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9008 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9009 (aarch64_gather_scale_operand_d): New predicates.
9010 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9011 (mask_gather_load<mode>): New insns.
9013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9014 Alan Hayward <alan.hayward@arm.com>
9015 David Sherwood <david.sherwood@arm.com>
9017 * optabs.def (fold_left_plus_optab): New optab.
9018 * doc/md.texi (fold_left_plus_@var{m}): Document.
9019 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9020 * internal-fn.c (fold_left_direct): Define.
9021 (expand_fold_left_optab_fn): Likewise.
9022 (direct_fold_left_optab_supported_p): Likewise.
9023 * fold-const-call.c (fold_const_fold_left): New function.
9024 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9025 * tree-parloops.c (valid_reduction_p): New function.
9026 (gather_scalar_reductions): Use it.
9027 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9028 (vect_finish_replace_stmt): Declare.
9029 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9030 (needs_fold_left_reduction_p): New function, split out from...
9031 (vect_is_simple_reduction): ...here. Accept reductions that
9032 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9033 (vect_force_simple_reduction): Also store the reduction type in
9034 the assignment's STMT_VINFO_REDUC_TYPE.
9035 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9036 (merge_with_identity): New function.
9037 (vect_expand_fold_left): Likewise.
9038 (vectorize_fold_left_reduction): Likewise.
9039 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9040 scalar phi in place for it. Check for target support and reject
9041 cases that would reassociate the operation. Defer the transform
9042 phase to vectorize_fold_left_reduction.
9043 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9044 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9045 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9049 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9050 call to ifc_temp_var.
9052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9053 Alan Hayward <alan.hayward@arm.com>
9054 David Sherwood <david.sherwood@arm.com>
9056 * target.def (legitimize_address_displacement): Take the original
9057 offset as a poly_int.
9058 * targhooks.h (default_legitimize_address_displacement): Update
9060 * targhooks.c (default_legitimize_address_displacement): Likewise.
9061 * doc/tm.texi: Regenerate.
9062 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9063 as an argument, moving assert of ad->disp == ad->disp_term to...
9064 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9065 Try calling targetm.legitimize_address_displacement before expanding
9066 the address rather than afterwards, and adjust for the new interface.
9067 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9068 Match the new hook interface. Handle SVE addresses.
9069 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9074 * Makefile.in (OBJS): Add early-remat.o.
9075 * target.def (select_early_remat_modes): New hook.
9076 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9077 * doc/tm.texi: Regenerate.
9078 * targhooks.h (default_select_early_remat_modes): Declare.
9079 * targhooks.c (default_select_early_remat_modes): New function.
9080 * timevar.def (TV_EARLY_REMAT): New timevar.
9081 * passes.def (pass_early_remat): New pass.
9082 * tree-pass.h (make_pass_early_remat): Declare.
9083 * early-remat.c: New file.
9084 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9086 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9089 Alan Hayward <alan.hayward@arm.com>
9090 David Sherwood <david.sherwood@arm.com>
9092 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9093 vfm1 with a bound_epilog parameter.
9094 (vect_do_peeling): Update calls accordingly, and move the prologue
9095 call earlier in the function. Treat the base bound_epilog as 0 for
9096 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9097 this base when peeling for gaps.
9098 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9099 with fully-masked loops.
9100 (vect_estimate_min_profitable_iters): Handle the single peeled
9101 iteration in that case.
9103 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9104 Alan Hayward <alan.hayward@arm.com>
9105 David Sherwood <david.sherwood@arm.com>
9107 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9108 single-element interleaving even if the size is not a power of 2.
9109 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9110 accesses for single-element interleaving if the group size is
9113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9114 Alan Hayward <alan.hayward@arm.com>
9115 David Sherwood <david.sherwood@arm.com>
9117 * doc/md.texi (fold_extract_last_@var{m}): Document.
9118 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9119 * optabs.def (fold_extract_last_optab): New optab.
9120 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9121 * internal-fn.c (fold_extract_direct): New macro.
9122 (expand_fold_extract_optab_fn): Likewise.
9123 (direct_fold_extract_optab_supported_p): Likewise.
9124 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9125 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9126 EXTRACT_LAST_REDUCTION.
9127 (get_initial_def_for_reduction): Do not create an initial vector
9128 for EXTRACT_LAST_REDUCTION reductions.
9129 (vectorizable_reduction): Leave the scalar phi in place for
9130 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9131 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9132 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9133 transform phase to vectorizable_condition.
9134 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9136 (vect_finish_stmt_generation): ...here.
9137 (vect_finish_replace_stmt): New function.
9138 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9139 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9141 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9144 Alan Hayward <alan.hayward@arm.com>
9145 David Sherwood <david.sherwood@arm.com>
9147 * doc/md.texi (extract_last_@var{m}): Document.
9148 * optabs.def (extract_last_optab): New optab.
9149 * internal-fn.def (EXTRACT_LAST): New internal function.
9150 * internal-fn.c (cond_unary_direct): New macro.
9151 (expand_cond_unary_optab_fn): Likewise.
9152 (direct_cond_unary_optab_supported_p): Likewise.
9153 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9154 loops using EXTRACT_LAST.
9155 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9156 (extract_last_<mode>): ...this optab.
9157 (vec_extract<mode><Vel>): Update accordingly.
9159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9160 Alan Hayward <alan.hayward@arm.com>
9161 David Sherwood <david.sherwood@arm.com>
9163 * target.def (empty_mask_is_expensive): New hook.
9164 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9165 * doc/tm.texi: Regenerate.
9166 * targhooks.h (default_empty_mask_is_expensive): Declare.
9167 * targhooks.c (default_empty_mask_is_expensive): New function.
9168 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9169 if the target says that empty masks are expensive.
9170 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9172 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9175 Alan Hayward <alan.hayward@arm.com>
9176 David Sherwood <david.sherwood@arm.com>
9178 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9179 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9180 (vect_use_loop_mask_for_alignment_p): New function.
9181 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9182 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9183 niters_skip argument. Make sure that the first niters_skip elements
9184 of the first iteration are inactive.
9185 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9186 Update call to vect_set_loop_masks_directly.
9187 (get_misalign_in_elems): New function, split out from...
9188 (vect_gen_prolog_loop_niters): ...here.
9189 (vect_update_init_of_dr): Take a code argument that specifies whether
9190 the adjustment should be added or subtracted.
9191 (vect_update_init_of_drs): Likewise.
9192 (vect_prepare_for_masked_peels): New function.
9193 (vect_do_peeling): Skip prologue peeling if we're using a mask
9194 instead. Update call to vect_update_inits_of_drs.
9195 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9197 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9198 alignment. Do not include the number of peeled iterations in
9199 the minimum threshold in that case.
9200 (vectorizable_induction): Adjust the start value down by
9201 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9202 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9203 Take the number of skipped iterations into account when calculating
9205 * tree-vect-stmts.c (vect_gen_while_not): New function.
9207 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9208 Alan Hayward <alan.hayward@arm.com>
9209 David Sherwood <david.sherwood@arm.com>
9211 * doc/sourcebuild.texi (vect_fully_masked): Document.
9212 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9214 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9216 (vect_analyze_loop_2): ...here. Don't check the vectorization
9217 factor against the number of loop iterations if the loop is
9220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9221 Alan Hayward <alan.hayward@arm.com>
9222 David Sherwood <david.sherwood@arm.com>
9224 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9225 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9226 (dump_groups): Update accordingly.
9227 (iv_use::mem_type): New member variable.
9228 (address_p): New function.
9229 (record_use): Add a mem_type argument and initialize the new
9231 (record_group_use): Add a mem_type argument. Use address_p.
9232 Remove obsolete null checks of base_object. Update call to record_use.
9233 (find_interesting_uses_op): Update call to record_group_use.
9234 (find_interesting_uses_cond): Likewise.
9235 (find_interesting_uses_address): Likewise.
9236 (get_mem_type_for_internal_fn): New function.
9237 (find_address_like_use): Likewise.
9238 (find_interesting_uses_stmt): Try find_address_like_use before
9239 calling find_interesting_uses_op.
9240 (addr_offset_valid_p): Use the iv mem_type field as the type
9241 of the addressed memory.
9242 (add_autoinc_candidates): Likewise.
9243 (get_address_cost): Likewise.
9244 (split_small_address_groups_p): Use address_p.
9245 (split_address_groups): Likewise.
9246 (add_iv_candidate_for_use): Likewise.
9247 (autoinc_possible_for_pair): Likewise.
9248 (rewrite_groups): Likewise.
9249 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9250 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9251 (get_alias_ptr_type_for_ptr_address): New function.
9252 (rewrite_use_address): Rewrite address uses in calls that were
9253 identified by find_address_like_use.
9255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9256 Alan Hayward <alan.hayward@arm.com>
9257 David Sherwood <david.sherwood@arm.com>
9259 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9261 * gimple-expr.h (is_gimple_addressable: Likewise.
9262 * gimple-expr.c (is_gimple_address): Likewise.
9263 * internal-fn.c (expand_call_mem_ref): New function.
9264 (expand_mask_load_optab_fn): Use it.
9265 (expand_mask_store_optab_fn): Likewise.
9267 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9268 Alan Hayward <alan.hayward@arm.com>
9269 David Sherwood <david.sherwood@arm.com>
9271 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9272 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9273 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9274 (cond_umax@var{mode}): Document.
9275 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9276 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9277 (cond_umin_optab, cond_umax_optab): New optabs.
9278 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9279 (COND_IOR, COND_XOR): New internal functions.
9280 * internal-fn.h (get_conditional_internal_fn): Declare.
9281 * internal-fn.c (cond_binary_direct): New macro.
9282 (expand_cond_binary_optab_fn): Likewise.
9283 (direct_cond_binary_optab_supported_p): Likewise.
9284 (get_conditional_internal_fn): New function.
9285 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9286 Cope with reduction statements that are vectorized as calls rather
9288 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9289 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9290 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9291 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9292 (UNSPEC_COND_EOR): New unspecs.
9293 (optab): Add mappings for them.
9294 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9295 (sve_int_op, sve_fp_op): New int attributes.
9297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9298 Alan Hayward <alan.hayward@arm.com>
9299 David Sherwood <david.sherwood@arm.com>
9301 * optabs.def (while_ult_optab): New optab.
9302 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9303 * internal-fn.def (WHILE_ULT): New internal function.
9304 * internal-fn.h (direct_internal_fn_supported_p): New override
9305 that takes two types as argument.
9306 * internal-fn.c (while_direct): New macro.
9307 (expand_while_optab_fn): New function.
9308 (convert_optab_supported_p): Likewise.
9309 (direct_while_optab_supported_p): New macro.
9310 * wide-int.h (wi::udiv_ceil): New function.
9311 * tree-vectorizer.h (rgroup_masks): New structure.
9312 (vec_loop_masks): New typedef.
9313 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9315 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9316 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9317 (vect_max_vf): New function.
9318 (slpeel_make_loop_iterate_ntimes): Delete.
9319 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9320 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9321 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9322 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9323 internal-fn.h, stor-layout.h and optabs-query.h.
9324 (vect_set_loop_mask): New function.
9325 (add_preheader_seq): Likewise.
9326 (add_header_seq): Likewise.
9327 (interleave_supported_p): Likewise.
9328 (vect_maybe_permute_loop_masks): Likewise.
9329 (vect_set_loop_masks_directly): Likewise.
9330 (vect_set_loop_condition_masked): Likewise.
9331 (vect_set_loop_condition_unmasked): New function, split out from
9332 slpeel_make_loop_iterate_ntimes.
9333 (slpeel_make_loop_iterate_ntimes): Rename to..
9334 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9335 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9336 (vect_do_peeling): Update call accordingly.
9337 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9339 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9340 mask_compare_type, can_fully_mask_p and fully_masked_p.
9341 (release_vec_loop_masks): New function.
9342 (_loop_vec_info): Use it to free the loop masks.
9343 (can_produce_all_loop_masks_p): New function.
9344 (vect_get_max_nscalars_per_iter): Likewise.
9345 (vect_verify_full_masking): Likewise.
9346 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9347 retries, and free the mask rgroups before retrying. Check loop-wide
9348 reasons for disallowing fully-masked loops. Make the final decision
9349 about whether use a fully-masked loop or not.
9350 (vect_estimate_min_profitable_iters): Do not assume that peeling
9351 for the number of iterations will be needed for fully-masked loops.
9352 (vectorizable_reduction): Disable fully-masked loops.
9353 (vectorizable_live_operation): Likewise.
9354 (vect_halve_mask_nunits): New function.
9355 (vect_double_mask_nunits): Likewise.
9356 (vect_record_loop_mask): Likewise.
9357 (vect_get_loop_mask): Likewise.
9358 (vect_transform_loop): Handle the case in which the final loop
9359 iteration might handle a partial vector. Call vect_set_loop_condition
9360 instead of slpeel_make_loop_iterate_ntimes.
9361 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9362 (check_load_store_masking): New function.
9363 (prepare_load_store_mask): Likewise.
9364 (vectorizable_store): Handle fully-masked loops.
9365 (vectorizable_load): Likewise.
9366 (supportable_widening_operation): Use vect_halve_mask_nunits for
9368 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9369 (vect_gen_while): New function.
9370 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9371 (aarch64_uqdec<mode>): New insn.
9373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9374 Alan Hayward <alan.hayward@arm.com>
9375 David Sherwood <david.sherwood@arm.com>
9377 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9378 (reduc_xor_scal_optab): New optabs.
9379 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9380 (reduc_xor_scal_@var{m}): Document.
9381 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9382 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9384 * fold-const-call.c (fold_const_call): Handle them.
9385 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9386 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9387 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9388 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9389 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9390 (UNSPEC_XORV): New unspecs.
9391 (optab): Add entries for them.
9392 (BITWISEV): New int iterator.
9393 (bit_reduc_op): New int attributes.
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 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9400 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9401 * optabs.def (vec_shl_insert_optab): New optab.
9402 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9403 (duplicate_and_interleave): Likewise.
9404 * tree-vect-loop.c: Include internal-fn.h.
9405 (neutral_op_for_slp_reduction): New function, split out from
9406 get_initial_defs_for_reduction.
9407 (get_initial_def_for_reduction): Handle option 2 for variable-length
9408 vectors by loading the neutral value into a vector and then shifting
9409 the initial value into element 0.
9410 (get_initial_defs_for_reduction): Replace the code argument with
9411 the neutral value calculated by neutral_op_for_slp_reduction.
9412 Use gimple_build_vector for constant-length vectors.
9413 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9414 but the first group_size elements have a neutral value.
9415 Use duplicate_and_interleave otherwise.
9416 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9417 Update call to get_initial_defs_for_reduction. Handle SLP
9418 reductions for variable-length vectors by creating one vector
9419 result for each scalar result, with the elements associated
9420 with other scalar results stubbed out with the neutral value.
9421 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9422 Require IFN_VEC_SHL_INSERT for double reductions on
9423 variable-length vectors, or SLP reductions that have
9424 a neutral value. Require can_duplicate_and_interleave_p
9425 support for variable-length unchained SLP reductions if there
9426 is no neutral value, such as for MIN/MAX reductions. Also require
9427 the number of vector elements to be a multiple of the number of
9428 SLP statements when doing variable-length unchained SLP reductions.
9429 Update call to vect_create_epilog_for_reduction.
9430 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9431 and remove initial values.
9432 (duplicate_and_interleave): Make public.
9433 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9434 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9437 Alan Hayward <alan.hayward@arm.com>
9438 David Sherwood <david.sherwood@arm.com>
9440 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9441 (can_duplicate_and_interleave_p): New function.
9442 (vect_get_and_check_slp_defs): Take the vector of statements
9443 rather than just the current one. Remove excess parentheses.
9444 Restriction rejectinon of vect_constant_def and vect_external_def
9445 for variable-length vectors to boolean types, or types for which
9446 can_duplicate_and_interleave_p is false.
9447 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9448 (duplicate_and_interleave): New function.
9449 (vect_get_constant_vectors): Use gimple_build_vector for
9450 constant-length vectors and suitable variable-length constant
9451 vectors. Use duplicate_and_interleave for other variable-length
9452 vectors. Don't defer the update when inserting new statements.
9454 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9455 Alan Hayward <alan.hayward@arm.com>
9456 David Sherwood <david.sherwood@arm.com>
9458 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9459 min_profitable_iters doesn't go negative.
9461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9462 Alan Hayward <alan.hayward@arm.com>
9463 David Sherwood <david.sherwood@arm.com>
9465 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9466 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9467 * optabs.def (vec_mask_load_lanes_optab): New optab.
9468 (vec_mask_store_lanes_optab): Likewise.
9469 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9470 (MASK_STORE_LANES): Likewise.
9471 * internal-fn.c (mask_load_lanes_direct): New macro.
9472 (mask_store_lanes_direct): Likewise.
9473 (expand_mask_load_optab_fn): Handle masked operations.
9474 (expand_mask_load_lanes_optab_fn): New macro.
9475 (expand_mask_store_optab_fn): Handle masked operations.
9476 (expand_mask_store_lanes_optab_fn): New macro.
9477 (direct_mask_load_lanes_optab_supported_p): Likewise.
9478 (direct_mask_store_lanes_optab_supported_p): Likewise.
9479 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9481 (vect_load_lanes_supported): Likewise.
9482 * tree-vect-data-refs.c (strip_conversion): New function.
9483 (can_group_stmts_p): Likewise.
9484 (vect_analyze_data_ref_accesses): Use it instead of checking
9485 for a pair of assignments.
9486 (vect_store_lanes_supported): Take a masked_p parameter.
9487 (vect_load_lanes_supported): Likewise.
9488 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9489 vect_store_lanes_supported and vect_load_lanes_supported.
9490 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9491 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9492 parameter. Don't allow gaps for masked accesses.
9493 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9494 and vect_load_lanes_supported.
9495 (get_load_store_type): Take a masked_p parameter and update
9496 call to get_group_load_store_type.
9497 (vectorizable_store): Update call to get_load_store_type.
9498 Handle IFN_MASK_STORE_LANES.
9499 (vectorizable_load): Update call to get_load_store_type.
9500 Handle IFN_MASK_LOAD_LANES.
9502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9503 Alan Hayward <alan.hayward@arm.com>
9504 David Sherwood <david.sherwood@arm.com>
9506 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9508 * config/aarch64/aarch64-protos.h
9509 (aarch64_sve_struct_memory_operand_p): Declare.
9510 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9511 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9512 (VPRED, vpred): Handle SVE structure modes.
9513 * config/aarch64/constraints.md (Utx): New constraint.
9514 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9515 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9516 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9517 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9518 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9519 structure modes. Split into pieces after RA.
9520 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9521 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9523 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9524 SVE structure modes.
9525 (aarch64_classify_address): Likewise.
9526 (sizetochar): Move earlier in file.
9527 (aarch64_print_operand): Handle SVE register lists.
9528 (aarch64_array_mode): New function.
9529 (aarch64_sve_struct_memory_operand_p): Likewise.
9530 (TARGET_ARRAY_MODE): Redefine.
9532 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9533 Alan Hayward <alan.hayward@arm.com>
9534 David Sherwood <david.sherwood@arm.com>
9536 * target.def (array_mode): New target hook.
9537 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9538 * doc/tm.texi: Regenerate.
9539 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9540 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9541 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9543 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9546 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9547 Alan Hayward <alan.hayward@arm.com>
9548 David Sherwood <david.sherwood@arm.com>
9550 * fold-const.c (fold_binary_loc): Check the argument types
9551 rather than the result type when testing for a vector operation.
9553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9555 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9556 * doc/tm.texi: Regenerate.
9558 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9559 Alan Hayward <alan.hayward@arm.com>
9560 David Sherwood <david.sherwood@arm.com>
9562 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9563 (sve): Document new AArch64 extension.
9564 * doc/md.texi (w): Extend the description of the AArch64
9565 constraint to include SVE vectors.
9566 (Upl, Upa): Document new AArch64 predicate constraints.
9567 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9569 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9570 (msve-vector-bits=): New option.
9571 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9572 SVE when these are disabled.
9573 (sve): New extension.
9574 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9575 modes. Adjust their number of units based on aarch64_sve_vg.
9576 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9577 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9578 aarch64_addr_query_type.
9579 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9580 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9581 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9582 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9583 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9584 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9585 (aarch64_simd_imm_zero_p): Delete.
9586 (aarch64_check_zero_based_sve_index_immediate): Declare.
9587 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9588 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9589 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9590 (aarch64_sve_float_mul_immediate_p): Likewise.
9591 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9593 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9594 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9595 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9596 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9597 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9598 (aarch64_regmode_natural_size): Likewise.
9599 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9600 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9602 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9603 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9604 for VG and the SVE predicate registers.
9605 (V_ALIASES): Add a "z"-prefixed alias.
9606 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9607 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9608 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9609 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9610 (REG_CLASS_NAMES): Add entries for them.
9611 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9612 and the predicate registers.
9613 (aarch64_sve_vg): Declare.
9614 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9615 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9616 (REGMODE_NATURAL_SIZE): Define.
9617 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9619 * config/aarch64/aarch64.c: Include cfgrtl.h.
9620 (simd_immediate_info): Add a constructor for series vectors,
9621 and an associated step field.
9622 (aarch64_sve_vg): New variable.
9623 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9624 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9625 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9626 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9627 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9628 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9629 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9630 (aarch64_get_mask_mode): New functions.
9631 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9632 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9633 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9634 predicate modes and predicate registers. Explicitly restrict
9635 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9636 to store a vector mode if it is recognized by
9637 aarch64_classify_vector_mode.
9638 (aarch64_regmode_natural_size): New function.
9639 (aarch64_hard_regno_caller_save_mode): Return the original mode
9641 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9642 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9643 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9644 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9646 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9647 does not overlap dest if the function is frame-related. Handle
9649 (aarch64_split_add_offset): New function.
9650 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9651 them aarch64_add_offset.
9652 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9653 and update call to aarch64_sub_sp.
9654 (aarch64_add_cfa_expression): New function.
9655 (aarch64_expand_prologue): Pass extra temporary registers to the
9656 functions above. Handle the case in which we need to emit new
9657 DW_CFA_expressions for registers that were originally saved
9658 relative to the stack pointer, but now have to be expressed
9659 relative to the frame pointer.
9660 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9662 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9663 IP0 and IP1 values for SVE frames.
9664 (aarch64_expand_vec_series): New function.
9665 (aarch64_expand_sve_widened_duplicate): Likewise.
9666 (aarch64_expand_sve_const_vector): Likewise.
9667 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9668 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9669 into the register, rather than emitting a SET directly.
9670 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9671 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9672 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9673 (offset_9bit_signed_scaled_p): New functions.
9674 (aarch64_replicate_bitmask_imm): New function.
9675 (aarch64_bitmask_imm): Use it.
9676 (aarch64_cannot_force_const_mem): Reject expressions involving
9677 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9678 (aarch64_classify_index): Handle SVE indices, by requiring
9679 a plain register index with a scale that matches the element size.
9680 (aarch64_classify_address): Handle SVE addresses. Assert that
9681 the mode of the address is VOIDmode or an integer mode.
9682 Update call to aarch64_classify_symbol.
9683 (aarch64_classify_symbolic_expression): Update call to
9684 aarch64_classify_symbol.
9685 (aarch64_const_vec_all_in_range_p): New function.
9686 (aarch64_print_vector_float_operand): Likewise.
9687 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9688 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9689 and the FP immediates 1.0 and 0.5.
9690 (aarch64_print_address_internal): Handle SVE addresses.
9691 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9692 (aarch64_regno_regclass): Handle predicate registers.
9693 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9695 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9696 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9697 (aarch64_convert_sve_vector_bits): New function.
9698 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9699 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9701 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9702 Handle SVE vector and predicate modes. Accept VL-based constants
9703 that need only one temporary register, and VL offsets that require
9704 no temporary registers.
9705 (aarch64_conditional_register_usage): Mark the predicate registers
9706 as fixed if SVE isn't available.
9707 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9708 Return true for SVE vector and predicate modes.
9709 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9710 rather than an unsigned int. Handle SVE modes.
9711 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9713 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9715 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9716 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9717 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9718 (aarch64_sve_float_mul_immediate_p): New functions.
9719 (aarch64_sve_valid_immediate): New function.
9720 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9721 Explicitly reject structure modes. Check for INDEX constants.
9722 Handle PTRUE and PFALSE constants.
9723 (aarch64_check_zero_based_sve_index_immediate): New function.
9724 (aarch64_simd_imm_zero_p): Delete.
9725 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9726 vector modes. Accept constants in the range of CNT[BHWD].
9727 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9728 ask for an Advanced SIMD mode.
9729 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9730 (aarch64_simd_vector_alignment): Handle SVE predicates.
9731 (aarch64_vectorize_preferred_vector_alignment): New function.
9732 (aarch64_simd_vector_alignment_reachable): Use it instead of
9734 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9735 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9737 (MAX_VECT_LEN): Delete.
9738 (expand_vec_perm_d): Add a vec_flags field.
9739 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9740 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9741 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9743 (aarch64_evpc_rev): Rename to...
9744 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9745 (aarch64_evpc_rev_global): New function.
9746 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9747 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9749 (aarch64_evpc_sve_tbl): New function.
9750 (aarch64_expand_vec_perm_const_1): Update after rename of
9751 aarch64_evpc_rev. Handle SVE permutes too, trying
9752 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9753 than aarch64_evpc_tbl.
9754 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9755 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9756 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9757 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9758 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9759 (aarch64_expand_sve_vcond): New functions.
9760 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9761 of aarch64_vector_mode_p.
9762 (aarch64_dwarf_poly_indeterminate_value): New function.
9763 (aarch64_compute_pressure_classes): Likewise.
9764 (aarch64_can_change_mode_class): Likewise.
9765 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9766 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9767 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9768 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9769 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9770 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9771 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9772 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9774 (Dn, Dl, Dr): Accept const as well as const_vector.
9775 (Dz): Likewise. Compare against CONST0_RTX.
9776 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9777 of "vector" where appropriate.
9778 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9779 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9780 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9781 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9782 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9783 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9784 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9785 (v_int_equiv): Extend to SVE modes.
9786 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9788 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9789 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9790 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9791 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9792 (SVE_COND_FP_CMP): New int iterators.
9793 (perm_hilo): Handle the new unpack unspecs.
9794 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9796 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9797 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9798 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9799 (aarch64_equality_operator, aarch64_constant_vector_operand)
9800 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9801 (aarch64_sve_nonimmediate_operand): Likewise.
9802 (aarch64_sve_general_operand): Likewise.
9803 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9804 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9805 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9806 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9807 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9808 (aarch64_sve_float_arith_immediate): Likewise.
9809 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9810 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9811 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9812 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9813 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9814 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9815 (aarch64_sve_float_arith_operand): Likewise.
9816 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9817 (aarch64_sve_float_mul_operand): Likewise.
9818 (aarch64_sve_vec_perm_operand): Likewise.
9819 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9820 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9821 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9822 as well as const_vector.
9823 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9824 in file. Use CONST0_RTX and CONSTM1_RTX.
9825 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9826 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9827 Use aarch64_simd_imm_zero.
9828 * config/aarch64/aarch64-sve.md: New file.
9829 * config/aarch64/aarch64.md: Include it.
9830 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9831 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9832 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9833 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9834 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9835 (sve): New attribute.
9836 (enabled): Disable instructions with the sve attribute unless
9838 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9839 aarch64_expand_mov_immediate.
9840 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9841 CNT[BHSD] immediates.
9842 (movti): Split CONST_POLY_INT moves into two halves.
9843 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9844 Split additions that need a temporary here if the destination
9845 is the stack pointer.
9846 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9847 (*add<mode>3_poly_1): New instruction.
9848 (set_clobber_cc): New expander.
9850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9852 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9853 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9854 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9855 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9856 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9857 Change innermode from fixed_mode_size to machine_mode.
9858 (simplify_subreg): Update call accordingly. Handle a constant-sized
9859 subreg of a variable-length CONST_VECTOR.
9861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9862 Alan Hayward <alan.hayward@arm.com>
9863 David Sherwood <david.sherwood@arm.com>
9865 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9866 (add_offset_to_base): New function, split out from...
9867 (create_mem_ref): ...here. When handling a scale other than 1,
9868 check first whether the address is valid without the offset.
9869 Add it into the base if so, leaving the index and scale as-is.
9871 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9874 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9875 fold_for_warn before checking if arg2 is INTEGER_CST.
9877 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9879 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9880 (store_multiple_operation): Delete.
9881 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9882 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9883 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9884 guarded by TARGET_STRING.
9885 (rs6000_output_load_multiple): Delete.
9886 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9887 OPTION_MASK_STRING / TARGET_STRING handling.
9888 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9889 (const rs6000_opt_masks) <"string">: Change mask to 0.
9890 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9891 (MASK_STRING): Delete.
9892 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9894 (load_multiple): Delete.
9901 (store_multiple): Delete.
9908 (movmemsi_8reg): Delete.
9909 (corresponding unnamed define_insn): Delete.
9910 (movmemsi_6reg): Delete.
9911 (corresponding unnamed define_insn): Delete.
9912 (movmemsi_4reg): Delete.
9913 (corresponding unnamed define_insn): Delete.
9914 (movmemsi_2reg): Delete.
9915 (corresponding unnamed define_insn): Delete.
9916 (movmemsi_1reg): Delete.
9917 (corresponding unnamed define_insn): Delete.
9918 * config/rs6000/rs6000.opt (mno-string): New.
9919 (mstring): Replace by deprecation warning stub.
9920 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9922 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9924 * regrename.c (regrename_do_replace): If replacing the same
9925 reg multiple times, try to reuse last created gen_raw_REG.
9928 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9929 main to workaround a bug in GDB.
9931 2018-01-12 Tom de Vries <tom@codesourcery.com>
9934 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9936 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9938 PR rtl-optimization/80481
9939 * ira-color.c (get_cap_member): New function.
9940 (allocnos_conflict_by_live_ranges_p): Use it.
9941 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9942 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9944 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9947 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9948 (*saddl_se_1): Ditto.
9950 (*ssubl_se_1): Ditto.
9952 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9954 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9955 rather than wi::to_widest for DR_INITs.
9956 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9957 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9958 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9960 (vect_analyze_group_access_1): Note that here.
9962 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9964 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9965 polynomial type sizes.
9967 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9969 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9970 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9971 (gimple_add_tmp_var): Likewise.
9973 2018-01-12 Martin Liska <mliska@suse.cz>
9975 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9976 (gimple_alloc_sizes): Likewise.
9977 (dump_gimple_statistics): Use PRIu64 in printf format.
9978 * gimple.h: Change uint64_t to int.
9980 2018-01-12 Martin Liska <mliska@suse.cz>
9982 * tree-core.h: Use uint64_t instead of int.
9983 * tree.c (tree_node_counts): Likewise.
9984 (tree_node_sizes): Likewise.
9985 (dump_tree_statistics): Use PRIu64 in printf format.
9987 2018-01-12 Martin Liska <mliska@suse.cz>
9989 * Makefile.in: As qsort_chk is implemented in vec.c, add
9990 vec.o to linkage of gencfn-macros.
9991 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9992 passing the info to record_node_allocation_statistics.
9993 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9995 * ggc-common.c (struct ggc_usage): Add operator== and use
9996 it in operator< and compare function.
9997 * mem-stats.h (struct mem_usage): Likewise.
9998 * vec.c (struct vec_usage): Remove operator< and compare
9999 function. Can be simply inherited.
10001 2018-01-12 Martin Jambor <mjambor@suse.cz>
10004 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10005 * tree-ssa-math-opts.c: Include domwalk.h.
10006 (convert_mult_to_fma_1): New function.
10007 (fma_transformation_info): New type.
10008 (fma_deferring_state): Likewise.
10009 (cancel_fma_deferring): New function.
10010 (result_of_phi): Likewise.
10011 (last_fma_candidate_feeds_initial_phi): Likewise.
10012 (convert_mult_to_fma): Added deferring logic, split actual
10013 transformation to convert_mult_to_fma_1.
10014 (math_opts_dom_walker): New type.
10015 (math_opts_dom_walker::after_dom_children): New method, body moved
10016 here from pass_optimize_widening_mul::execute, added deferring logic
10018 (pass_optimize_widening_mul::execute): Moved most of code to
10019 math_opts_dom_walker::after_dom_children.
10020 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10021 * config/i386/i386.c (ix86_option_override_internal): Added
10022 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10024 2018-01-12 Richard Biener <rguenther@suse.de>
10027 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10028 inline instance vars.
10030 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10033 * config/rx/rx.c (rx_is_restricted_memory_address):
10034 Handle SUBREG case.
10036 2018-01-12 Richard Biener <rguenther@suse.de>
10038 PR tree-optimization/80846
10039 * target.def (split_reduction): New target hook.
10040 * targhooks.c (default_split_reduction): New function.
10041 * targhooks.h (default_split_reduction): Declare.
10042 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10043 target requests first reduce vectors by combining low and high
10045 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10046 (get_vectype_for_scalar_type_and_size): Export.
10047 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10048 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10049 * doc/tm.texi: Regenerate.
10050 * config/i386/i386.c (ix86_split_reduction): Implement
10051 TARGET_VECTORIZE_SPLIT_REDUCTION.
10053 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10056 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10057 in PIC mode except for TARGET_VXWORKS_RTP.
10058 * config/sparc/sparc.c: Include cfgrtl.h.
10059 (TARGET_INIT_PIC_REG): Define.
10060 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10061 (sparc_pic_register_p): New predicate.
10062 (sparc_legitimate_address_p): Use it.
10063 (sparc_legitimize_pic_address): Likewise.
10064 (sparc_delegitimize_address): Likewise.
10065 (sparc_mode_dependent_address_p): Likewise.
10066 (gen_load_pcrel_sym): Remove 4th parameter.
10067 (load_got_register): Adjust call to above. Remove obsolete stuff.
10068 (sparc_expand_prologue): Do not call load_got_register here.
10069 (sparc_flat_expand_prologue): Likewise.
10070 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10071 (sparc_use_pseudo_pic_reg): New function.
10072 (sparc_init_pic_reg): Likewise.
10073 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10074 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10076 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10078 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10079 Add item for branch_cost.
10081 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10083 PR rtl-optimization/83565
10084 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10085 not extend the result to a larger mode for rotate operations.
10086 (num_sign_bit_copies1): Likewise.
10088 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10091 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10093 Use values-Xc.o for -pedantic.
10094 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10096 2018-01-12 Martin Liska <mliska@suse.cz>
10099 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10101 (possible_polymorphic_call_targets): Use it.
10102 (ipa_devirt): Likewise.
10104 2018-01-12 Martin Liska <mliska@suse.cz>
10106 * profile-count.h (enum profile_quality): Use 0 as invalid
10107 enum value of profile_quality.
10109 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10111 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10112 -mext-string options.
10114 2018-01-12 Richard Biener <rguenther@suse.de>
10116 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10117 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10118 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10120 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10122 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10124 * configure.ac (--with-long-double-format): Add support for the
10125 configuration option to change the default long double format on
10127 * config.gcc (powerpc*-linux*-*): Likewise.
10128 * configure: Regenerate.
10129 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10130 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10131 used without modification.
10133 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10135 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10136 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10137 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10138 MISC_BUILTIN_SPEC_BARRIER.
10139 (rs6000_init_builtins): Likewise.
10140 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10142 (speculation_barrier): New define_insn.
10143 * doc/extend.texi: Document __builtin_speculation_barrier.
10145 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10148 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10149 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10150 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10152 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10153 integral modes instead of "ss" and "sd".
10154 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10155 vectors with 32-bit and 64-bit elements.
10156 (vecdupssescalarmodesuffix): New mode attribute.
10157 (vec_dup<mode>): Use it.
10159 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10162 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10163 frame if argument is passed on stack.
10165 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10168 * ree.c (combine_reaching_defs): Optimize also
10169 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10170 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10172 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10174 PR middle-end/83189
10175 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10177 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10179 PR middle-end/83718
10180 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10181 after they are computed.
10183 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10185 PR tree-optimization/83695
10186 * gimple-loop-linterchange.cc
10187 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10188 reset cached scev information after interchange.
10189 (pass_linterchange::execute): Remove call to scev_reset_htab.
10191 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10193 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10194 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10195 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10196 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10197 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10198 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10199 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10200 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10201 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10202 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10203 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10204 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10205 (V_lane_reg): Likewise.
10206 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10208 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10209 (vfmal_lane_low<mode>_intrinsic,
10210 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10211 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10212 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10213 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10214 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10215 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10217 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10219 * config/arm/arm-cpus.in (fp16fml): New feature.
10220 (ALL_SIMD): Add fp16fml.
10221 (armv8.2-a): Add fp16fml as an option.
10222 (armv8.3-a): Likewise.
10223 (armv8.4-a): Add fp16fml as part of fp16.
10224 * config/arm/arm.h (TARGET_FP16FML): Define.
10225 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10227 * config/arm/arm-modes.def (V2HF): Define.
10228 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10229 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10230 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10231 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10232 vfmsl_low, vfmsl_high): New set of builtins.
10233 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10234 (vfml_op): New code attribute.
10235 (VFMLHALVES): New int iterator.
10236 (VFML, VFMLSEL): New mode attributes.
10237 (V_reg): Define mapping for V2HF.
10238 (V_hi, V_lo): New mode attributes.
10239 (VF_constraint): Likewise.
10240 (vfml_half, vfml_half_selector): New int attributes.
10241 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10243 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10244 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10246 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10247 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10248 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10249 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10251 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10252 Document new effective target and option set.
10254 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10256 * config/arm/arm-cpus.in (armv8_4): New feature.
10257 (ARMv8_4a): New fgroup.
10258 (armv8.4-a): New arch.
10259 * config/arm/arm-tables.opt: Regenerate.
10260 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10261 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10262 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10263 Add matching rules for -march=armv8.4-a and extensions.
10264 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10266 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10269 * config/rx/rx.md (BW): New mode attribute.
10270 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10272 2018-01-11 Richard Biener <rguenther@suse.de>
10274 PR tree-optimization/83435
10275 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10276 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10277 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10279 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10280 Alan Hayward <alan.hayward@arm.com>
10281 David Sherwood <david.sherwood@arm.com>
10283 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10285 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10286 (aarch64_print_address_internal): Use it to check for a zero offset.
10288 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10289 Alan Hayward <alan.hayward@arm.com>
10290 David Sherwood <david.sherwood@arm.com>
10292 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10293 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10294 Return a poly_int64 rather than a HOST_WIDE_INT.
10295 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10296 rather than a HOST_WIDE_INT.
10297 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10298 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10299 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10300 final_offset from HOST_WIDE_INT to poly_int64.
10301 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10302 to_constant when getting the number of units in an Advanced SIMD
10304 (aarch64_builtin_vectorized_function): Check for a constant number
10306 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10308 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10309 attribute instead of GET_MODE_NUNITS.
10310 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10311 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10312 GET_MODE_SIZE for fixed-size registers.
10313 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10314 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10315 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10316 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10317 (aarch64_print_operand, aarch64_print_address_internal)
10318 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10319 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10320 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10321 Handle polynomial GET_MODE_SIZE.
10322 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10323 wider than SImode without modification.
10324 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10325 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10326 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10327 passing and returning SVE modes.
10328 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10329 rather than GEN_INT.
10330 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10331 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10332 (aarch64_allocate_and_probe_stack_space): Likewise.
10333 (aarch64_layout_frame): Cope with polynomial offsets.
10334 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10335 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10336 polynomial offsets.
10337 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10338 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10339 poly_int64 rather than a HOST_WIDE_INT.
10340 (aarch64_get_separate_components, aarch64_process_components)
10341 (aarch64_expand_prologue, aarch64_expand_epilogue)
10342 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10343 (aarch64_anchor_offset): New function, split out from...
10344 (aarch64_legitimize_address): ...here.
10345 (aarch64_builtin_vectorization_cost): Handle polynomial
10346 TYPE_VECTOR_SUBPARTS.
10347 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10349 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10350 number of elements from the PARALLEL rather than the mode.
10351 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10352 rather than GET_MODE_BITSIZE.
10353 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10354 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10355 (aarch64_expand_vec_perm_const_1): Handle polynomial
10356 d->perm.length () and d->perm elements.
10357 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10358 Apply to_constant to d->perm elements.
10359 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10360 polynomial CONST_VECTOR_NUNITS.
10361 (aarch64_move_pointer): Take amount as a poly_int64 rather
10363 (aarch64_progress_pointer): Avoid temporary variable.
10364 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10365 the mode attribute instead of GET_MODE.
10367 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10368 Alan Hayward <alan.hayward@arm.com>
10369 David Sherwood <david.sherwood@arm.com>
10371 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10372 x exists before using it.
10373 (aarch64_add_constant_internal): Rename to...
10374 (aarch64_add_offset_1): ...this. Replace regnum with separate
10375 src and dest rtxes. Handle the case in which they're different,
10376 including when the offset is zero. Replace scratchreg with an rtx.
10377 Use 2 additions if there is no spare register into which we can
10378 move a 16-bit constant.
10379 (aarch64_add_constant): Delete.
10380 (aarch64_add_offset): Replace reg with separate src and dest
10381 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10382 Use aarch64_add_offset_1.
10383 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10384 an rtx rather than an int. Take the delta as a poly_int64
10385 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10386 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10387 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10388 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10389 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10390 and aarch64_add_sp.
10391 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10392 aarch64_add_constant.
10394 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10396 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10397 Use scalar_float_mode.
10399 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10401 * config/aarch64/aarch64-simd.md
10402 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10403 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10404 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10405 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10406 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10407 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10408 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10409 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10410 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10411 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10413 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10416 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10417 targ_options->x_arm_arch_string is non NULL.
10419 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10421 * config/aarch64/aarch64.h
10422 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10424 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10427 * expmed.c (emit_store_flag_force): Swap if const op0
10428 and change VOIDmode to mode of op0.
10430 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10432 PR rtl-optimization/83761
10433 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10434 than bytes to mode_for_size.
10436 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10438 PR middle-end/83189
10439 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10440 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10443 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10445 PR middle-end/83575
10446 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10447 when in layout mode.
10448 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10449 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10452 2018-01-10 Michael Collison <michael.collison@arm.com>
10454 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10455 * config/aarch64/aarch64-option-extension.def: Add
10456 AARCH64_OPT_EXTENSION of 'fp16fml'.
10457 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10458 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10459 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10460 * config/aarch64/constraints.md (Ui7): New constraint.
10461 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10462 (VFMLA_SEL_W): Ditto.
10465 (VFMLA16_LOW): New int iterator.
10466 (VFMLA16_HIGH): Ditto.
10467 (UNSPEC_FMLAL): New unspec.
10468 (UNSPEC_FMLSL): Ditto.
10469 (UNSPEC_FMLAL2): Ditto.
10470 (UNSPEC_FMLSL2): Ditto.
10471 (f16mac): New code attribute.
10472 * config/aarch64/aarch64-simd-builtins.def
10473 (aarch64_fmlal_lowv2sf): Ditto.
10474 (aarch64_fmlsl_lowv2sf): Ditto.
10475 (aarch64_fmlalq_lowv4sf): Ditto.
10476 (aarch64_fmlslq_lowv4sf): Ditto.
10477 (aarch64_fmlal_highv2sf): Ditto.
10478 (aarch64_fmlsl_highv2sf): Ditto.
10479 (aarch64_fmlalq_highv4sf): Ditto.
10480 (aarch64_fmlslq_highv4sf): Ditto.
10481 (aarch64_fmlal_lane_lowv2sf): Ditto.
10482 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10483 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10484 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10485 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10486 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10487 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10488 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10489 (aarch64_fmlal_lane_highv2sf): Ditto.
10490 (aarch64_fmlsl_lane_highv2sf): Ditto.
10491 (aarch64_fmlal_laneq_highv2sf): Ditto.
10492 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10493 (aarch64_fmlalq_lane_highv4sf): Ditto.
10494 (aarch64_fmlsl_lane_highv4sf): Ditto.
10495 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10496 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10497 * config/aarch64/aarch64-simd.md:
10498 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10499 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10500 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10501 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10502 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10503 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10504 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10505 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10506 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10507 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10508 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10509 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10510 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10511 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10512 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10513 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10514 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10515 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10516 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10517 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10518 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10519 (vfmlsl_low_u32): Ditto.
10520 (vfmlalq_low_u32): Ditto.
10521 (vfmlslq_low_u32): Ditto.
10522 (vfmlal_high_u32): Ditto.
10523 (vfmlsl_high_u32): Ditto.
10524 (vfmlalq_high_u32): Ditto.
10525 (vfmlslq_high_u32): Ditto.
10526 (vfmlal_lane_low_u32): Ditto.
10527 (vfmlsl_lane_low_u32): Ditto.
10528 (vfmlal_laneq_low_u32): Ditto.
10529 (vfmlsl_laneq_low_u32): Ditto.
10530 (vfmlalq_lane_low_u32): Ditto.
10531 (vfmlslq_lane_low_u32): Ditto.
10532 (vfmlalq_laneq_low_u32): Ditto.
10533 (vfmlslq_laneq_low_u32): Ditto.
10534 (vfmlal_lane_high_u32): Ditto.
10535 (vfmlsl_lane_high_u32): Ditto.
10536 (vfmlal_laneq_high_u32): Ditto.
10537 (vfmlsl_laneq_high_u32): Ditto.
10538 (vfmlalq_lane_high_u32): Ditto.
10539 (vfmlslq_lane_high_u32): Ditto.
10540 (vfmlalq_laneq_high_u32): Ditto.
10541 (vfmlslq_laneq_high_u32): Ditto.
10542 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10543 (AARCH64_FL_FOR_ARCH8_4): New.
10544 (AARCH64_ISA_F16FML): New ISA flag.
10545 (TARGET_F16FML): New feature flag for fp16fml.
10546 (doc/invoke.texi): Document new fp16fml option.
10548 2018-01-10 Michael Collison <michael.collison@arm.com>
10550 * config/aarch64/aarch64-builtins.c:
10551 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10552 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10553 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10554 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10555 (AARCH64_ISA_SHA3): New ISA flag.
10556 (TARGET_SHA3): New feature flag for sha3.
10557 * config/aarch64/iterators.md (sha512_op): New int attribute.
10558 (CRYPTO_SHA512): New int iterator.
10559 (UNSPEC_SHA512H): New unspec.
10560 (UNSPEC_SHA512H2): Ditto.
10561 (UNSPEC_SHA512SU0): Ditto.
10562 (UNSPEC_SHA512SU1): Ditto.
10563 * config/aarch64/aarch64-simd-builtins.def
10564 (aarch64_crypto_sha512hqv2di): New builtin.
10565 (aarch64_crypto_sha512h2qv2di): Ditto.
10566 (aarch64_crypto_sha512su0qv2di): Ditto.
10567 (aarch64_crypto_sha512su1qv2di): Ditto.
10568 (aarch64_eor3qv8hi): Ditto.
10569 (aarch64_rax1qv2di): Ditto.
10570 (aarch64_xarqv2di): Ditto.
10571 (aarch64_bcaxqv8hi): Ditto.
10572 * config/aarch64/aarch64-simd.md:
10573 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10574 (aarch64_crypto_sha512su0qv2di): Ditto.
10575 (aarch64_crypto_sha512su1qv2di): Ditto.
10576 (aarch64_eor3qv8hi): Ditto.
10577 (aarch64_rax1qv2di): Ditto.
10578 (aarch64_xarqv2di): Ditto.
10579 (aarch64_bcaxqv8hi): Ditto.
10580 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10581 (vsha512h2q_u64): Ditto.
10582 (vsha512su0q_u64): Ditto.
10583 (vsha512su1q_u64): Ditto.
10584 (veor3q_u16): Ditto.
10585 (vrax1q_u64): Ditto.
10586 (vxarq_u64): Ditto.
10587 (vbcaxq_u16): Ditto.
10588 * config/arm/types.md (crypto_sha512): New type attribute.
10589 (crypto_sha3): Ditto.
10590 (doc/invoke.texi): Document new sha3 option.
10592 2018-01-10 Michael Collison <michael.collison@arm.com>
10594 * config/aarch64/aarch64-builtins.c:
10595 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10596 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10597 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10598 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10599 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10600 (AARCH64_ISA_SM4): New ISA flag.
10601 (TARGET_SM4): New feature flag for sm4.
10602 * config/aarch64/aarch64-simd-builtins.def
10603 (aarch64_sm3ss1qv4si): Ditto.
10604 (aarch64_sm3tt1aq4si): Ditto.
10605 (aarch64_sm3tt1bq4si): Ditto.
10606 (aarch64_sm3tt2aq4si): Ditto.
10607 (aarch64_sm3tt2bq4si): Ditto.
10608 (aarch64_sm3partw1qv4si): Ditto.
10609 (aarch64_sm3partw2qv4si): Ditto.
10610 (aarch64_sm4eqv4si): Ditto.
10611 (aarch64_sm4ekeyqv4si): Ditto.
10612 * config/aarch64/aarch64-simd.md:
10613 (aarch64_sm3ss1qv4si): Ditto.
10614 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10615 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10616 (aarch64_sm4eqv4si): Ditto.
10617 (aarch64_sm4ekeyqv4si): Ditto.
10618 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10619 (sm3part_op): Ditto.
10620 (CRYPTO_SM3TT): Ditto.
10621 (CRYPTO_SM3PART): Ditto.
10622 (UNSPEC_SM3SS1): New unspec.
10623 (UNSPEC_SM3TT1A): Ditto.
10624 (UNSPEC_SM3TT1B): Ditto.
10625 (UNSPEC_SM3TT2A): Ditto.
10626 (UNSPEC_SM3TT2B): Ditto.
10627 (UNSPEC_SM3PARTW1): Ditto.
10628 (UNSPEC_SM3PARTW2): Ditto.
10629 (UNSPEC_SM4E): Ditto.
10630 (UNSPEC_SM4EKEY): Ditto.
10631 * config/aarch64/constraints.md (Ui2): New constraint.
10632 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10633 * config/arm/types.md (crypto_sm3): New type attribute.
10634 (crypto_sm4): Ditto.
10635 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10636 (vsm3tt1aq_u32): Ditto.
10637 (vsm3tt1bq_u32): Ditto.
10638 (vsm3tt2aq_u32): Ditto.
10639 (vsm3tt2bq_u32): Ditto.
10640 (vsm3partw1q_u32): Ditto.
10641 (vsm3partw2q_u32): Ditto.
10642 (vsm4eq_u32): Ditto.
10643 (vsm4ekeyq_u32): Ditto.
10644 (doc/invoke.texi): Document new sm4 option.
10646 2018-01-10 Michael Collison <michael.collison@arm.com>
10648 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10649 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10650 (AARCH64_FL_FOR_ARCH8_4): New.
10651 (AARCH64_FL_V8_4): New flag.
10652 (doc/invoke.texi): Document new armv8.4-a option.
10654 2018-01-10 Michael Collison <michael.collison@arm.com>
10656 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10657 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10658 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10659 * config/aarch64/aarch64-option-extension.def: Add
10660 AARCH64_OPT_EXTENSION of 'sha2'.
10661 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10662 (crypto): Disable sha2 and aes if crypto disabled.
10663 (crypto): Enable aes and sha2 if enabled.
10664 (simd): Disable sha2 and aes if simd disabled.
10665 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10667 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10668 (TARGET_SHA2): New feature flag for sha2.
10669 (TARGET_AES): New feature flag for aes.
10670 * config/aarch64/aarch64-simd.md:
10671 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10672 conditional on TARGET_AES.
10673 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10674 (aarch64_crypto_sha1hsi): Make pattern conditional
10676 (aarch64_crypto_sha1hv4si): Ditto.
10677 (aarch64_be_crypto_sha1hv4si): Ditto.
10678 (aarch64_crypto_sha1su1v4si): Ditto.
10679 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10680 (aarch64_crypto_sha1su0v4si): Ditto.
10681 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10682 (aarch64_crypto_sha256su0v4si): Ditto.
10683 (aarch64_crypto_sha256su1v4si): Ditto.
10684 (doc/invoke.texi): Document new aes and sha2 options.
10686 2018-01-10 Martin Sebor <msebor@redhat.com>
10688 PR tree-optimization/83781
10689 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10692 2018-01-11 Martin Sebor <msebor@gmail.com>
10693 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10695 PR tree-optimization/83501
10696 PR tree-optimization/81703
10698 * tree-ssa-strlen.c (get_string_cst): Rename...
10699 (get_string_len): ...to this. Handle global constants.
10700 (handle_char_store): Adjust.
10702 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10703 Jim Wilson <jimw@sifive.com>
10705 * config/riscv/riscv-protos.h (riscv_output_return): New.
10706 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10707 (riscv_attribute_table, riscv_output_return),
10708 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10709 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10710 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10712 (riscv_expand_prologue): Add early return for naked function.
10713 (riscv_expand_epilogue): Likewise.
10714 (riscv_function_ok_for_sibcall): Return false for naked function.
10715 (riscv_set_current_function): New.
10716 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10717 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10718 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10719 * doc/extend.texi (RISC-V Function Attributes): New.
10721 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10723 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10724 check for 128-bit long double before checking TCmode.
10725 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10726 128-bit long doubles before checking TFmode or TCmode.
10727 (FLOAT128_IBM_P): Likewise.
10729 2018-01-10 Martin Sebor <msebor@redhat.com>
10731 PR tree-optimization/83671
10732 * builtins.c (c_strlen): Unconditionally return zero for the empty
10734 Use -Warray-bounds for warnings.
10735 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10736 for non-constant array indices with COMPONENT_REF, arrays of
10737 arrays, and pointers to arrays.
10738 (gimple_fold_builtin_strlen): Determine and set length range for
10739 non-constant character arrays.
10741 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10743 PR middle-end/81897
10744 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10747 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10749 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10751 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10754 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10755 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10756 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10757 indexed_or_indirect_operand predicate.
10758 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10759 (*vsx_le_perm_load_v8hi): Likewise.
10760 (*vsx_le_perm_load_v16qi): Likewise.
10761 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10762 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10763 (*vsx_le_perm_store_v8hi): Likewise.
10764 (*vsx_le_perm_store_v16qi): Likewise.
10765 (eight unnamed splitters): Likewise.
10767 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10769 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10770 * config/rs6000/emmintrin.h: Likewise.
10771 * config/rs6000/mmintrin.h: Likewise.
10772 * config/rs6000/xmmintrin.h: Likewise.
10774 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10777 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10779 * tree.c (tree_nop_conversion): Return true for location wrapper
10781 (maybe_wrap_with_location): New function.
10782 (selftest::check_strip_nops): New function.
10783 (selftest::test_location_wrappers): New function.
10784 (selftest::tree_c_tests): Call it.
10785 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10786 (maybe_wrap_with_location): New decl.
10787 (EXPR_LOCATION_WRAPPER_P): New macro.
10788 (location_wrapper_p): New inline function.
10789 (tree_strip_any_location_wrapper): New inline function.
10791 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10794 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10795 stack_realign_offset for the largest alignment of stack slot
10797 (ix86_find_max_used_stack_alignment): New function.
10798 (ix86_finalize_stack_frame_flags): Use it. Set
10799 max_used_stack_alignment if we don't realign stack.
10800 * config/i386/i386.h (machine_function): Add
10801 max_used_stack_alignment.
10803 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10805 * config/arm/arm.opt (-mbranch-cost): New option.
10806 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10809 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10812 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10813 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10815 2018-01-10 Richard Biener <rguenther@suse.de>
10818 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10819 early out so it also covers the case where we have a non-NULL
10822 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10824 PR tree-optimization/83753
10825 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10826 for non-strided grouped accesses if the number of elements is 1.
10828 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10831 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10832 * i386.h (TARGET_USE_GATHER): Define.
10833 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10835 2018-01-10 Martin Liska <mliska@suse.cz>
10838 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10839 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10841 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10842 CLEANUP_NO_PARTITIONING is not set.
10844 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10846 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10847 for vectors, as a partial revert of r254296.
10848 * rtl.h (const_vec_p): Delete.
10849 (const_vec_duplicate_p): Don't test for vector CONSTs.
10850 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10851 * expmed.c (make_tree): Likewise.
10854 * common.md (E, F): Use CONSTANT_P instead of checking for
10856 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10857 checking for CONST_VECTOR.
10859 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10861 PR middle-end/83575
10862 * predict.c (force_edge_cold): Handle in more sane way edges
10863 with no prediction.
10865 2018-01-09 Carl Love <cel@us.ibm.com>
10867 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10869 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10870 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10871 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10872 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10873 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10874 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10875 * config/rs6000/rs6000-protos.h: Add extern defition for
10876 rs6000_generate_float2_double_code.
10877 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10879 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10880 (float2_v2df): Add define_expand.
10882 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10885 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10886 op_mode in the force_to_mode call.
10888 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10890 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10891 instead of checking each element individually.
10892 (aarch64_evpc_uzp): Likewise.
10893 (aarch64_evpc_zip): Likewise.
10894 (aarch64_evpc_ext): Likewise.
10895 (aarch64_evpc_rev): Likewise.
10896 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10897 instead of checking each element individually. Return true without
10899 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10900 whether all selected elements come from the same input, instead of
10901 checking each element individually. Remove calls to gen_rtx_REG,
10902 start_sequence and end_sequence and instead assert that no rtl is
10905 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10907 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10908 order of HIGH and CONST checks.
10910 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10912 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10913 if the destination isn't an SSA_NAME.
10915 2018-01-09 Richard Biener <rguenther@suse.de>
10917 PR tree-optimization/83668
10918 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10920 (canonicalize_loop_form): ... here, renamed from ...
10921 (canonicalize_loop_closed_ssa_form): ... this and amended to
10922 swap successor edges for loop exit blocks to make us use
10923 the RPO order we need for initial schedule generation.
10925 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10927 PR tree-optimization/64811
10928 * match.pd: When optimizing comparisons with Inf, avoid
10929 introducing or losing exceptions from comparisons with NaN.
10931 2018-01-09 Martin Liska <mliska@suse.cz>
10934 * asan.c (shadow_mem_size): Add gcc_assert.
10936 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10938 Don't save registers in main().
10941 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10942 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10943 * config/avr/avr.c (avr_set_current_function): Don't error if
10944 naked, OS_task or OS_main are specified at the same time.
10945 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10947 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10949 * common/config/avr/avr-common.c (avr_option_optimization_table):
10950 Switch on -mmain-is-OS_task for optimizing compilations.
10952 2018-01-09 Richard Biener <rguenther@suse.de>
10954 PR tree-optimization/83572
10955 * graphite.c: Include cfganal.h.
10956 (graphite_transform_loops): Connect infinite loops to exit
10957 and remove fake edges at the end.
10959 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10961 * ipa-inline.c (edge_badness): Revert accidental checkin.
10963 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10966 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10967 symbols; not inline clones.
10969 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10972 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10973 hard registers. Formatting fixes.
10975 PR preprocessor/83722
10976 * gcc.c (try_generate_repro): Pass
10977 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10978 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10981 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10982 Kito Cheng <kito.cheng@gmail.com>
10984 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10985 (riscv_leaf_function_p): Delete.
10986 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10988 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10990 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10992 (do_ifelse): New function.
10993 (do_isel): New function.
10994 (do_sub3): New function.
10995 (do_add3): New function.
10996 (do_load_mask_compare): New function.
10997 (do_overlap_load_compare): New function.
10998 (expand_compare_loop): New function.
10999 (expand_block_compare): Call expand_compare_loop() when appropriate.
11000 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11001 option description.
11002 (-mblock-compare-inline-loop-limit): New option.
11004 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11007 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11008 Reverse order of second and third operands in first alternative.
11009 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11010 of first and second elements in UNSPEC_VPERMR vector.
11011 (altivec_expand_vec_perm_le): Likewise.
11013 2018-01-08 Jeff Law <law@redhat.com>
11015 PR rtl-optimizatin/81308
11016 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11017 (process_switch): If group_case_labels makes a change, then set
11019 (pass_convert_switch::execute): If a switch is converted, then
11020 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11022 PR rtl-optimization/81308
11023 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11026 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11028 PR target/83663 - Revert r255946
11029 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11030 generation for cases where splatting a value is not useful.
11031 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11032 across a vec_duplicate and a paradoxical subreg forming a vector
11033 mode to a vec_concat.
11035 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11037 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11038 -march=armv8.3-a variants.
11039 * config/arm/t-multilib: Likewise.
11040 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11042 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11044 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11046 (cceq_ior_compare_complement): Give it a name so I can use it, and
11047 change boolean_or_operator predicate to boolean_operator so it can
11048 be used to generate a crand.
11049 (eqne): New code iterator.
11050 (bd/bd_neg): New code_attrs.
11051 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11052 a single define_insn.
11053 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11054 decrement (bdnzt/bdnzf/bdzt/bdzf).
11055 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11056 with the new names of the branch decrement patterns, and added the
11057 names of the branch decrement conditional patterns.
11059 2018-01-08 Richard Biener <rguenther@suse.de>
11061 PR tree-optimization/83563
11062 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11065 2018-01-08 Richard Biener <rguenther@suse.de>
11067 PR middle-end/83713
11068 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11070 2018-01-08 Richard Biener <rguenther@suse.de>
11072 PR tree-optimization/83685
11073 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11074 references to abnormals.
11076 2018-01-08 Richard Biener <rguenther@suse.de>
11079 * dwarf2out.c (output_indirect_strings): Handle empty
11080 skeleton_debug_str_hash.
11081 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11083 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11085 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11086 (emit_store_direct): Likewise.
11087 (arc_trampoline_adjust_address): Likewise.
11088 (arc_asm_trampoline_template): New function.
11089 (arc_initialize_trampoline): Use asm_trampoline_template.
11090 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11091 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11092 * config/arc/arc.md (flush_icache): Delete pattern.
11094 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11096 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11097 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11100 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11103 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11104 by not USED_FOR_TARGET.
11105 (make_pass_resolve_sw_modes): Likewise.
11107 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11109 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11112 2018-01-08 Richard Biener <rguenther@suse.de>
11114 PR middle-end/83580
11115 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11117 2018-01-08 Richard Biener <rguenther@suse.de>
11119 PR middle-end/83517
11120 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11122 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11124 PR middle-end/81897
11125 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11126 basic blocks with a small number of successors.
11127 (convert_control_dep_chain_into_preds): Improve handling of
11129 (dump_predicates): Split apart into...
11130 (dump_pred_chain): ...here...
11131 (dump_pred_info): ...and here.
11132 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11133 (can_chain_union_be_invalidated_p): Improve check for invalidation
11135 (uninit_uses_cannot_happen): Avoid unnecessary if
11136 convert_control_dep_chain_into_preds yielded nothing.
11138 2018-01-06 Martin Sebor <msebor@redhat.com>
11140 PR tree-optimization/83640
11141 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11142 subtracting negative offset from size.
11143 (builtin_access::overlap): Adjust offset bounds of the access to fall
11144 within the size of the object if possible.
11146 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11148 PR rtl-optimization/83699
11149 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11150 extract_bit_field_as_subreg to cases in which the extracted
11151 value is also a vector.
11153 * lra-constraints.c (process_alt_operands): Test for the equivalence
11154 substitutions when detecting a possible reload cycle.
11156 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11159 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11160 by default if flag_selective_schedling{,2}. Formatting fixes.
11162 PR rtl-optimization/83682
11163 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11164 if it has non-VECTOR_MODE element mode.
11165 (vec_duplicate_p): Likewise.
11167 PR middle-end/83694
11168 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11169 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11171 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11174 * config/i386/i386-builtin.def
11175 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11176 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11177 Require also OPTION_MASK_ISA_AVX512F in addition to
11178 OPTION_MASK_ISA_GFNI.
11179 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11180 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11181 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11182 to OPTION_MASK_ISA_GFNI.
11183 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11184 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11185 OPTION_MASK_ISA_AVX512BW.
11186 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11187 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11188 addition to OPTION_MASK_ISA_GFNI.
11189 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11190 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11191 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11192 to OPTION_MASK_ISA_GFNI.
11193 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11194 a requirement for all ISAs rather than any of them with a few
11196 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11198 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11199 bitmasks to be enabled with 3 exceptions, instead of requiring any
11200 enabled ISA with lots of exceptions.
11201 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11202 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11203 Change avx512bw in isa attribute to avx512f.
11204 * config/i386/sgxintrin.h: Add license boilerplate.
11205 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11206 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11207 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11208 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11210 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11211 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11212 temporarily sse2 rather than sse if not enabled already.
11215 * config/i386/sse.md (VI248_VLBW): Rename to ...
11216 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11217 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11218 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11219 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11220 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11221 mode iterator instead of VI248_VLBW.
11223 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11225 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11226 (record_modified): Skip clobbers; add debug output.
11227 (param_change_prob): Use sreal frequencies.
11229 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11231 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11232 punt for user-aligned variables.
11234 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11236 * tree-chrec.c (chrec_contains_symbols): Return true for
11239 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11242 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11243 of (x|y) == x for BICS pattern.
11245 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11247 PR tree-optimization/83605
11248 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11249 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11252 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11254 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11255 * config/epiphany/rtems.h: New file.
11257 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11258 Uros Bizjak <ubizjak@gmail.com>
11261 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11262 QIreg_operand instead of register_operand predicate.
11263 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11264 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11265 comments instead of -fmitigate[-_]rop.
11267 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11270 * cgraphunit.c (symbol_table::compile): Switch to text_section
11271 before calling assembly_start debug hook.
11272 * run-rtl-passes.c (run_rtl_passes): Likewise.
11275 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11277 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11278 range_int_cst_p rather than !symbolic_range_p before calling
11279 extract_range_from_multiplicative_op_1.
11281 2018-01-04 Jeff Law <law@redhat.com>
11283 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11284 redundant test in assertion.
11286 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11288 * doc/rtl.texi: Document machine_mode wrapper classes.
11290 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11292 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11293 using tree_to_uhwi.
11295 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11297 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11298 the VEC_PERM_EXPR fold to fail.
11300 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11303 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11304 to switched_sections.
11306 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11309 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11310 test for d.testing.
11312 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11315 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11316 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11318 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11321 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11322 is BLKmode and bitpos not zero or mode change is needed.
11324 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11327 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11330 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11333 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11334 instead of MULT rtx. Update all corresponding splitters.
11335 (*saddl_se): Ditto.
11336 (*ssub<modesuffix>): Ditto.
11337 (*ssubl_se): Ditto.
11338 (*cmp_sadd_di): Update split patterns.
11339 (*cmp_sadd_si): Ditto.
11340 (*cmp_sadd_sidi): Ditto.
11341 (*cmp_ssub_di): Ditto.
11342 (*cmp_ssub_si): Ditto.
11343 (*cmp_ssub_sidi): Ditto.
11344 * config/alpha/predicates.md (const23_operand): New predicate.
11345 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11346 Look for ASHIFT, not MULT inner operand.
11347 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11349 2018-01-04 Martin Liska <mliska@suse.cz>
11351 PR gcov-profile/83669
11352 * gcov.c (output_intermediate_file): Add version to intermediate
11354 * doc/gcov.texi: Document new field 'version' in intermediate
11355 file format. Fix location of '-k' option of gcov command.
11357 2018-01-04 Martin Liska <mliska@suse.cz>
11360 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11362 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11364 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11366 2018-01-03 Martin Sebor <msebor@redhat.com>
11368 PR tree-optimization/83655
11369 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11370 checking calls with invalid arguments.
11372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11374 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11375 (vectorizable_mask_load_store): Delete.
11376 (vectorizable_call): Return false for masked loads and stores.
11377 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11378 instead of gimple_assign_rhs1.
11379 (vectorizable_load): Handle IFN_MASK_LOAD.
11380 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11384 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11386 (vectorizable_mask_load_store): ...here.
11387 (vectorizable_load): ...and here.
11389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11391 * tree-vect-stmts.c (vect_build_all_ones_mask)
11392 (vect_build_zero_merge_argument): New functions, split out from...
11393 (vectorizable_load): ...here.
11395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11397 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11399 (vectorizable_mask_load_store): ...here.
11400 (vectorizable_store): ...and here.
11402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11404 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11406 (vectorizable_mask_load_store): ...here.
11408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11410 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11411 (vect_model_store_cost): Take a vec_load_store_type instead of a
11413 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11414 (vect_model_store_cost): Take a vec_load_store_type instead of a
11416 (vectorizable_mask_load_store): Update accordingly.
11417 (vectorizable_store): Likewise.
11418 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11422 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11423 IFN_MASK_LOAD calls here rather than...
11424 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11427 Alan Hayward <alan.hayward@arm.com>
11428 David Sherwood <david.sherwood@arm.com>
11430 * expmed.c (extract_bit_field_1): For vector extracts,
11431 fall back to extract_bit_field_as_subreg if vec_extract
11434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11435 Alan Hayward <alan.hayward@arm.com>
11436 David Sherwood <david.sherwood@arm.com>
11438 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11439 they are variable or constant sized.
11440 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11441 slots for constant-sized data.
11443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11444 Alan Hayward <alan.hayward@arm.com>
11445 David Sherwood <david.sherwood@arm.com>
11447 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11448 handling COND_EXPRs with boolean comparisons, try to find a better
11449 basis for the mask type than the boolean itself.
11451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11453 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11454 is calculated and how it can be overridden.
11455 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11456 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11458 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11462 Alan Hayward <alan.hayward@arm.com>
11463 David Sherwood <david.sherwood@arm.com>
11465 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11466 Remove the mode argument.
11467 (aarch64_simd_valid_immediate): Remove the mode and inverse
11469 * config/aarch64/iterators.md (bitsize): New iterator.
11470 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11471 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11472 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11473 aarch64_simd_valid_immediate.
11474 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11475 (aarch64_reg_or_bic_imm): Likewise.
11476 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11477 with an insn_type enum and msl with a modifier_type enum.
11478 Replace element_width with a scalar_mode. Change the shift
11479 to unsigned int. Add constructors for scalar_float_mode and
11480 scalar_int_mode elements.
11481 (aarch64_vect_float_const_representable_p): Delete.
11482 (aarch64_can_const_movi_rtx_p)
11483 (aarch64_simd_scalar_immediate_valid_for_move)
11484 (aarch64_simd_make_constant): Update call to
11485 aarch64_simd_valid_immediate.
11486 (aarch64_advsimd_valid_immediate_hs): New function.
11487 (aarch64_advsimd_valid_immediate): Likewise.
11488 (aarch64_simd_valid_immediate): Remove mode and inverse
11489 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11490 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11491 and aarch64_float_const_representable_p on the result.
11492 (aarch64_output_simd_mov_immediate): Remove mode argument.
11493 Update call to aarch64_simd_valid_immediate and use of
11494 simd_immediate_info.
11495 (aarch64_output_scalar_simd_mov_immediate): Update call
11498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11499 Alan Hayward <alan.hayward@arm.com>
11500 David Sherwood <david.sherwood@arm.com>
11502 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11503 (mode_nunits): Likewise CONST_MODE_NUNITS.
11504 * machmode.def (ADJUST_NUNITS): Document.
11505 * genmodes.c (mode_data::need_nunits_adj): New field.
11506 (blank_mode): Update accordingly.
11507 (adj_nunits): New variable.
11508 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11510 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11511 listed in adj_nunits.
11512 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11513 listed in adj_nunits. Don't emit case statements for such modes.
11514 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11515 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11516 nothing if adj_nunits is nonnull.
11517 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11518 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11519 (emit_mode_fbit): Update use of print_maybe_const_decl.
11520 (emit_move_size): Likewise. Treat the array as non-const
11522 (emit_mode_adjustments): Handle adj_nunits.
11524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11526 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11527 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11528 (VECTOR_MODES): Use it.
11529 (make_vector_modes): Take the prefix as an argument.
11531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11532 Alan Hayward <alan.hayward@arm.com>
11533 David Sherwood <david.sherwood@arm.com>
11535 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11536 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11537 for MODE_VECTOR_BOOL.
11538 * machmode.def (VECTOR_BOOL_MODE): Document.
11539 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11540 (make_vector_bool_mode): New function.
11541 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11543 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11544 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11546 * stor-layout.c (int_mode_for_mode): Likewise.
11547 * tree.c (build_vector_type_for_mode): Likewise.
11548 * varasm.c (output_constant_pool_2): Likewise.
11549 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11550 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11551 for MODE_VECTOR_BOOL.
11552 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11553 of mode class checks.
11554 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11555 instead of a list of mode class checks.
11556 (expand_vector_scalar_condition): Likewise.
11557 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11560 Alan Hayward <alan.hayward@arm.com>
11561 David Sherwood <david.sherwood@arm.com>
11563 * machmode.h (mode_size): Change from unsigned short to
11565 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11566 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11567 or if measurement_type is not polynomial.
11568 (fixed_size_mode::includes_p): Check for constant-sized modes.
11569 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11570 return a poly_uint16 rather than an unsigned short.
11571 (emit_mode_size): Change the type of mode_size from unsigned short
11572 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11573 (emit_mode_adjustments): Cope with polynomial vector sizes.
11574 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11576 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11578 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11579 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11580 * caller-save.c (setup_save_areas): Likewise.
11581 (replace_reg_with_saved_mem): Likewise.
11582 * calls.c (emit_library_call_value_1): Likewise.
11583 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11584 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11585 (gen_lowpart_for_combine): Likewise.
11586 * convert.c (convert_to_integer_1): Likewise.
11587 * cse.c (equiv_constant, cse_insn): Likewise.
11588 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11589 (cselib_subst_to_values): Likewise.
11590 * dce.c (word_dce_process_block): Likewise.
11591 * df-problems.c (df_word_lr_mark_ref): Likewise.
11592 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11593 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11594 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11595 (rtl_for_decl_location): Likewise.
11596 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11597 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11598 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11599 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11600 (expand_expr_real_1): Likewise.
11601 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11602 (pad_below): Likewise.
11603 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11604 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11605 * ira.c (get_subreg_tracking_sizes): Likewise.
11606 * ira-build.c (ira_create_allocno_objects): Likewise.
11607 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11608 (ira_sort_regnos_for_alter_reg): Likewise.
11609 * ira-costs.c (record_operand_costs): Likewise.
11610 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11611 (resolve_simple_move): Likewise.
11612 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11613 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11614 (lra_constraints): Likewise.
11615 (CONST_POOL_OK_P): Reject variable-sized modes.
11616 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11617 (add_pseudo_to_slot, lra_spill): Likewise.
11618 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11619 * optabs-query.c (get_best_extraction_insn): Likewise.
11620 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11621 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11622 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11623 * recog.c (offsettable_address_addr_space_p): Likewise.
11624 * regcprop.c (maybe_mode_change): Likewise.
11625 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11626 * regrename.c (build_def_use): Likewise.
11627 * regstat.c (dump_reg_info): Likewise.
11628 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11629 (find_reloads, find_reloads_subreg_address): Likewise.
11630 * reload1.c (eliminate_regs_1): Likewise.
11631 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11632 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11633 (simplify_binary_operation_1, simplify_subreg): Likewise.
11634 * targhooks.c (default_function_arg_padding): Likewise.
11635 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11636 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11637 (verify_gimple_assign_ternary): Likewise.
11638 * tree-inline.c (estimate_move_cost): Likewise.
11639 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11640 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11641 (get_address_cost_ainc): Likewise.
11642 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11643 (vect_supportable_dr_alignment): Likewise.
11644 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11645 (vectorizable_reduction): Likewise.
11646 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11647 (vectorizable_operation, vectorizable_load): Likewise.
11648 * tree.c (build_same_sized_truth_vector_type): Likewise.
11649 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11650 * var-tracking.c (emit_note_insn_var_location): Likewise.
11651 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11652 (ADDR_VEC_ALIGN): Likewise.
11654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11655 Alan Hayward <alan.hayward@arm.com>
11656 David Sherwood <david.sherwood@arm.com>
11658 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11660 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11661 or if measurement_type is polynomial.
11662 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11663 * combine.c (make_extraction): Likewise.
11664 * dse.c (find_shift_sequence): Likewise.
11665 * dwarf2out.c (mem_loc_descriptor): Likewise.
11666 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11667 (extract_bit_field, extract_low_bits): Likewise.
11668 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11669 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11670 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11671 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11672 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11673 * reload.c (find_reloads): Likewise.
11674 * reload1.c (alter_reg): Likewise.
11675 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11676 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11677 * tree-if-conv.c (predicate_mem_writes): Likewise.
11678 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11679 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11680 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11681 * valtrack.c (dead_debug_insert_temp): Likewise.
11682 * varasm.c (mergeable_constant_section): Likewise.
11683 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11686 Alan Hayward <alan.hayward@arm.com>
11687 David Sherwood <david.sherwood@arm.com>
11689 * expr.c (expand_assignment): Cope with polynomial mode sizes
11690 when assigning to a CONCAT.
11692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11693 Alan Hayward <alan.hayward@arm.com>
11694 David Sherwood <david.sherwood@arm.com>
11696 * machmode.h (mode_precision): Change from unsigned short to
11698 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11700 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11701 or if measurement_type is not polynomial.
11702 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11703 in which the mode is already known to be a scalar_int_mode.
11704 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11705 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11707 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11708 for GET_MODE_PRECISION.
11709 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11710 for GET_MODE_PRECISION.
11711 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11713 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11714 (expand_field_assignment, make_extraction): Likewise.
11715 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11716 (get_last_value): Likewise.
11717 * convert.c (convert_to_integer_1): Likewise.
11718 * cse.c (cse_insn): Likewise.
11719 * expr.c (expand_expr_real_1): Likewise.
11720 * lra-constraints.c (simplify_operand_subreg): Likewise.
11721 * optabs-query.c (can_atomic_load_p): Likewise.
11722 * optabs.c (expand_atomic_load): Likewise.
11723 (expand_atomic_store): Likewise.
11724 * ree.c (combine_reaching_defs): Likewise.
11725 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11726 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11727 * tree.h (type_has_mode_precision_p): Likewise.
11728 * ubsan.c (instrument_si_overflow): Likewise.
11730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11731 Alan Hayward <alan.hayward@arm.com>
11732 David Sherwood <david.sherwood@arm.com>
11734 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11735 polynomial numbers of units.
11736 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11737 (valid_vector_subparts_p): New function.
11738 (build_vector_type): Remove temporary shim and take the number
11739 of units as a poly_uint64 rather than an int.
11740 (build_opaque_vector_type): Take the number of units as a
11741 poly_uint64 rather than an int.
11742 * tree.c (build_vector_from_ctor): Handle polynomial
11743 TYPE_VECTOR_SUBPARTS.
11744 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11745 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11746 (build_vector_from_val): If the number of units is variable,
11747 use build_vec_duplicate_cst for constant operands and
11748 VEC_DUPLICATE_EXPR otherwise.
11749 (make_vector_type): Remove temporary is_constant ().
11750 (build_vector_type, build_opaque_vector_type): Take the number of
11751 units as a poly_uint64 rather than an int.
11752 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11754 * cfgexpand.c (expand_debug_expr): Likewise.
11755 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11756 (store_constructor, expand_expr_real_1): Likewise.
11757 (const_scalar_mask_from_tree): Likewise.
11758 * fold-const-call.c (fold_const_reduction): Likewise.
11759 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11760 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11761 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11762 (fold_relational_const): Likewise.
11763 (native_interpret_vector): Likewise. Change the size from an
11764 int to an unsigned int.
11765 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11766 TYPE_VECTOR_SUBPARTS.
11767 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11768 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11769 duplicating a non-constant operand into a variable-length vector.
11770 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11771 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11772 * ipa-icf.c (sem_variable::equals): Likewise.
11773 * match.pd: Likewise.
11774 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11775 * print-tree.c (print_node): Likewise.
11776 * stor-layout.c (layout_type): Likewise.
11777 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11778 * tree-cfg.c (verify_gimple_comparison): Likewise.
11779 (verify_gimple_assign_binary): Likewise.
11780 (verify_gimple_assign_ternary): Likewise.
11781 (verify_gimple_assign_single): Likewise.
11782 * tree-pretty-print.c (dump_generic_node): Likewise.
11783 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11784 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11785 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11786 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11787 (vect_shift_permute_load_chain): Likewise.
11788 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11789 (expand_vector_condition, optimize_vector_constructor): Likewise.
11790 (lower_vec_perm, get_compute_type): Likewise.
11791 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11792 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11793 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11794 (vect_recog_mask_conversion_pattern): Likewise.
11795 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11796 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11797 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11798 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11799 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11800 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11801 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11802 (supportable_widening_operation): Likewise.
11803 (supportable_narrowing_operation): Likewise.
11804 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11806 * varasm.c (output_constant): Likewise.
11808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11809 Alan Hayward <alan.hayward@arm.com>
11810 David Sherwood <david.sherwood@arm.com>
11812 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11813 so that both the length == 3 and length != 3 cases set up their
11814 own permute vectors. Add comments explaining why we know the
11815 number of elements is constant.
11816 (vect_permute_load_chain): Likewise.
11818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11819 Alan Hayward <alan.hayward@arm.com>
11820 David Sherwood <david.sherwood@arm.com>
11822 * machmode.h (mode_nunits): Change from unsigned char to
11824 (ONLY_FIXED_SIZE_MODES): New macro.
11825 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11826 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11827 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11829 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11830 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11831 or if measurement_type is not polynomial.
11832 * genmodes.c (ZERO_COEFFS): New macro.
11833 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11835 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11836 Use ZERO_COEFFS when emitting initializers.
11837 * data-streamer.h (bp_pack_poly_value): New function.
11838 (bp_unpack_poly_value): Likewise.
11839 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11840 for GET_MODE_NUNITS.
11841 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11842 for GET_MODE_NUNITS.
11843 * tree.c (make_vector_type): Remove temporary shim and make
11844 the real function take the number of units as a poly_uint64
11845 rather than an int.
11846 (build_vector_type_for_mode): Handle polynomial nunits.
11847 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11848 * emit-rtl.c (const_vec_series_p_1): Likewise.
11849 (gen_rtx_CONST_VECTOR): Likewise.
11850 * fold-const.c (test_vec_duplicate_folding): Likewise.
11851 * genrecog.c (validate_pattern): Likewise.
11852 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11853 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11854 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11855 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11856 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11857 * rtlanal.c (subreg_get_info): Likewise.
11858 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11859 (vect_grouped_load_supported): Likewise.
11860 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11861 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11862 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11863 (simplify_const_unary_operation, simplify_binary_operation_1)
11864 (simplify_const_binary_operation, simplify_ternary_operation)
11865 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11866 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11867 instead of CONST_VECTOR_NUNITS.
11868 * varasm.c (output_constant_pool_2): Likewise.
11869 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11870 explicit-encoded elements in the XVEC for variable-length vectors.
11872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11874 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11877 Alan Hayward <alan.hayward@arm.com>
11878 David Sherwood <david.sherwood@arm.com>
11880 * coretypes.h (fixed_size_mode): Declare.
11881 (fixed_size_mode_pod): New typedef.
11882 * builtins.h (target_builtins::x_apply_args_mode)
11883 (target_builtins::x_apply_result_mode): Change type to
11884 fixed_size_mode_pod.
11885 * builtins.c (apply_args_size, apply_result_size, result_vector)
11886 (expand_builtin_apply_args_1, expand_builtin_apply)
11887 (expand_builtin_return): Update accordingly.
11889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11891 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11892 * cselib.c (cselib_hash_rtx): Likewise.
11893 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11894 CONST_VECTOR encoding.
11896 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11897 Jeff Law <law@redhat.com>
11900 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11901 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11902 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11903 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11906 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11907 explicitly probe *sp in a noreturn function if there were any callee
11908 register saves or frame pointer is needed.
11910 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11913 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11914 BLKmode for ternary, binary or unary expressions.
11917 * var-tracking.c (delete_vta_debug_insn): New inline function.
11918 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11919 insns from get_insns () to NULL instead of each bb separately.
11920 Use delete_vta_debug_insn. No longer static.
11921 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11922 delete_vta_debug_insns callers.
11923 * rtl.h (delete_vta_debug_insns): Declare.
11924 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11925 instead of variable_tracking_main.
11927 2018-01-03 Martin Sebor <msebor@redhat.com>
11929 PR tree-optimization/83603
11930 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11931 arguments past the endof the argument list in functions declared
11932 without a prototype.
11933 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11934 Avoid checking when arguments are null.
11936 2018-01-03 Martin Sebor <msebor@redhat.com>
11939 * doc/extend.texi (attribute const): Fix a typo.
11940 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11941 issuing -Wsuggest-attribute for void functions.
11943 2018-01-03 Martin Sebor <msebor@redhat.com>
11945 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11946 offset_int::from instead of wide_int::to_shwi.
11947 (maybe_diag_overlap): Remove assertion.
11948 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11949 * gimple-ssa-sprintf.c (format_directive): Same.
11950 (parse_directive): Same.
11951 (sprintf_dom_walker::compute_format_length): Same.
11952 (try_substitute_return_value): Same.
11954 2018-01-03 Jeff Law <law@redhat.com>
11956 PR middle-end/83654
11957 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11958 non-constant residual for zero at runtime and avoid probing in
11959 that case. Reorganize code for trailing problem to mirror handling
11962 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11964 PR tree-optimization/83501
11965 * tree-ssa-strlen.c (get_string_cst): New.
11966 (handle_char_store): Call get_string_cst.
11968 2018-01-03 Martin Liska <mliska@suse.cz>
11970 PR tree-optimization/83593
11971 * tree-ssa-strlen.c: Include tree-cfg.h.
11972 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11973 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11974 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11976 (strlen_dom_walker::before_dom_children): Call
11977 gimple_purge_dead_eh_edges. Dump tranformation with details
11979 (strlen_dom_walker::before_dom_children): Update call by adding
11980 new argument cleanup_eh.
11981 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11983 2018-01-03 Martin Liska <mliska@suse.cz>
11986 * cif-code.def (VARIADIC_THUNK): New enum value.
11987 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11990 2018-01-03 Jan Beulich <jbeulich@suse.com>
11992 * sse.md (mov<mode>_internal): Tighten condition for when to use
11993 vmovdqu<ssescalarsize> for TI and OI modes.
11995 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11997 Update copyright years.
11999 2018-01-03 Martin Liska <mliska@suse.cz>
12002 * ipa-visibility.c (function_and_variable_visibility): Skip
12003 functions with noipa attribure.
12005 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12007 * gcc.c (process_command): Update copyright notice dates.
12008 * gcov-dump.c (print_version): Ditto.
12009 * gcov.c (print_version): Ditto.
12010 * gcov-tool.c (print_version): Ditto.
12011 * gengtype.c (create_file): Ditto.
12012 * doc/cpp.texi: Bump @copying's copyright year.
12013 * doc/cppinternals.texi: Ditto.
12014 * doc/gcc.texi: Ditto.
12015 * doc/gccint.texi: Ditto.
12016 * doc/gcov.texi: Ditto.
12017 * doc/install.texi: Ditto.
12018 * doc/invoke.texi: Ditto.
12020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12022 * vector-builder.h (vector_builder::m_full_nelts): Change from
12023 unsigned int to poly_uint64.
12024 (vector_builder::full_nelts): Update prototype accordingly.
12025 (vector_builder::new_vector): Likewise.
12026 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12027 (vector_builder::operator ==): Likewise.
12028 (vector_builder::finalize): Likewise.
12029 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12030 Take the number of elements as a poly_uint64 rather than an
12032 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12033 from unsigned int to poly_uint64.
12034 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12035 (vec_perm_indices::new_vector): Likewise.
12036 (vec_perm_indices::length): Likewise.
12037 (vec_perm_indices::nelts_per_input): Likewise.
12038 (vec_perm_indices::input_nelts): Likewise.
12039 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12040 number of elements per input as a poly_uint64 rather than an
12041 unsigned int. Use the original encoding for variable-length
12042 vectors, rather than clamping each individual element.
12043 For the second and subsequent elements in each pattern,
12044 clamp the step and base before clamping their sum.
12045 (vec_perm_indices::series_p): Handle polynomial element counts.
12046 (vec_perm_indices::all_in_range_p): Likewise.
12047 (vec_perm_indices_to_tree): Likewise.
12048 (vec_perm_indices_to_rtx): Likewise.
12049 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12050 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12051 (tree_vector_builder::new_binary_operation): Handle polynomial
12052 element counts. Return false if we need to know the number
12053 of elements at compile time.
12054 * fold-const.c (fold_vec_perm): Punt if the number of elements
12055 isn't known at compile time.
12057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12059 * vec-perm-indices.h (vec_perm_builder): Change element type
12060 from HOST_WIDE_INT to poly_int64.
12061 (vec_perm_indices::element_type): Update accordingly.
12062 (vec_perm_indices::clamp): Handle polynomial element_types.
12063 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12064 (vec_perm_indices::all_in_range_p): Likewise.
12065 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12067 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12068 polynomial vec_perm_indices element types.
12069 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12070 * fold-const.c (fold_vec_perm): Likewise.
12071 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12072 * tree-vect-generic.c (lower_vec_perm): Likewise.
12073 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12074 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12075 element type to HOST_WIDE_INT.
12077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12078 Alan Hayward <alan.hayward@arm.com>
12079 David Sherwood <david.sherwood@arm.com>
12081 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12082 rather than an int. Use plus_constant.
12083 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12084 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12087 Alan Hayward <alan.hayward@arm.com>
12088 David Sherwood <david.sherwood@arm.com>
12090 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12091 a HOST_WIDE_INT to a poly_int64.
12093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12094 Alan Hayward <alan.hayward@arm.com>
12095 David Sherwood <david.sherwood@arm.com>
12097 * calls.c (load_register_parameters): Cope with polynomial
12098 mode sizes. Require a constant size for BLKmode parameters
12099 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12100 forces a parameter to be padded at the lsb end in order to
12101 fill a complete number of words, require the parameter size
12102 to be ordered wrt UNITS_PER_WORD.
12104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12105 Alan Hayward <alan.hayward@arm.com>
12106 David Sherwood <david.sherwood@arm.com>
12108 * reload1.c (spill_stack_slot_width): Change element type
12109 from unsigned int to poly_uint64_pod.
12110 (alter_reg): Treat mode sizes as polynomial.
12112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12113 Alan Hayward <alan.hayward@arm.com>
12114 David Sherwood <david.sherwood@arm.com>
12116 * reload.c (complex_word_subreg_p): New function.
12117 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
12124 operands whose sizes aren't ordered.
12125 (match_reload): Refer to this check here.
12127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12128 Alan Hayward <alan.hayward@arm.com>
12129 David Sherwood <david.sherwood@arm.com>
12131 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12132 that the mode size is in the set {1, 2, 4, 8, 16}.
12134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12135 Alan Hayward <alan.hayward@arm.com>
12136 David Sherwood <david.sherwood@arm.com>
12138 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12139 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12146 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12147 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12148 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12149 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12150 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12151 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12152 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12153 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12154 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12156 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12157 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12158 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12159 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12160 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12161 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12162 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12163 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12164 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12165 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12167 * expr.c (emit_move_resolve_push): Treat the input and result
12168 of PUSH_ROUNDING as a poly_int64.
12169 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12170 (emit_push_insn): Likewise.
12171 * lra-eliminations.c (mark_not_eliminable): Likewise.
12172 * recog.c (push_operand): Likewise.
12173 * reload1.c (elimination_effects): Likewise.
12174 * rtlanal.c (nonzero_bits1): Likewise.
12175 * calls.c (store_one_arg): Likewise. Require the padding to be
12176 known at compile time.
12178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12179 Alan Hayward <alan.hayward@arm.com>
12180 David Sherwood <david.sherwood@arm.com>
12182 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12183 Use plus_constant instead of gen_rtx_PLUS.
12185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12186 Alan Hayward <alan.hayward@arm.com>
12187 David Sherwood <david.sherwood@arm.com>
12189 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12190 rather than an int.
12192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12193 Alan Hayward <alan.hayward@arm.com>
12194 David Sherwood <david.sherwood@arm.com>
12196 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12197 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12198 via stack temporaries. Treat the mode size as polynomial too.
12200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12201 Alan Hayward <alan.hayward@arm.com>
12202 David Sherwood <david.sherwood@arm.com>
12204 * expr.c (expand_expr_real_2): When handling conversions involving
12205 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12206 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12207 as a poly_uint64 too.
12209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12210 Alan Hayward <alan.hayward@arm.com>
12211 David Sherwood <david.sherwood@arm.com>
12213 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12216 Alan Hayward <alan.hayward@arm.com>
12217 David Sherwood <david.sherwood@arm.com>
12219 * combine.c (can_change_dest_mode): Handle polynomial
12220 REGMODE_NATURAL_SIZE.
12221 * expmed.c (store_bit_field_1): Likewise.
12222 * expr.c (store_constructor): Likewise.
12223 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12224 and polynomial REGMODE_NATURAL_SIZE.
12225 (gen_lowpart_common): Likewise.
12226 * reginfo.c (record_subregs_of_mode): Likewise.
12227 * rtlanal.c (read_modify_subreg_p): 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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12234 numbers of elements.
12236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12237 Alan Hayward <alan.hayward@arm.com>
12238 David Sherwood <david.sherwood@arm.com>
12240 * match.pd: Cope with polynomial numbers of vector elements.
12242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12243 Alan Hayward <alan.hayward@arm.com>
12244 David Sherwood <david.sherwood@arm.com>
12246 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12247 in a POINTER_PLUS_EXPR.
12249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12250 Alan Hayward <alan.hayward@arm.com>
12251 David Sherwood <david.sherwood@arm.com>
12253 * omp-simd-clone.c (simd_clone_subparts): New function.
12254 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12255 (ipa_simd_modify_function_body): Likewise.
12257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12258 Alan Hayward <alan.hayward@arm.com>
12259 David Sherwood <david.sherwood@arm.com>
12261 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12262 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12263 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12264 (expand_vector_condition, vector_element): Likewise.
12265 (subparts_gt): New function.
12266 (get_compute_type): Use subparts_gt.
12267 (count_type_subparts): Delete.
12268 (expand_vector_operations_1): Use subparts_gt instead of
12269 count_type_subparts.
12271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12272 Alan Hayward <alan.hayward@arm.com>
12273 David Sherwood <david.sherwood@arm.com>
12275 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12276 (vect_compile_time_alias): ...this new function. Do the calculation
12277 on poly_ints rather than trees.
12278 (vect_prune_runtime_alias_test_list): Update call accordingly.
12280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12281 Alan Hayward <alan.hayward@arm.com>
12282 David Sherwood <david.sherwood@arm.com>
12284 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12286 (vect_schedule_slp_instance): Likewise.
12288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12289 Alan Hayward <alan.hayward@arm.com>
12290 David Sherwood <david.sherwood@arm.com>
12292 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12293 constant and extern definitions for variable-length vectors.
12294 (vect_get_constant_vectors): Note that the number of units
12295 is known to be constant.
12297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12298 Alan Hayward <alan.hayward@arm.com>
12299 David Sherwood <david.sherwood@arm.com>
12301 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12302 of units as polynomial. Choose between WIDE and NARROW based
12305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12306 Alan Hayward <alan.hayward@arm.com>
12307 David Sherwood <david.sherwood@arm.com>
12309 * tree-vect-stmts.c (simd_clone_subparts): New function.
12310 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12313 Alan Hayward <alan.hayward@arm.com>
12314 David Sherwood <david.sherwood@arm.com>
12316 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12317 vectors as polynomial. Use build_index_vector for
12318 IFN_GOMP_SIMD_LANE.
12320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12321 Alan Hayward <alan.hayward@arm.com>
12322 David Sherwood <david.sherwood@arm.com>
12324 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12325 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12326 for variable-length vectors.
12327 (vectorizable_mask_load_store): Treat the number of units as
12328 polynomial, asserting that it is constant if the condition has
12329 already been enforced.
12330 (vectorizable_store, vectorizable_load): Likewise.
12332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12333 Alan Hayward <alan.hayward@arm.com>
12334 David Sherwood <david.sherwood@arm.com>
12336 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12337 of units as polynomial. Punt if we can't tell at compile time
12338 which vector contains the final result.
12340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12341 Alan Hayward <alan.hayward@arm.com>
12342 David Sherwood <david.sherwood@arm.com>
12344 * tree-vect-loop.c (vectorizable_induction): Treat the number
12345 of units as polynomial. Punt on SLP inductions. Use an integer
12346 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12347 cast of such a series for variable-length floating-point
12350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12351 Alan Hayward <alan.hayward@arm.com>
12352 David Sherwood <david.sherwood@arm.com>
12354 * tree.h (build_index_vector): Declare.
12355 * tree.c (build_index_vector): New function.
12356 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12357 of units as polynomial, forcibly converting it to a constant if
12358 vectorizable_reduction has already enforced the condition.
12359 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12360 to create a {1,2,3,...} vector.
12361 (vectorizable_reduction): Treat the number of units as polynomial.
12362 Choose vectype_in based on the largest scalar element size rather
12363 than the smallest number of units. Enforce the restrictions
12366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12367 Alan Hayward <alan.hayward@arm.com>
12368 David Sherwood <david.sherwood@arm.com>
12370 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12371 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12378 * target.def (autovectorize_vector_sizes): Return the vector sizes
12379 by pointer, using vector_sizes rather than a bitmask.
12380 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12381 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12382 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12384 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12385 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12386 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12387 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12388 * omp-general.c (omp_max_vf): Likewise.
12389 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12390 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12391 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12392 * tree-vect-slp.c (vect_slp_bb): Likewise.
12393 * doc/tm.texi: Regenerate.
12394 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12396 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12397 the vector size as a poly_uint64 rather than an unsigned int.
12398 (current_vector_size): Change from an unsigned int to a poly_uint64.
12399 (get_vectype_for_scalar_type): Update accordingly.
12400 * tree.h (build_truth_vector_type): Take the size and number of
12401 units as a poly_uint64 rather than an unsigned int.
12402 (build_vector_type): Add a temporary overload that takes
12403 the number of units as a poly_uint64 rather than an unsigned int.
12404 * tree.c (make_vector_type): Likewise.
12405 (build_truth_vector_type): Take the number of units as a poly_uint64
12406 rather than an unsigned int.
12408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12409 Alan Hayward <alan.hayward@arm.com>
12410 David Sherwood <david.sherwood@arm.com>
12412 * target.def (get_mask_mode): Take the number of units and length
12413 as poly_uint64s rather than unsigned ints.
12414 * targhooks.h (default_get_mask_mode): Update accordingly.
12415 * targhooks.c (default_get_mask_mode): Likewise.
12416 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12417 * doc/tm.texi: Regenerate.
12419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12420 Alan Hayward <alan.hayward@arm.com>
12421 David Sherwood <david.sherwood@arm.com>
12423 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12424 * omp-general.c (omp_max_vf): Likewise.
12425 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12426 (expand_omp_simd): Handle polynomial safelen.
12427 * omp-low.c (omplow_simd_context): Add a default constructor.
12428 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12429 (lower_rec_simd_input_clauses): Update accordingly.
12430 (lower_rec_input_clauses): Likewise.
12432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12433 Alan Hayward <alan.hayward@arm.com>
12434 David Sherwood <david.sherwood@arm.com>
12436 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12437 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12438 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12439 (vect_analyze_slp_cost): Likewise.
12440 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12441 (vect_model_load_cost): Likewise.
12443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12444 Alan Hayward <alan.hayward@arm.com>
12445 David Sherwood <david.sherwood@arm.com>
12447 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12448 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12449 from an unsigned int * to a poly_uint64_pod *.
12450 (calculate_unrolling_factor): New function.
12451 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12454 Alan Hayward <alan.hayward@arm.com>
12455 David Sherwood <david.sherwood@arm.com>
12457 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12458 from an unsigned int to a poly_uint64.
12459 (_loop_vec_info::slp_unrolling_factor): Likewise.
12460 (_loop_vec_info::vectorization_factor): Change from an int
12462 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12463 (vect_get_num_vectors): New function.
12464 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12465 (vect_get_num_copies): Use vect_get_num_vectors.
12466 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12467 to an unsigned int *.
12468 (vect_analyze_data_refs): Change min_vf from an int * to a
12470 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12471 than an unsigned HOST_WIDE_INT.
12472 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12473 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12474 to an unsigned int *.
12475 (vect_analyze_data_ref_dependences): Likewise.
12476 (vect_compute_data_ref_alignment): Handle polynomial vf.
12477 (vect_enhance_data_refs_alignment): Likewise.
12478 (vect_prune_runtime_alias_test_list): Likewise.
12479 (vect_shift_permute_load_chain): Likewise.
12480 (vect_supportable_dr_alignment): Likewise.
12481 (dependence_distance_ge_vf): Take the vectorization factor as a
12482 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12483 (vect_analyze_data_refs): Change min_vf from an int * to a
12485 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12486 vfm1 as a poly_uint64 rather than an int. Make the same change
12487 for the returned bound_scalar.
12488 (vect_gen_vector_loop_niters): Handle polynomial vf.
12489 (vect_do_peeling): Likewise. Update call to
12490 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12491 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12493 * tree-vect-loop.c (vect_determine_vectorization_factor)
12494 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12495 (vect_get_known_peeling_cost): Likewise.
12496 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12497 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12498 (vect_transform_loop): Likewise. Use the lowest possible VF when
12499 updating the upper bounds of the loop.
12500 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12501 rather than an int.
12502 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12503 polynomial unroll factors.
12504 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12505 (vect_make_slp_decision): Likewise.
12506 (vect_supported_load_permutation_p): Likewise, and polynomial
12508 (vect_analyze_slp_cost): Handle polynomial vf.
12509 (vect_slp_analyze_node_operations): Likewise.
12510 (vect_slp_analyze_bb_1): Likewise.
12511 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12512 than an unsigned HOST_WIDE_INT.
12513 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12514 (vectorizable_load): Handle polynomial vf.
12515 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12517 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12520 Alan Hayward <alan.hayward@arm.com>
12521 David Sherwood <david.sherwood@arm.com>
12523 * match.pd: Handle bit operations involving three constants
12524 and try to fold one pair.
12526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12528 * tree-vect-loop-manip.c: Include gimple-fold.h.
12529 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12530 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12531 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12532 Add a path that uses a step of VF instead of 1, but disable it
12534 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12535 and niters_no_overflow parameters. Update calls to
12536 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12537 Create a new SSA name if the latter choses to use a ste other
12538 than zero, and return it via niters_vector_mult_vf_var.
12539 * tree-vect-loop.c (vect_transform_loop): Update calls to
12540 vect_do_peeling, vect_gen_vector_loop_niters and
12541 slpeel_make_loop_iterate_ntimes.
12542 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12543 (vect_gen_vector_loop_niters): Update declarations after above changes.
12545 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12547 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12548 128-bit round to integer instructions.
12549 (ceil<mode>2): Likewise.
12550 (btrunc<mode>2): Likewise.
12551 (round<mode>2): Likewise.
12553 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12555 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12556 unaligned VSX load/store on P8/P9.
12557 (expand_block_clear): Allow the use of unaligned VSX
12558 load/store on P8/P9.
12560 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12562 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12564 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12565 swap associated with both a load and a store.
12567 2018-01-02 Andrew Waterman <andrew@sifive.com>
12569 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12570 * config/riscv/riscv.md (clear_cache): Use it.
12572 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12574 * web.c: Remove out-of-date comment.
12576 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12578 * expr.c (fixup_args_size_notes): Check that any existing
12579 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12580 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12581 (emit_single_push_insn): ...here.
12583 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12585 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12586 (const_vector_encoded_nelts): New function.
12587 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12588 (const_vector_int_elt, const_vector_elt): Declare.
12589 * emit-rtl.c (const_vector_int_elt_1): New function.
12590 (const_vector_elt): Likewise.
12591 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12592 of CONST_VECTOR_ELT.
12594 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12596 * expr.c: Include rtx-vector-builder.h.
12597 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12598 directly on the tree encoding.
12599 (const_vector_from_tree): Likewise.
12600 * optabs.c: Include rtx-vector-builder.h.
12601 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12602 sequence of "u" values.
12603 * vec-perm-indices.c: Include rtx-vector-builder.h.
12604 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12605 directly on the vec_perm_indices encoding.
12607 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12609 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12610 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12611 * rtx-vector-builder.h: New file.
12612 * rtx-vector-builder.c: Likewise.
12613 * rtl.h (rtx_def::u2): Add a const_vector field.
12614 (CONST_VECTOR_NPATTERNS): New macro.
12615 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12616 (CONST_VECTOR_DUPLICATE_P): Likewise.
12617 (CONST_VECTOR_STEPPED_P): Likewise.
12618 (CONST_VECTOR_ENCODED_ELT): Likewise.
12619 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12620 (unwrap_const_vec_duplicate): Likewise.
12621 (const_vec_series_p): Check for a non-duplicated vector encoding.
12622 Say that the function only returns true for integer vectors.
12623 * emit-rtl.c: Include rtx-vector-builder.h.
12624 (gen_const_vec_duplicate_1): Delete.
12625 (gen_const_vector): Call gen_const_vec_duplicate instead of
12626 gen_const_vec_duplicate_1.
12627 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12628 (gen_const_vec_duplicate): Use rtx_vector_builder.
12629 (gen_const_vec_series): Likewise.
12630 (gen_rtx_CONST_VECTOR): Likewise.
12631 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12632 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12633 Build a new vector rather than modifying a CONST_VECTOR in-place.
12634 (handle_special_swappables): Update call accordingly.
12635 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12636 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12637 Build a new vector rather than modifying a CONST_VECTOR in-place.
12638 (handle_special_swappables): Update call accordingly.
12640 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12642 * simplify-rtx.c (simplify_const_binary_operation): Use
12643 CONST_VECTOR_ELT instead of XVECEXP.
12645 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12647 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12648 the selector elements to be different from the data elements
12649 if the selector is a VECTOR_CST.
12650 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12651 ssizetype for the selector.
12653 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12655 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12656 before testing each element individually.
12657 * tree-vect-generic.c (lower_vec_perm): Likewise.
12659 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12661 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12662 * selftest-run-tests.c (selftest::run_tests): Call it.
12663 * vector-builder.h (vector_builder::operator ==): New function.
12664 (vector_builder::operator !=): Likewise.
12665 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12666 (vec_perm_indices::all_from_input_p): New function.
12667 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12668 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12669 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12670 instead of reading the VECTOR_CST directly. Detect whether both
12671 vector inputs are the same before constructing the vec_perm_indices,
12672 and update the number of inputs argument accordingly. Use the
12673 utility functions added above. Only construct sel2 if we need to.
12675 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12677 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12678 the broadcast of the low byte.
12679 (expand_mult_highpart): Use an explicit encoding for the permutes.
12680 * optabs-query.c (can_mult_highpart_p): Likewise.
12681 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12682 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12683 (vectorizable_bswap): Likewise.
12684 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12685 explicit encoding for the power-of-2 permutes.
12686 (vect_permute_store_chain): Likewise.
12687 (vect_grouped_load_supported): Likewise.
12688 (vect_permute_load_chain): Likewise.
12690 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12692 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12693 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12694 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12695 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12696 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12697 (vect_gen_perm_mask_any): Likewise.
12699 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12701 * int-vector-builder.h: New file.
12702 * vec-perm-indices.h: Include int-vector-builder.h.
12703 (vec_perm_indices): Redefine as an int_vector_builder.
12704 (auto_vec_perm_indices): Delete.
12705 (vec_perm_builder): Redefine as a stand-alone class.
12706 (vec_perm_indices::vec_perm_indices): New function.
12707 (vec_perm_indices::clamp): Likewise.
12708 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12709 (vec_perm_indices::new_vector): New function.
12710 (vec_perm_indices::new_expanded_vector): Update for new
12711 vec_perm_indices class.
12712 (vec_perm_indices::rotate_inputs): New function.
12713 (vec_perm_indices::all_in_range_p): Operate directly on the
12714 encoded form, without computing elided elements.
12715 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12716 encoding. Update for new vec_perm_indices class.
12717 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12718 the given vec_perm_builder.
12719 (expand_vec_perm_var): Update vec_perm_builder constructor.
12720 (expand_mult_highpart): Use vec_perm_builder instead of
12721 auto_vec_perm_indices.
12722 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12723 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12724 or double series encoding as appropriate.
12725 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12726 vec_perm_indices instead of auto_vec_perm_indices.
12727 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12728 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12729 (vect_permute_store_chain): Likewise.
12730 (vect_grouped_load_supported): Likewise.
12731 (vect_permute_load_chain): Likewise.
12732 (vect_shift_permute_load_chain): Likewise.
12733 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12734 (vect_transform_slp_perm_load): Likewise.
12735 (vect_schedule_slp_instance): Likewise.
12736 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12737 (vectorizable_mask_load_store): Likewise.
12738 (vectorizable_bswap): Likewise.
12739 (vectorizable_store): Likewise.
12740 (vectorizable_load): Likewise.
12741 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12742 vec_perm_indices instead of auto_vec_perm_indices. Use
12743 tree_to_vec_perm_builder to read the vector from a tree.
12744 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12745 vec_perm_builder instead of a vec_perm_indices.
12746 (have_whole_vector_shift): Use vec_perm_builder and
12747 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12748 truncation to calc_vec_perm_mask_for_shift.
12749 (vect_create_epilog_for_reduction): Likewise.
12750 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12751 from auto_vec_perm_indices to vec_perm_indices.
12752 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12753 instead of changing individual elements.
12754 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12755 the vector in d.perm.
12756 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12757 from auto_vec_perm_indices to vec_perm_indices.
12758 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12759 instead of changing individual elements.
12760 (arm_vectorize_vec_perm_const): Use new_vector to install
12761 the vector in d.perm.
12762 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12763 Update vec_perm_builder constructor.
12764 (rs6000_expand_interleave): Likewise.
12765 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12766 (rs6000_expand_interleave): Likewise.
12768 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12770 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12771 to qimode could truncate the indices.
12772 * optabs.c (expand_vec_perm_var): Likewise.
12774 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12776 * Makefile.in (OBJS): Add vec-perm-indices.o.
12777 * vec-perm-indices.h: New file.
12778 * vec-perm-indices.c: Likewise.
12779 * target.h (vec_perm_indices): Replace with a forward class
12781 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12782 * optabs.h: Include vec-perm-indices.h.
12783 (expand_vec_perm): Delete.
12784 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12785 (expand_vec_perm_const): Declare.
12786 * target.def (vec_perm_const_ok): Replace with...
12787 (vec_perm_const): ...this new hook.
12788 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12789 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12790 * doc/tm.texi: Regenerate.
12791 * optabs.def (vec_perm_const): Delete.
12792 * doc/md.texi (vec_perm_const): Likewise.
12793 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12794 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12795 expand_vec_perm for constant permutation vectors. Assert that
12796 the mode of variable permutation vectors is the integer equivalent
12797 of the mode that is being permuted.
12798 * optabs-query.h (selector_fits_mode_p): Declare.
12799 * optabs-query.c: Include vec-perm-indices.h.
12800 (selector_fits_mode_p): New function.
12801 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12802 is defined, instead of checking whether the vec_perm_const_optab
12803 exists. Use targetm.vectorize.vec_perm_const instead of
12804 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12805 fit in the vector mode before using a variable permute.
12806 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12807 vec_perm_indices instead of an rtx.
12808 (expand_vec_perm): Replace with...
12809 (expand_vec_perm_const): ...this new function. Take the selector
12810 as a vec_perm_indices rather than an rtx. Also take the mode of
12811 the selector. Update call to shift_amt_for_vec_perm_mask.
12812 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12813 Use vec_perm_indices::new_expanded_vector to expand the original
12814 selector into bytes. Check whether the indices fit in the vector
12815 mode before using a variable permute.
12816 (expand_vec_perm_var): Make global.
12817 (expand_mult_highpart): Use expand_vec_perm_const.
12818 * fold-const.c: Includes vec-perm-indices.h.
12819 * tree-ssa-forwprop.c: Likewise.
12820 * tree-vect-data-refs.c: Likewise.
12821 * tree-vect-generic.c: Likewise.
12822 * tree-vect-loop.c: Likewise.
12823 * tree-vect-slp.c: Likewise.
12824 * tree-vect-stmts.c: Likewise.
12825 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12827 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12828 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12829 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12830 (aarch64_vectorize_vec_perm_const): ...this new function.
12831 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12832 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12833 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12834 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12835 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12836 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12837 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12839 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12840 check for NEON modes.
12841 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12842 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12843 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12844 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12846 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12847 the old VEC_PERM_CONST conditions.
12848 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12849 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12850 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12851 (ia64_vectorize_vec_perm_const_ok): Merge into...
12852 (ia64_vectorize_vec_perm_const): ...this new function.
12853 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12854 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12855 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12856 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12857 * config/mips/mips.c (mips_expand_vec_perm_const)
12858 (mips_vectorize_vec_perm_const_ok): Merge into...
12859 (mips_vectorize_vec_perm_const): ...this new function.
12860 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12861 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12862 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12863 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12864 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12865 (rs6000_expand_vec_perm_const): Delete.
12866 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12868 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12869 (altivec_expand_vec_perm_const_le): Take each operand individually.
12870 Operate on constant selectors rather than rtxes.
12871 (altivec_expand_vec_perm_const): Likewise. Update call to
12872 altivec_expand_vec_perm_const_le.
12873 (rs6000_expand_vec_perm_const): Delete.
12874 (rs6000_vectorize_vec_perm_const_ok): Delete.
12875 (rs6000_vectorize_vec_perm_const): New function.
12876 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12877 an element count and rtx array.
12878 (rs6000_expand_extract_even): Update call accordingly.
12879 (rs6000_expand_interleave): Likewise.
12880 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12881 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12882 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12883 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12884 (rs6000_expand_vec_perm_const): Delete.
12885 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12886 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12887 (altivec_expand_vec_perm_const_le): Take each operand individually.
12888 Operate on constant selectors rather than rtxes.
12889 (altivec_expand_vec_perm_const): Likewise. Update call to
12890 altivec_expand_vec_perm_const_le.
12891 (rs6000_expand_vec_perm_const): Delete.
12892 (rs6000_vectorize_vec_perm_const_ok): Delete.
12893 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12894 reference to the SPE evmerge intructions.
12895 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12896 an element count and rtx array.
12897 (rs6000_expand_extract_even): Update call accordingly.
12898 (rs6000_expand_interleave): Likewise.
12899 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12900 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12902 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12904 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12906 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12907 vector mode and that that mode matches the mode of the data
12909 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12910 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12911 directly using expand_vec_perm_1 when forcing selectors into
12913 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12915 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12917 * optabs-query.h (can_vec_perm_p): Delete.
12918 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12919 * optabs-query.c (can_vec_perm_p): Split into...
12920 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12921 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12922 particular selector is valid.
12923 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12924 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12925 (vect_grouped_load_supported): Likewise.
12926 (vect_shift_permute_load_chain): Likewise.
12927 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12928 (vect_transform_slp_perm_load): Likewise.
12929 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12930 (vectorizable_bswap): Likewise.
12931 (vect_gen_perm_mask_checked): Likewise.
12932 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12933 implementations of variable permutation vectors into account
12934 when deciding which selector to use.
12935 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12936 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12937 with a false third argument.
12938 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12939 to test whether the constant selector is valid and can_vec_perm_var_p
12940 to test whether a variable selector is valid.
12942 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12944 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12945 * optabs-query.c (can_vec_perm_p): Likewise.
12946 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12947 instead of vec_perm_indices.
12948 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12949 (vect_gen_perm_mask_checked): Likewise,
12950 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12951 (vect_gen_perm_mask_checked): Likewise,
12953 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12955 * optabs-query.h (qimode_for_vec_perm): Declare.
12956 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12957 (qimode_for_vec_perm): ...this new function.
12958 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12960 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12962 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12963 does not have a conditional at the top.
12965 2018-01-02 Richard Biener <rguenther@suse.de>
12967 * ipa-inline.c (big_speedup_p): Fix expression.
12969 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12972 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12975 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12979 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12980 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12981 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12982 cond_taken_branch_cost 3->4.
12984 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12986 PR tree-optimization/83581
12987 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12988 TODO_cleanup_cfg if any changes have been made.
12990 PR middle-end/83608
12991 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12992 convert_modes if target mode has the right side, but different mode
12995 PR middle-end/83609
12996 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12997 last argument when extracting from CONCAT. If either from_real or
12998 from_imag is NULL, use expansion through memory. If result is not
12999 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13000 the parts directly to inner mode, if even that fails, use expansion
13003 PR middle-end/83623
13004 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13005 check for bswap in mode rather than HImode and use that in expand_unop
13008 Copyright (C) 2018 Free Software Foundation, Inc.
13010 Copying and distribution of this file, with or without modification,
13011 are permitted in any medium without royalty provided the copyright
13012 notice and this notice are preserved.